On 15 May 2026, the Democratic Republic of the Congo declared its 17th Ebola outbreak. Within 48 hours, the World Health Organization declared a Public Health Emergency of International Concern. As of late May, more than 1,200 suspected and confirmed cases had been reported with over 260 deaths. The outbreak is caused by Bundibugyo virus, a rare Ebola strain for which no approved vaccine yet exists. Imported cases have been confirmed in Uganda, Germany, and the Czech Republic.

For GeoPoll, the news triggered an immediate question that has driven our work for the past twelve years: how do we collect reliable, representative data from communities that field teams cannot safely or easily reach, fast enough to inform a live response.

This article walks through GeoPoll’s published experience supporting responses to Ebola, COVID-19, cholera, and Mpox across Africa and Asia, and lays out what we offer partners now.

Where it began: West Africa, 2014

The 2014 to 2016 West Africa Ebola outbreak killed nearly 12,000 people across Guinea, Liberia, and Sierra Leone. It also became the moment that mobile data collection in humanitarian crises moved from promising idea to operational reality.

When the outbreak peaked, GeoPoll was finalising its SMS survey system in Liberia. As we documented in the Journal of Health Communication, that timing meant we could begin running surveys immediately. We did not have to build infrastructure from scratch in the middle of a crisis. The same is true today. Our platform, panel, and mobile network operator integrations are in place in the affected countries before the next outbreak begins.

What we did across the West Africa outbreak

In the years that followed, our SMS and CATI surveys in Liberia, Sierra Leone, and Guinea covered a range of programme questions. Several of these projects are documented in published case studies and peer-reviewed work:

• Food security tracking with the United Nations World Food Programme. Over three months in Sierra Leone, Liberia, and Guinea, we collected indicators on food prices, wages, and household coping. The work adapted the reduced Coping Strategies Index for mobile delivery, with prior validation showing no significant difference between mobile and face-to-face collection. Case study.
• Market functionality monitoring for the Famine Early Warning Systems Network. Panel-based SMS surveys with market traders in Sierra Leone and Liberia, tracking market sizes, operating costs, stock levels, and agricultural activity through ten rounds. Case study.
• Long-term economic impact surveys for the USAID Bureau for Africa and FHI360. Thirteen rounds of nationally stratified surveys in Liberia and Sierra Leone between January and June 2015, tracking income, employment, food prices, and schooling. Sample base of 1.8 million in Sierra Leone and 1.6 million in Liberia, with 1,000 completes per country per round. Case study.
• Health communications research with Johns Hopkins University in Liberia. SMS-based community dialogue and rumour tracking, supporting Ebola risk communication and community engagement. Documented in the academic literature.
• Community perceptions in Sierra Leone with Keystone Accountability. Assessing how the population viewed the international community’s response in real time.

Across the West Africa outbreak, GeoPoll reached more than 100,000 people. The methods worked because the people we surveyed already had access to mobile phones, our network operator integrations meant respondents incurred no cost to participate, and the SMS and voice modes did not require enumerators to enter quarantine zones or treatment areas.

What we learned

Three operational lessons from 2014 to 2016 still shape how we run surveys during outbreaks today:

• Keep surveys short. On SMS, response rates drop sharply beyond 12 to 15 questions. The constraint forces discipline on what we ask.
• Pre-code open-ended questions. 160-character limits and noisy environments mean structured response options outperform free text for most use cases.
• Always offer airtime credit on completion. Small incentives (we have typically used the local equivalent of about USD 0.50) significantly improve completion rates among low-income respondents.

Beyond West Africa: outbreaks in the DRC and the eastern corridor

Between 2018 and 2020, the DRC experienced two more large Ebola outbreaks in the eastern part of the country, primarily in North Kivu and Ituri. GeoPoll deployed mobile surveys during these outbreaks as well, focused on socio-economic impact and information flow. By the time we entered the COVID-19 era in 2020, we had effectively built a playbook for outbreak response work and applied it across an expanding set of geographies and health threats.

Our experience now spans the major health emergencies of the last decade:

• Ebola: Liberia, Sierra Leone, Guinea (2014 to 2016) and the DRC (2018 to 2020)
• COVID-19: 30+ countries across sub-Saharan Africa, the Middle East and North Africa, and Asia
• Cholera: Zambia (2024) and other African geographies
• Mpox: DRC, Burundi, Rwanda, Uganda, Central African Republic, and Kenya (2024)
• Other infectious disease and vaccine work: malaria, polio, measles, yellow fever, and routine immunisation studies across multiple African countries

COVID-19: when the playbook scaled

When COVID-19 reached sub-Saharan Africa in 2020, the methods we had refined for Ebola scaled up overnight. Between 2020 and 2022, GeoPoll ran self-funded and partner-funded research across more than 30 countries, covering economic impact, food security, vaccine acceptance, and risk communication. Findings from our November 2020 vaccine acceptance study across Cote d’Ivoire, the DRC, Kenya, Mozambique, Nigeria, and South Africa were archived publicly in ICPSR and used by researchers and policy makers globally.

We continued tracking vaccine perceptions across multiple rounds. The April 2021 follow-up, reported on the GeoPoll blog, found that fewer than half of respondents (48 percent) felt they had been given enough trustworthy information about the vaccine, a finding that mirrored what we were seeing on the ground.

The COVID work cemented two principles we now apply by default in outbreak research:

• Multi-mode is non-negotiable. SMS reaches the broadest base but limits depth. CATI handles longer instruments and complex skip logic. Mobile web reaches smartphone-heavy segments. In-person fills gaps for offline populations. The best outbreak studies combine modes by design, not as a fallback.
• Trust matters more than reach. A representative sample of people who refuse to answer honestly is not a sample. We invest in respondent identity verification, plain-language consent, and call-centre training in local languages because trust at the moment of the interview drives data quality.

Mpox: turning prior experience into rapid mobilisation

When mpox began spreading through Central and Eastern Africa in 2024, GeoPoll moved into the response within weeks. As we wrote at the time, the parallels with earlier outbreaks were clear: a disease moving faster than traditional surveillance, vaccine hesitancy reshaping its trajectory, and demand from public health partners and pharmaceutical companies for granular, real-time data.

Through late 2024 we ran mpox vaccine acceptance and behaviour monitoring rounds across six African countries: DRC, Burundi, Rwanda, Uganda, Central African Republic, and Kenya. The DRC mpox vaccine acceptance work has since been published in peer-reviewed medical literature and remains one of the largest mobile-based mpox studies on record from that period.

Cholera Zambia: a public-good data drop in the middle of a crisis

In early 2024, while Zambia was managing a cholera outbreak that had infected more than 21,000 people and caused over 700 deaths, GeoPoll ran a self-funded nationwide CATI survey to understand public awareness, water and sanitation access, and behaviour change. The findings were released as a public report on ReliefWeb with an interactive dashboard. The study used a stratified random sample of 400 respondents drawn from our Zambia panel, delivered in English, Bemba, and Nyanja from our Lusaka call centre.

The point of that work was not commercial. It was to demonstrate something that we believe matters more than any single study: in a crisis, the right response is to gather and share data quickly, even when there is no client paying for it.

What we offer partners responding to the 2026 outbreak

The capability that an organisation needs during an outbreak is not abstract. It is a short list of practical things, done quickly and well. Here is what we offer.

Mobile data collection across multiple modes

We run surveys through the channels respondents actually use. Most outbreak studies blend these by design:

• SMS surveys: Free-to-user via mobile network operator integrations. Best for broad reach, short instruments, and reaching low-income or rural populations. Used heavily in our Ebola, COVID, and cholera work.
• Computer Assisted Telephone Interviewing (CATI): Live calls from our call centres in Nairobi, Lusaka, Dar es Salaam, Johannesburg, and Panama City. Best for longer instruments, complex skip logic, sensitive topics, and qualitative depth.
• Mobile web (link-based): Surveys delivered via WhatsApp, SMS link, or other distribution. Best for smartphone-heavy segments, image-based questions, and longer self-completion.
• GeoPoll App: Our smartphone application supports longer panels and incentivised tracking studies.
• In-person interviewing: Where offline populations or sensitive observations are needed, we deploy trained field teams. Used selectively in our outbreak work, primarily for qualitative and validation purposes.

Reach across affected geographies

GeoPoll has more than 5 million profiled panelists and access to over 250 million individuals across 64 countries. In the geographies most relevant to the current Ebola outbreak, our panel and infrastructure are operational today:

• Democratic Republic of the Congo: active panel and call-centre capacity. French, Lingala, Kiswahili, and Kinande supported.
• Uganda: active panel, English and major local languages.
• Adjacent at-risk countries: Rwanda, Burundi, Tanzania, South Sudan, Central African Republic, and Kenya all have operational panels.

Speed when speed matters

Outbreak response cannot wait three months for fieldwork. Typical timelines for GeoPoll outbreak studies:

Methodology that holds up to scrutiny

Outbreak research is read by epidemiologists, donors, and ethics committees. Our default methodology is designed to pass that scrutiny:

• CDC-aligned KAP frameworks. We design knowledge, attitudes, and practice instruments to be compatible with established disease-response frameworks.
• Stratified random sampling. By gender, age, and geography. We report margins of error and confidence intervals consistently.
• IRB experience. We have participated in institutional review board processes with universities and research partners. Our research follows ESOMAR and WAPOR ethical standards.
• Transparent reporting. Every study reports its sample size, margin of error, languages, mode, and field period. We do not hide methodology.

Senselytic for real-time qualitative analysis

Outbreaks generate a lot of qualitative signal: open-ended responses, call-centre notes, social listening, focus group transcripts. Senselytic, our AI-powered qualitative analysis tool, helps partners extract patterns from this material in hours instead of weeks. We used it to support analysis on multi-country COVID and mpox studies, and it is a core capability for the current Ebola response.

Two ways partners can engage with us

For the current Bundibugyo outbreak, we are offering two complementary engagement options. They can stand alone or run in parallel:

1. Commissioned research

Bespoke studies designed around a single partner’s questions. Suitable when you have specific decision needs, geographic priorities, or contractual reporting requirements. Examples we are equipped to run today include vaccine acceptance and intent, risk communication effectiveness, healthcare-seeking behaviour, rumour and misinformation surveillance, food security and economic impact in affected zones, and case investigation support.

2. Ebola Outbreak Omnibus Survey

A shared, nationally representative DRC survey where multiple organisations contribute custom questions and receive their own answers plus common themes. Costs are shared, fielding is faster, and results are comparable across participating organisations. Suitable for partners who need data but do not require a full standalone study. A parallel Uganda omnibus will run if there is sufficient interest.

Get in Touch

Bundibugyo Ebola has no approved vaccine. The response will succeed or fail on case finding, contact tracing, risk communication, and community trust. All four depend on understanding what people in affected areas actually believe, know, fear, and need. That understanding cannot be assumed and it cannot be sampled from clinic registers alone. It has to be collected from people, in their own language, on a platform they already use.

GeoPoll has been collecting that kind of data through every major African outbreak of the last twelve years. The infrastructure is in place. The methodology is documented. The team is mobilised. We are ready to support partners working on this response, from public-good monitoring to bespoke programme evaluation, from rapid omnibus participation to long-term tracking studies.

In every outbreak we have worked on, the lesson has been the same: speed compounds. Decisions made on Day 7 with imperfect data are usually better than decisions made on Day 30 with perfect data. We are built to deliver on Day 7.

To learn more, discuss commissioned research, or to participate in the Ebola Outbreak Omnibus Survey, contact us.

The post A decade on the front line: what mobile data has taught us about responding to Ebola and other outbreaks appeared first on GeoPoll.

The term borrows from traditional survey research, where “enumeration” refers to the act of collecting data from respondents in the field, by phone, or through mobile channels. AI enumeration applies the same function to a new mode of delivery.

For research teams operating at scale across multiple languages and time zones, AI enumeration is one of the most significant methodological shifts since the move from face-to-face interviewing to computer-assisted telephone interviewing (CATI). But like any new method, it works well in some contexts and poorly in others, and understanding the difference is what separates useful adoption from expensive experimentation.

This guide covers what AI enumeration is, how it works, where it adds value, where it falls short, and why research expertise and verified respondent panels remain essential even as the interview itself becomes automated.

How AI enumeration works

At a mechanical level, AI enumeration systems combine three technologies: speech recognition to understand what the respondent says, a large language model to interpret meaning and generate follow-up questions, and text-to-speech to deliver questions in a natural voice.

The AI follows a structured questionnaire, just as a CATI interviewer would, but it can adapt within defined boundaries. If a respondent gives an unclear answer to an open-ended question, the AI can probe for clarification. If a respondent mentions something worth exploring, the AI can branch into a follow-up. And if the respondent speaks a different dialect or code-switches between languages, modern systems can often keep up.

The respondent experience varies. Some AI enumeration deployments use voice over the phone, mirroring traditional CATI. Others use voice through WhatsApp or messaging apps. A few use text-based chat interfaces. The common thread is that the interview feels like a conversation rather than a form.

AI enumeration versus traditional enumeration

Traditional enumeration relies on trained human interviewers. It is proven, flexible, and capable of handling almost any research context, but it is also expensive, slow to scale, and subject to variability between interviewers.

AI enumeration flips several of these tradeoffs. It scales almost instantly, runs consistently across thousands of interviews, and operates in any language the model supports, at any hour, without fatigue. What it gives up, at least for now, is the human judgment that skilled enumerators bring to difficult interviews: reading hesitation, building rapport with reluctant respondents, and knowing when to push and when to step back.

Neither method is universally better. The useful question is which method fits which study, and for many projects the answer is a thoughtful combination of both.

Advantages of AI enumeration

• Cost efficiency at scale. Human enumeration costs scale roughly linearly with sample size. AI enumeration has a higher fixed setup cost but much lower marginal cost per interview, which makes it economical for large samples, tracking studies, and high-frequency research. A study that would require hundreds of call center hours can often be completed in a fraction of the time at a fraction of the cost.
• Speed to field and speed to data. An AI enumerator can start interviews as soon as the questionnaire is approved and the sample is ready. There is no enumerator training, no briefing, no staffing up for peak periods. Fielding windows that used to take two to three weeks can close in days, and because the AI transcribes and codes as it goes, clean data is available almost immediately after the last interview completes.
• Consistency across interviews. Every respondent hears the same question in the same tone with the same phrasing. Interviewer effects, which are a real and often underdiscussed source of measurement error, are largely eliminated. This matters especially for tracking studies, where even small shifts in enumerator behavior between waves can create noise and bias that look like signals.
• Language and dialect coverage. Multilingual studies have traditionally required recruiting, training, and managing enumerators in each language. AI systems trained on sufficiently large speech datasets can handle dozens of languages, including low-resource languages that are difficult to staff for. This is a particularly meaningful advantage in regions like Sub-Saharan Africa, where a single national study might need to run in five or more languages.
• Respondent candor on sensitive topics. There is a growing body of evidence that respondents disclose more openly to AI interviewers on sensitive subjects, including health behaviors, financial status, political attitudes, and experiences of discrimination or violence. The absence of social judgment seems to reduce the performative element of responses that skews sensitive-topic data.
• 24/7 availability. AI enumerators do not have shifts. Respondents in rural areas who are only reachable in the evening, or business owners who can only talk after closing, can be interviewed whenever they are available. This expands the reachable universe and reduces the bias introduced by sampling only people who answer during call center hours.
• Scalability without quality degradation. In traditional enumeration, scaling a study often means hiring less experienced interviewers, which degrades quality at exactly the moment you need it most. AI enumeration holds quality constant regardless of sample size.

Drawbacks and considerations

• Rapport limits. Human enumerators build trust through small cues: warmth, acknowledgment, cultural references, shared language. AI systems are getting better at this, but they still struggle with the kind of rapport that gets a reluctant respondent to open up or a busy executive to stay on the line. For studies where participation depends on rapport, human enumeration is still the better choice.
• Complex probing and narrative elicitation. AI enumerators can probe effectively on structured open-ends, but they might fall short in deep narrative elicitation, especially when not well trained, where the interviewer needs to follow an unexpected thread, understand implicit meaning, or recognize when a respondent is circling back to something they have not yet said. Ethnographic and deeply qualitative work remains firmly in human territory.
• Respondent trust and consent. Respondents have a right to know they are speaking with an AI. Disclosure is both an ethical and, increasingly, a regulatory requirement. Studies need to handle this transparently without suppressing participation.
• Data security and model choice. AI enumeration involves sending the respondent’s speech to speech recognition and language models. The choice of models, where they are hosted, and how respondent data flows through the system are all material questions, particularly for studies involving vulnerable populations or regulated data.

Why research expertise still matters

AI enumeration automates the interview. It does not automate research.

Designing a study that yields valid, useful insights still requires methodological judgment: framing the research question, selecting the appropriate methodology, designing a questionnaire that avoids leading and double-barreled items, setting quotas that reflect population realities, defining weighting schemes that correct for known sample biases, and interpreting results in context. None of this is done by the AI.

If the questionnaire is poorly designed, an AI enumerator will execute it flawlessly and produce flawless garbage. If the sampling frame is biased, running the interviews through AI will produce precise estimates of the wrong quantity.

To get value from AI enumeration, researchers must pair it with genuine research expertise. If you treat AI enumeration as a replacement for research thinking, you will ship studies faster and be wrong faster.

Why a respondent database still matters

The second thing AI enumeration does not solve is the sample.

An AI enumerator needs someone to interview. That means a reachable, representative, profiled, and willing respondent base. Building such a base takes years and requires serious investment in recruitment, verification, profiling, re-engagement, and incentive management. It is not commodity infrastructure, and it cannot be conjured at the moment a study is commissioned.

In regions where traditional sampling frames are incomplete and where reaching specific demographic segments requires deliberate panel construction, the quality of the underlying respondent database largely determines the quality of any study run on top of it. An AI interviewer that calls the wrong people efficiently is not useful.

This is the pattern likely to play out across the industry: AI enumeration will become widely available, but the research buyers who get meaningful results will be the ones working with providers who own and actively maintain the respondent relationships the interviews depend on.

This is where organizations like GeoPoll, which has access to over 300 million mobile subscribers, come in. To provide a diverse enough sample to produce good research.

Best practices for AI-enumerated studies

• Pilot before you scale. Always run a pilot of at least 50 to 100 interviews before a full rollout. Listen to the recordings. Check the transcriptions. Identify the questions where respondents are confused, the probes that are not firing, and the moments where the AI misinterprets an answer. Fix before scaling.
• Design questionnaires for voice. Questionnaires that work on self-complete mobile surveys do not always work for voice. Long question stems, complex scales, and nested skip patterns that are fine for a human enumerator can confuse both the AI and the respondent. Shorter, cleaner, more conversational phrasing produces better results.
• Plan QA before fielding, not after. Decide in advance what proportion of interviews will be reviewed, what flags will trigger review, and who owns the review process. Budget time and cost for it.
• Use hybrid designs deliberately. AI for the scalable, structured portion of the study; human enumerators for the harder segments (rural, elderly, sensitive follow-ups, and qualitative deep dives). The best hybrid designs are intentional about which mode handles which respondent type.
• Be transparent with respondents. Disclose at the start that the interview is being conducted by an AI. Give respondents the option to decline. Respondents who participate under clear consent give more reliable data than those who feel tricked.
• Measure mode effects. If you are transitioning a tracking study from human CATI to AI enumeration, run a bridge study. Mode effects are real and measurable, and pretending they do not exist is how tracking data quietly loses its comparability.

Use cases for AI enumeration

• Large-scale tracking studies. Brand health, political opinion, consumer confidence, and public health tracking studies all benefit from AI enumeration’s consistency and cost efficiency, particularly when they run monthly or quarterly across multiple markets.
• Multilingual research in emerging markets. Studies that span multiple countries or multiple languages within a country, including African markets where staffing enumerators across five or more languages is a recurring operational challenge, can be run more cheaply and consistently with AI enumeration.
• Rapid-turnaround studies. Crisis response research, reaction studies around news events, and tight-deadline commercial studies all benefit from the speed advantages of AI fielding.
• Sensitive-topic research. Studies on health behaviors, financial vulnerability, gender-based violence, and political attitudes can produce more candid data through AI enumeration, though with strong ethical guardrails and clear pathways to human support where relevant.
• Panel recontact and longitudinal work. Reaching existing panel members for follow-up waves is operationally expensive with human enumerators. AI enumeration lowers the cost enough to make more frequent, lighter-touch recontact viable.
• Hard-to-reach schedules. Research with business owners, healthcare workers, farmers during harvest, or parents with young children requires flexibility that fixed call center hours cannot easily provide. AI enumeration’s always-on availability changes what is reachable.

Where AI enumeration is headed

AI enumeration will not replace human enumerators across the board. It will be for specific kinds of work, at specific scales, in specific contexts, while expanding the total volume of research that is economically viable. Integrating AI enumeration into a broader research offering rather than treating it as a standalone product is the current stance.

Powered by the ASR models we have been creating over the last few years using GeoPoll AI Data Streams, GeoPoll is currently running AI enumeration across our own survey platform.  Our focus on multilingual performance in Africa, Asia, and Latin America, and on the quality controls that make AI-collected data fit for client use.

If you are thinking about AI enumeration for your research project, or if you would like to discuss a pilot, get in touch with the GeoPoll team.

The post What is AI Enumeration? A Practitioner’s Guide to AI-Led Survey Interviews appeared first on GeoPoll.

Broadcast television and radio remain central to daily life in Tanzania, with the market shaped by distinctive features: a powerful sports-led TV ecosystem, a single evening peak that dwarfs every other viewing window, and a radio dial where three stations account for most of the national listening share.

GeoPoll’s Audience Measurement (GAM) data from January to March 2026 provides a picture of who is watching, who is listening, and when. This report distills the Q1 2026 numbers into the trends that matter for broadcasters, advertisers, and media planners operating in Tanzania.

As you read through this overview, remember that this is a high-level summary of national reach and share for Q1 2026. Weekly and monthly shifts, regional breakdowns, content-level performance, advertising effectiveness studies, and hourly unique-audience analysis are not covered here. For deeper insights, please reach out.

Top TV Stations in Tanzania

Television in Tanzania is a mass-reach medium, and the competitive landscape at the top is tight. The Top 5 stations are separated by only four percentage points on reach, with Azam Two leading by a narrow margin over ITV, the two Azam Sports channels, and TBC1.

Top 10 TV Stations by Reach (January – March 2026)

One of the most striking features of the Tanzanian list is how much of it belongs to a single broadcaster group. Azam Media occupies three of the Top 5 positions on reach (Azam Two, Azam Sports 1, Azam Sports 2) and a fourth in the Top 10. This cluster gives the group an unusually strong position in any national reach calculation and a near-unavoidable presence in sports-led media plans.

TV Viewership by Dayparts

Reach tells us who tuned in over the quarter. Daypart share of viewing tells us when each station was winning. The numbers below represent each station’s share of total Top 10 viewing within the time block.

• Early Morning (6:00 AM – 9:00 AM): Breakfast TV is the most fragmented block of the day. TBC1 leads with 16% share of Top 10 morning viewing, with ITV (15%), Clouds (12%), and Azam Two (11%) close behind. Seven different stations hold between 8% and 16%, which means morning audiences in Tanzania are genuinely split across the dial rather than concentrated on one or two options.
• Afternoon (4:00 PM – 6:00 PM): This is where the Azam Sports effect becomes very clear. Azam Sports 1 commands 37% of Top 10 viewing in this block, and when combined with Azam Sports 2 and SuperSport, sports channels take more than 60% of all Top 10 afternoon viewing. Live football and pre-match programming drive this concentration, and it is one of the most commercially distinct dayparts in the Tanzanian market.
• Prime-Time (7:00 PM – 10:00 PM): This is the single most important commercial block of the day, and it is dominated by three names. Azam Sports 1 takes 20% share, followed by ITV (16%) and Azam Sports 2 (15%). The Azam cluster combined (including Azam Two at 12%) takes roughly 47% of all Top 10 prime-time viewing, which is one of the most concentrated prime-time positions of any East African market.

Weekday Prime-Time Share of Viewing (7:00 PM – 10:00 PM)

Late Night (10:00 PM – Midnight): The sports channels extend their lead into late-night viewing. Azam Sports 1 (29%) and Azam Sports 2 (20%) together account for nearly half of all late-night Top 10 viewing, with Azam Two adding another 16%. Late-night in Tanzania is, effectively, a sports block.

Weekend and Special Programming Trends

Weekends amplify what is already true about weekday viewing in Tanzania: the evening peak becomes even sharper, and the sports channels maintain their structural advantage across the day.

• ITV leads weekend prime-time with 22% share, reflecting the pull of local drama and entertainment programming on Saturday and Sunday evenings.
• The 8:00 PM weekend half-hour is by far the largest viewing window in the week, with total Top 10 audience roughly 1.6× the next-largest weekend half-hour and well above any single weekday moment.
• Morning reach is significantly larger on weekends, with the 6:30 AM block drawing audiences that rival mid-afternoon weekday numbers. This suggests weekend morning content (children’s programming, religious broadcasts, magazine shows) remains a meaningful commercial window.

Top Radio Stations in Tanzania

Radio in Tanzania is concentrated at the top to a degree that is rare in the region. Three stations account for the bulk of national listening, and the drop-off from third to fourth place is the steepest on the chart.

Top 10 Radio Stations by Reach (January – March 2026)

TBC Taifa Radio leads at 47%, followed by Clouds (44%) and Radio Free Africa (38%). The gap to the fourth station, Wasafi FM, is more than 12 percentage points. For national advertisers, this concentration means that the top three stations are effectively must-include for any broad-reach Tanzanian radio plan.

Radio Listenership by Dayparts

Breakfast (6:00 AM – 9:00 AM): The breakfast block is dominated by the top three. TBC Taifa Radio leads with 26% share of Top 10 morning listening, followed by Radio Free Africa (22%) and Clouds (18%). Together, these three stations account for roughly two-thirds of all Top 10 breakfast listening on weekdays.

Weekday Breakfast Share of Listening (6:00 AM – 9:00 AM)

Daytime (9:00 AM – 4:00 PM): Listening levels dip after breakfast but stabilise through the middle of the day, with Clouds holding a particularly strong position in the late-morning and lunchtime blocks. Radio Free Africa also sustains strong midday listening, reflecting loyal audiences who stay tuned through the workday.

Drive-Time (4:00 PM – 7:00 PM): The afternoon drive block tightens the race. TBC Taifa Radio (24%) holds a narrow lead over Clouds (22%) and Radio Free Africa (20%), with Wasafi FM (10%) a clear fourth. The top three are separated by less than five percentage points during drive-time, making it one of the most competitive listening windows of the day.

Evening (7:00 PM – 10:00 PM): The evening sees TBC Taifa Radio and Clouds move essentially level at roughly 26% share each, with Radio Free Africa dropping back to 12%. This is also where Wasafi FM’s 9:00 PM spike shows up, reflecting its entertainment-led evening programming pulling a distinct audience segment.

Weekend Radio Trends

Weekend listening follows a similar shape to weekdays but with a flatter distribution across the day. TBC Taifa Radio remains the weekend leader at breakfast, and Radio Free Africa’s weekend breakfast position strengthens, reflecting the role of weekend talk and religious programming.

Insights from the Q1 2026 Tanzania Media Numbers

A few structural themes emerge from the Tanzanian data:

1. Prime-time has a single peak, and it is unusually sharp. The 8:00 PM half-hour on weekdays is more than 1.5× the size of any other weekday moment, and on weekends it is even larger. Unlike markets with multiple evening peaks, Tanzania concentrates its evening audience into one narrow window. This is an important peak for media plans.
2. Sports is a structural pillar of Tanzanian TV, not a niche category. Azam Sports 1 alone takes more than a third of afternoon viewing and a fifth of prime-time viewing. Any national media plan that excludes sports inventory is effectively excluding a major portion of the available TV audience.
3. Breakfast is bigger than drive-time on radio, by a wide margin. Total Top 10 radio listening at 6:00 AM on weekdays is roughly 4× the level at 5:00 PM. This is a common pattern, but Tanzania’s morning peak is especially concentrated into the 6:00 – 7:00 AM hour. For advertisers, the morning hour is the single most valuable radio block of the day.
4. Radio listening is more concentrated than TV viewing. Three stations account for the majority of national radio listening across every daypart, while TV’s Top 10 is more evenly distributed. Advertisers can build meaningful TV reach across five or six channels, but radio reach in Tanzania requires the top three.
5. Cluster-brand effects matter more here than elsewhere. The Azam cluster’s combined share of prime-time TV viewing (around 47%) is a pattern that does not appear in most other East African markets. For competing broadcasters, advertisers, and media regulators, the shape of the market is not just about individual station performance but about group-level concentration.

GeoPoll Audience Measurement

GeoPoll provides real-time, data-driven insights into media consumption in Tanzania and across more than 120 countries. Whether you need daily audience measurement, on-demand analytics, a campaign effectiveness study, or a custom report covering a specific audience segment or time period, GeoPoll Audience Measurement gives broadcasters, advertisers, and media planners the tools to make informed decisions.

How to access more insights:

• Subscribe for daily TV and radio ratings to track audience behavior in real time
• Request a custom report for any time period or audience segment
• Measure advertising impact and campaign performance
• Analyze viewership trends by content type, daypart, and broadcaster

Get in touch today!

The post The Top TV and Radio Stations in Tanzania – Q1 2026 appeared first on GeoPoll.

GeoPoll’s Audience Measurement (GAM) data from January to March 2026 gives us a clear picture of who is watching, who is listening, and when. This report distills the Q1 2026 numbers into the trends that matter for broadcasters, advertisers, and media planners.

As you read through this overview, remember that this is a high-level summary of national reach and share for Q1 2026. Weekly and monthly shifts, regional breakdowns, content-level performance, advertising effectiveness studies, hourly unique-audience analysis, demographics, and actual numbers are not covered here. For deeper insights, please reach out.

Top TV Stations in Kenya

Television remains a mass-reach medium in Kenya, and the top of the table in Q1 2026 looks familiar. Citizen TV held on to the number one position, reaching three in every four Kenyan TV viewers over the quarter. NTV, KTN, SuperSport, and TV 47 rounded out the Top 5, with SuperSport’s continued climb into the top tier showing both the strength of live sport and the appetite for premium football programming.

Top 10 TV Stations by Reach (January – March 2026)

A few things stand out when compared to our previous report. SuperSport has climbed into the Top 4 on the back of a packed sporting calendar. TV 47 continues to consolidate its position as a go-to news and current affairs option, and Al Jazeera’s appearance in the Top 10 reflects strong interest in international news coverage during an eventful first quarter, especially with the Iran-Israel-US conflict.

TV Viewership by Dayparts

Reach tells us who tuned in over the quarter. Daypart share of viewing tells us when each station was winning. The numbers below represent each station’s share of total Top 10 viewing within the time block.

Early Morning (6:00 AM – 9:00 AM): Breakfast TV belongs to Citizen TV, which captures roughly 26% of Top 10 viewing during the morning block. NTV follows at 11.5%, with TV 47, KTN, and K24 clustered closely as Kenyans catch up with morning bulletins and talk shows before heading out.

Mid-Morning to Afternoon (9:00 AM – 6:00 PM): The daytime block is fragmented. Citizen TV still leads, but Maisha Magic East makes a strong showing with local drama and soap operas pulling in at-home audiences, while Inooro TV sustains a loyal vernacular following.

Prime-Time (7:00 PM – 9:30 PM): This remains the most competitive and most commercially valuable block of the day. Citizen TV commands 35.5% of Top 10 prime-time viewing on weekdays, powered by the 7:00 PM Swahili news and the programming that follows. NTV (13.4%) is the clear number two, followed by Maisha Magic East (8.9%), KTN (8.5%), SuperSport (7.3%), and TV 47 (6.9%).

Weekday Prime-Time Share of Viewing (7:00 PM – 9:30 PM)

Late Night (10:00 PM – Midnight): This is where the ranking changes most dramatically. SuperSport takes 37% of late-night viewing on weekdays, often more than any other station, as football fans stay up for European fixtures. Citizen TV holds second place at about 21%, with Maisha Magic East third.

Weekend and Special Programming Trends

Weekends shuffle the deck in favour of sport and family programming.

• SuperSport’s share of weekend prime-time viewing climbs to 16.9%, nearly double its weekday prime-time share, driven by weekend football and other live sports.
• In the weekend afternoon block (2:00 PM – 6:30 PM), SuperSport and Citizen TV run neck and neck, each capturing roughly 22–23% of Top 10 viewing, a pattern that reflects sports coverage running alongside family viewing.
• Maisha Magic East maintains a strong weekend footprint with local drama.
• Citizen TV and NTV continue to anchor weekend news bulletins and current affairs panels.

Top Radio Stations in Kenya

Radio continues to be remarkably resilient in Kenya. Despite the trends around podcasts and social audio, nearly two in three Kenyan radio listeners tuned into Citizen Radio in Q1 2026, with a dense pack of competitors close behind.

Top 10 Radio Stations by Reach (January – March 2026)

The top four stations are separated by a narrow margin, and all of them recorded their peak audiences at 7:00 AM – confirming that breakfast radio remains the most valuable block on the dial as listeners commute to work. Radio Jambo, interestingly, is the only Top 10 station whose peak hour shifts to 8:00 AM, pointing to a later-starting commuter audience, probably driven by the popular Patanisho segment.

Radio Listenership by Dayparts

Breakfast (6:00 AM – 9:00 AM): The breakfast block is a four-station contest. Classic 105 leads with 18.8% of Top 10 morning listening on weekdays, narrowly edging Citizen (17.4%), Radio Jambo (15.0%), and Radio Maisha (12.0%). Lifestyle and political talk shows drive this block, and it is where brands typically invest most heavily.

Weekday Breakfast Share of Listening (6:00 AM – 9:00 AM)

Daytime (9:00 AM – 4:00 PM): Listening levels dip compared to breakfast, but Citizen and Classic 105 continue to lead. Kameme shows particular strength in the mid-morning block, reflecting its loyal Central Kenya audience staying tuned through the day.

Drive-Time (4:00 PM – 7:00 PM): Citizen pulls ahead in the afternoon drive block with 20.4% share of Top 10 listening, followed by Classic 105 (14.5%), Radio Jambo (14.1%), and Radio Maisha (11.9%).

Evening (7:00 PM – 10:00 PM): Citizen’s lead widens further in the evening to 26.1%, with Classic 105, Radio Maisha, and Radio Jambo filling out the top four.

Weekend Radio Trends

The weekend pattern mirrors weekdays at the top, but with a gentler peak and more even distribution across the day. Citizen (19.3%) and Classic 105 (18.1%) are essentially tied for weekend breakfast leadership, while Radio Jambo and Radio Maisha maintain their strong third and fourth positions.

What the Q1 2026 Numbers Tell Us

A few themes emerge from the data:

1. The gap behind the leaders is narrower than the top line suggests. Citizen TV and Citizen Radio lead their respective rankings. It is worth noting that in several dayparts, the second and third-place stations are within single-digit share points, meaning daypart selection matters as much as station selection for most campaigns for media planners.

2. Sport is now a Top 10 force on television. SuperSport’s weekend prime-time share (16.9%) and late-night dominance (37%) signal that live sports are among the few categories consistently pulling Kenyan viewers back to linear TV.
3. Breakfast remains the most contested slot on radio. Four stations are within a seven-point range at breakfast, and any share shift during this block has outsized commercial implications.
4. Vernacular and international stations have real weight. Inooro TV, Kameme, and Al Jazeera all made meaningful appearances in national rankings and in specific dayparts, a reminder that Kenya’s media market is more nuanced than the top line suggests.

GeoPoll Audience Measurement

GeoPoll provides real-time, data-driven insights into media consumption in Kenya and across more than 120 countries. Whether you need daily audience measurement, on-demand analytics, a campaign effectiveness study, or a custom report covering a specific audience segment or time period, GeoPoll Audience Measurement gives broadcasters, advertisers, and media planners the tools to make informed decisions.

How to access more media insights for Kenya:

• Subscribe for daily TV and radio ratings to track audience behavior in real time
• Request a custom report for any time period or audience segment
• Measure advertising impact and campaign performance
• Analyze viewership trends by content type, daypart, and broadcaster

Get in touch.

The post The Top TV and Radio Stations in Kenya – Q1 2026 appeared first on GeoPoll.

The Iran–Israel–United States conflict has been reshaping global energy markets since late 2025. For citizens in the Global South, the consequences are neither abstract nor distant. In Pakistan, the government implemented a historic Rs 55 per litre fuel price increase on 6 March 2026. In Kenya, the Energy and Petroleum Regulatory Authority (EPRA) announced on 14 April 2026 the largest fuel price adjustment in over 21 years of regulatory records — a KSh 28.69 per litre increase for petrol and KSh 40.30 for diesel, effective 15 April. In Egypt, subsidised fuel prices were revised upward for the third time in twelve months. In South Africa, the inland price of 95-octane petrol is set to breach some of the highest prices ever seen in the country . These are not coincidences. They are the downstream effects of a single geopolitical shock.

In early March 2026, GeoPoll surveyed 3,754 citizens across Egypt, Kenya, Nigeria, Pakistan, Saudi Arabia, and South Africa as part of our Caught in the Crossfire? citizen perceptions study. Among the study’s most striking findings: the economic dimension of the conflict is being felt acutely and immediately, with fuel prices at the centre of public concern.

Across the six-country sample, 70% of respondents report that the conflict has affected fuel prices in their country, with 42% characterising the impact as significant. The finding is consistent across diverse economic contexts – from oil-importing economies such as Pakistan and Kenya to the oil-exporting economy of Saudi Arabia, where 46% still report an impact.

The variation across countries reflects both the degree of energy dependence and the extent of government intervention. Pakistan, where the government passed through the full cost of disrupted imports, registers the highest impact at 85%. Saudi Arabia, which benefits from domestic production and price controls, registers the lowest at 46% – though this figure is notable in itself for a major oil producer.

Respondents Reporting or expecting Fuel Price Impact by Country

Country Analysis: The Fuel Crisis on the Ground

Pakistan: The Highest Impact in the Dataset

Pakistan registers the most severe fuel price impact of any country in the study, with 85% of respondents reporting or expecting an effect. The finding is consistent with on-the-ground realities: on 6 March 2026, the government implemented a Rs 55 per litre fuel price increase – among the largest single adjustments in the country’s recent history – directly attributed to rising import costs resulting from conflict-related supply disruptions.

Fifty percent of Pakistani respondents identify inflation and cost of living as the single most significant economic consequence of the conflict, the highest figure for any country on this measure. Pakistan’s dependence on imported crude oil, combined with a depreciating rupee and constrained foreign exchange reserves, creates a transmission mechanism that converts global oil price shocks directly into consumer-level inflation.

Pakistan also brokered the short-lived ceasefire between Iran and the United States that took effect on 8 April 2026 before collapsing on 12 April. The ceasefire’s failure has further complicated Pakistan’s diplomatic positioning and reinforced public anxiety about prolonged economic disruption.

Kenya: From Shortage to Record-Breaking Price Adjustment

Kenya presents a particularly instructive case study. At the time of surveying in March 2026, Kenya’s fuel prices were government-regulated through the EPRA pricing mechanism, which had effectively absorbed global price increases without passing them to consumers. However, 79% of Kenyan respondents still reported fuel price impact – because the economic strain was manifesting not through prices but through supply disruptions.

By early April, a severe fuel shortage had spread across at least 13 counties. In response, the government deployed KSh 6.2 billion in emergency subsidies and reduced VAT on fuel from 16% to 13%. These measures proved insufficient to contain the crisis.

On 14 April 2026, EPRA announced the largest fuel price adjustment in over 21 years of regulatory records: super petrol now retails at KSh 206.97 per litre in Nairobi, up KSh 28.69 from KSh 178.28, while diesel rises KSh 40.30 to an all-time high of KSh 206.84, effective 15 April. EPRA data indicate that the landed cost of imported super petrol rose 41.5% and diesel 68.7% during the review period. The regulatory body cited “significant increases in the prices of petroleum products in the international market” as the primary driver.

The magnitude of these adjustments points to the unsustainability of shielding consumers from global price shocks through regulation alone, and validates the concerns expressed by the 79% of Kenyan respondents who identified fuel price impact before the price adjustment was officially announced.

Egypt: Inflation Compounds an Existing Crisis

Seventy-eight percent of Egyptian respondents report fuel price impact. Egypt, which floated its currency in March 2024 and has experienced sustained inflationary pressure, is particularly vulnerable to energy price shocks. The government has raised subsidised fuel prices three times in the past twelve months. Brent crude’s rise from approximately $70 per barrel in late 2025 to over $128 per barrel in March 2026 has placed severe strain on Egypt’s import bill and fiscal position.

Forty-eight percent of Egyptian respondents cite inflation as the most significant economic consequence – the second-highest figure after Pakistan (50%). Nineteen percent identify food prices specifically, the highest of any country, reflecting the compounding effect of energy costs on food production and transport.

South Africa: A Slow-Burning Crisis

Sixty-eight percent of South African respondents report fuel price impact. The inland price of 95-octane petrol exceeded R30 per litre in March 2026. South Africa’s fuel pricing mechanism adjusts monthly based on international crude prices, the rand–dollar exchange rate, and shipping costs – all three of which have moved unfavourably. The Automobile Association of South Africa warned in April that further significant increases are expected for May 2026.

Twenty-two percent of South African respondents cite employment and job losses as the most significant economic consequence – the highest figure for any country on this measure – reflecting broader structural vulnerabilities in an economy already contending with 32% unemployment.

Nigeria: A Producer Still Feeling the Pressure

Despite being Africa’s largest oil producer, 56% of Nigerian respondents report fuel price impact. Nigeria’s Dangote refinery, which began operations in late 2024, has partially insulated the domestic market from global price shocks. However, the naira’s weakness and continued import dependence for refined products mean that global price movements still transmit to consumers, albeit with a lag.

The relatively lower figure compared to other countries in the sample may reflect some insulating effect of domestic production, but 56% still represents a majority reporting impact – a finding that challenges any assumption that oil-producing nations are immune to the conflict’s economic consequences.

Saudi Arabia: Impact Even for the Region’s Largest Producer

Saudi Arabia registers the lowest fuel price impact at 46%, consistent with its position as the world’s largest oil exporter with heavily subsidised domestic fuel prices. That nearly half of Saudi respondents still report an impact suggests the conflict’s economic effects extend beyond fuel pricing to broader cost-of-living increases and market uncertainty.

Inflation is the Primary Economic Consequence

When asked to identify the single most significant economic consequence of the conflict, respondents across all six countries point to inflation and cost of living (43%), followed by fuel prices specifically (27%), food prices (15%), and employment or job losses (13%). The pattern is consistent across countries, though the relative weighting varies with national economic conditions.

The Strait of Hormuz: A Global Chokepoint Under Pressure

The economic dynamics documented in this study are inseparable from developments in the Strait of Hormuz, through which approximately 20% of the world’s daily oil supply transits. Following the collapse of the Pakistan-brokered ceasefire on 12 April 2026, the U.S. Navy intensified its maritime operations in the Persian Gulf, raising the operational risk premium on all crude oil shipped through the strait.

Brent crude prices rose from approximately $70 per barrel in late 2025 to over $128 per barrel by mid-March 2026. While prices have fluctuated with diplomatic developments, the U.S. Energy Information Administration’s revised 2026 forecast of $96 per barrel (up from $74) signals that markets anticipate sustained disruption. For import-dependent economies such as Pakistan, Kenya, and Egypt, each dollar increase in the Brent price translates directly into higher landed costs for fuel, food, fertiliser, and manufactured goods.

Citizen Perspectives

The survey included open-ended responses that contextualise the quantitative findings. The following responses are representative of the concerns expressed across the six-country sample:

Why This Matters

The data presented in this report demonstrate that the Iran–Israel–U.S. conflict is not merely a geopolitical crisis confined to the Middle East. It is an economic event with measurable, immediate consequences for populations across the Global South. The 70% of respondents reporting fuel price impact, the 92% expressing cost-of-living concern, and the cascading effects on food, transport, and employment represent a humanitarian and policy challenge that extends well beyond the direct conflict zone.

For policymakers, the findings underscore the limits of domestic price controls and subsidies in the face of sustained global energy price shocks. Kenya’s trajectory, from regulated prices to nationwide shortage to record-breaking price adjustment, illustrates the unsustainability of shielding consumers indefinitely from global market forces.

For international organisations and development agencies, the data provide an empirical basis for understanding how distant conflicts translate into lived experience for citizens in Africa, South Asia, and the Middle East. The economic consequences documented here are likely to intensify if the conflict continues or escalates.

Methodology

This report draws on data from GeoPoll’s Caught in the Crossfire? citizen perceptions study, conducted in early March 2026. The study surveyed 3,754 respondents across six countries: Egypt (n = 626), Kenya (n = 627), Nigeria (n = 625), Pakistan (n = 626), Saudi Arabia (n = 624), and South Africa (n = 626).

Respondents were recruited through GeoPoll’s proprietary mobile panel, which uses random sampling from mobile network operator databases to reach nationally representative populations. Surveys were administered via mobile-based interviewing across multiple modes, including CATI, SMS, and mobile web. All respondents were aged 18 and above.

The margin of error for country-level estimates is approximately ±3.9% at a 95% confidence level. Cross-country comparisons should be interpreted with awareness of differing national contexts, including variation in government fuel pricing policies, currency stability, and import dependence.

Access the full 37-page report:

Caught in the Crossfire? A Six-Country Citizen Perceptions Study on the Iran–Israel–U.S. Conflict

For enquiries about country-specific data, custom analysis, or partnership opportunities, contact us.

The post The Price of Conflict: How the Iran-Isreal-U.S. War is Affecting Fuel Costs and Supply and Cost of Living appeared first on GeoPoll.

Pakistan occupies a unique position in the Iran–Israel–U.S. conflict. As Iran’s immediate neighbour, a nuclear-armed state, a major recipient of Chinese investment through CPEC, and a long-standing U.S. security partner, it sits at the intersection of every fault line this conflict has exposed. In early March 2026, GeoPoll surveyed 626 Pakistani citizens as part of a broader six-country study. The results reveal the most polarised and consistently intense responses of any country in the dataset.

Who Is to Blame for the Conflict? Israel, Overwhelmingly

Sixty-three percent of Pakistani respondents hold Israel most responsible for the conflict – the highest figure for any country in the survey and nearly double the six-country average (38%). A further 20% blame the United States, while only 5% point to Iran. The framing is clear: for most Pakistanis, this is a war waged by Israel and the U.S., not against them.

63%

of Pakistanis blame Israel: the highest of any country surveyed

Sympathy Tilts Sharply Toward Iran

Related to the finding above, Pakistan registers the strongest pro-Iran sympathy in the dataset at 82% – nearly double the overall average of 43% and the single highest figure for any question-country combination in the study. Just 3% express sympathy for Israel. This reflects deep religious and cultural affinity, shared borders, and decades of people-to-people ties. It is also consistent with Pakistan’s diplomatic positioning, which has historically sought to balance its Western security partnerships with its identity as a Muslim-majority state.

Fear of Escalation Is Acute

Eighty-six percent of Pakistani respondents believe the conflict could lead to a wider global war, which is not abstract anxiety. Pakistan shares a border with Iran, has its own nuclear arsenal, and is acutely aware that regional escalation could draw it in directly. The nuclear dimension resonates deeply, with 70% of respondents across all six countries view nuclear weapons use as likely, and Pakistan’s own nuclear status makes this fear personal rather than theoretical.

86%

fear the conflict could escalate into a global war

Pakistan is Experiencing the Worst Economic Pain

Pakistan reports the most severe economic impact of any country surveyed. Eighty-five percent say the conflict has significantly affected fuel prices, consistent with the government’s historic Rs 55 per liter fuel price hike on 6 March 2026, directly attributed to regional supply disruptions. Fifty percent cite inflation and cost of living as the single most significant economic consequence, the highest of any country.

Nearly three-quarters (73%) of Pakistani respondents report some form of personal impact from the conflict, far exceeding any other country. As one respondent put it: “It’s worrisome as we are in alliance with the States so we could be hit next.”

A Geopolitical Realignment: China Up, U.S. Complicated

Pakistan’s relationship with global powers is shifting. Forty-four percent now view China more favourably as a result of the conflict – the largest positive shift of any country – and 49% trust China most to act in Pakistan’s and the world’s interests. This is more than double any other country-body combination except Kenya’s and Nigeria’s trust in the UN.

Views of the U.S. are more nuanced than might be expected. While 33% view the U.S. less favourably, an equal 33% say their view is unchanged, and 20% actually view the U.S. more favourably. Pakistan is the only country where a significant share (39%) believes the U.S. is primarily serving Israel’s interests rather than its own.

Russia also benefits: 33% of Pakistanis view Russia more favourably, the highest positive shift for Russia in the dataset. China’s non-interventionist positioning and its deep economic ties through CPEC are clearly paying reputational dividends.

What Pakistani Citizens Want: Peace – and Support for Iran

Pakistan is the only country in the survey where a large share (42%) explicitly wants their government to support Iran, nearly equalling the 44% who call for peace negotiations. In every other country, peace negotiations command a clear majority. This split reflects the depth of religious and geopolitical solidarity with Iran, and suggests that any Pakistani government response perceived as neutral or Western-aligned could face significant public backlash.

On institutional trust, 35% look to the UN and 31% to China as most capable of resolving the conflict. Only 9% trust the United States. As one respondent noted: “Our country Pakistan is still sitting silent and foreign policy support is neutral.”

When asked which country or international body they trust most to act in the best interests of Pakistan and the world, China was, interestingly, overwhelmingly the most trusted (49%) – over twice that of even the United Nations.

Why This Matters

Pakistan’s profile in this data is the most extreme and internally consistent of any country surveyed. The combination of overwhelming blame on Israel, near-universal sympathy for Iran, the highest economic pain, the strongest China alignment, and the most acute fears of global escalation paints a picture of a population that feels deeply, personally affected by a conflict in which it has no formal role.

For policymakers and international organizations, the implications are clear. Pakistan’s public opinion is not a passive backdrop – it is a force that shapes the government’s room to maneuver on foreign policy, alliance decisions, and economic stabilization. Any diplomatic strategy that ignores this sentiment risks misreading one of the most consequential populations in the conflict’s broader orbit.

Methodology

This report draws on data from GeoPoll’s multi-country online survey conducted in the first week of March 2026, during a period of active military escalation in the Middle East following the launch of U.S.-Israeli Operation Epic Fury on 28 February 2026.

The survey collected 3,754 responses across six countries: Egypt, Kenya, Nigeria, Pakistan, Saudi Arabia, and South Africa. The Pakistan subsample comprises 626 respondents. Respondents were recruited through GeoPoll’s online panel and surveyed via a structured questionnaire covering conflict awareness, attribution, sympathy, escalation concerns, economic impact, views on global powers, institutional trust, media consumption, personal security, and preferred government responses. GeoPoll administered the questionnaire in both English and Urdu in Pakistan.

All percentage figures are rounded to the nearest whole number. Country-level breakdowns are unweighted.

Read the full 37-page report

This is a sub-report of GeoPoll’s report, “Caught in the Crossfire?“ A Six-Country Citizen Perceptions Study on the Iran–Israel–U.S. Conflict.

To read the full 37-page report, including detailed cross-country comparisons, verbatim citizen voices, and policy recommendations, visit: www.geopoll.com/blog/iran-israel-us-conflict-report

For enquiries about country-specific data, custom analysis, or partnership opportunities, contact [email protected]

The post Pakistan: Caught in the Iran-Israel-US Crossfire? appeared first on GeoPoll.

Back Home

Report: What Citizens Across the Global South Really Think About the Iran–Israel–U.S. Conflict

Deep economic anxiety, stark regional divides on blame and sympathy, declining trust in Western powers and media, and an overwhelming demand for peace.

The Middle East conflict has consumed headlines, but one voice has been largely missing from the conversation: that of the billions of people across the Global South who are bearing the economic and social fallout.

In March 2026, GeoPoll surveyed citizens across  Pakistan, Saudi Arabia, Egypt, Kenya, Nigeria, and South Africa, to understand how they perceive, experience, and respond to the escalating Iran–Israel–U.S. conflict. The findings are striking and carry direct implications for governments, international organisations, and media institutions.

100% are aware of the conflict. Very high.

70% believe the use of nuclear weapons is likely

72% believe this might be the start of World War 3

38% blame Israel for the war; 29% the US;18% Iran

43% view the U.S. less favourably due to the war

25% say Western media is misleading

70% say fuel prices significantly affected

69% are "very concerned" about the cost of living

54% want their govt to call for peace

The free 37-page report has detailed country-level breakdowns, cross-tabulations, open-ended citizen responses in three languages, comparisons with on-the-ground realities, and actionable policy recommendations, and is essential reading whether you are a policymaker, diplomat, development practitioner, journalist, researcher, or simply curious about how this conflict is reverberating far beyond the Middle East

Fill in this form to download the full report (free):

The post Report: Global South Perceptions the Iran–Israel–U.S. Conflict appeared first on GeoPoll.

The email lands on a Monday morning. A client, let’s say a development organization working across Africa, needs to understand how communities are adapting to climate shocks. They have funding, a timeline, and a genuine need for answers. What they often lack is a fully developed research design.

“We trust you to figure out the best approach,” they write. “You are the experts.”

This is how most research projects begin. Not with a polished methodology section, but with a problem that needs solving and a partner trusted to translate that problem into rigorous inquiry. The space between “we need to understand X” and a fieldwork-ready research design is where some of the most consequential decisions get made.

It is also where AI is proving unexpectedly useful.

The Messy Reality of Research Design

Research design isn’t linear. It is iterative, collaborative, and often constrained by factors that have nothing to do with methodological purity, such as budget limits, timeline pressures, data availability, political sensitivities, and client expectations.

The process typically involves:

• Clarifying what the client actually needs to know (which isn’t always what they initially ask for)
• Understanding what’s already known about the topic
• Identifying the right questions to answer the underlying need
• Determining what methodology will yield credible answers given real-world constraints
• Anticipating what could go wrong and designing around it

Experienced researchers carry much of this in their heads – pattern-matched from dozens of similar projects. But that expertise is hard to scale, and even veterans have blind spots.

This is where AI enters the picture. Not as a replacement for research expertise, but as a thinking partner that can hasten and strengthen each stage of the design process.

Vague Brief to Sharp Research Questions

Let’s return to our climate adaptation project. The client’s initial brief is broad: “understand how communities are adapting to climate shocks.” That’s a starting point, not a research question.

The first task is understanding what they actually need. Are they interested in documenting existing adaptation strategies? Measuring their effectiveness? Understanding barriers to adoption? Identifying which populations are most vulnerable? All of these could fall under “climate adaptation,” but each implies a different study.

AI can help here by:

Generating structured questions that surface unstated assumptions. Feed the brief into a well-prompted model, and it will return a list of clarifying questions the research team should ask: What types of climate shocks? What timeframe? Which communities? What decisions will this research inform?

Mapping the problem space. AI can quickly generate a conceptual map of related variables, potential frameworks, and dimensions worth considering. This isn’t definitive. It’s a starting point for discussion that ensures nothing obvious gets overlooked.

Suggesting alternative framings. Sometimes, the most valuable thing a research partner can do is reframe the question. A model trained on diverse research, such as GeoPoll’s specifically tuned AI Engine, can propose angles the client hadn’t considered, shifting the focus from “how are communities adapting?” to “what predicts successful adaptation?” or “where are adaptation efforts failing, and why?”

None of this replaces the conversation with the client. But it compresses what might take several rounds of back-and-forth into a more focused initial discussion.

What’s Already Known, and AI-Assisted Literature Review

Good research design requires understanding the existing landscape. What have others found? What methodologies have worked? Where are the gaps?

Traditional literature review is time-intensive. Researchers spend hours searching databases, scanning abstracts, reading papers, and synthesizing findings. For a well-funded academic study, this investment is appropriate. For a rapid-turnaround applied project with a six-week timeline, it’s often impractical.

AI doesn’t replace rigorous literature review, but it dramatically accelerates preliminary synthesis:

Rapid landscape mapping. Within minutes, AI can summarize what’s broadly known about a topic, identify key debates, and flag seminal studies worth reading in full. This gets the research team to baseline understanding faster.

Identifying methodological precedents. “How have others studied climate adaptation in Africa?” is a question AI can answer with reasonable accuracy, pointing toward approaches that have worked and those that have faced criticism.

Surfacing gaps. AI can synthesize what exists and help identify what doesn’t: unanswered questions, understudied populations, and untried methodologies. These gaps often become the most valuable research opportunities.

Cross-disciplinary connections. AI doesn’t respect academic silos. It might surface relevant work from behavioral economics, anthropology, or public health that a researcher siloed in their own discipline might miss.

The important caveat is that AI-generated literature summaries require verification. Models can hallucinate citations, mischaracterize findings, or miss recent work. The output is a starting point for human review, not a finished product.

Designing for Constraints

Every research project operates within constraints. Budget caps what’s possible. Timelines limit depth. Access determines who can be reached. Political sensitivities shape what can be asked.

Experienced researchers chart these tradeoffs intuitively. AI can make that navigation more systematic:

Scenario modeling. Given a fixed budget, what sample sizes are achievable across different methodological approaches? A trained AI model can quickly model tradeoffs – a larger sample with phone surveys versus a smaller sample with in-person interviews, helping teams make informed decisions.

Risk identification. What could go wrong? AI can generate a preliminary risk register based on the project parameters: potential for low response rates in certain regions, sensitivity of particular questions, logistical challenges in specific geographies. This isn’t exhaustive, but it prompts the team to think through contingencies.

Methodology matching. Given the research questions, constraints, and context, what methodological approaches make most sense? AI can suggest options the team might not have considered and flag potential limitations of each.

Pressure-Testing Assumptions

Every research design rests on assumptions, about respondent behavior, about data quality, about what questions will actually measure what you intend them to measure.

AI is useful for stress-testing these assumptions before fieldwork begins:

Anticipating respondent interpretation. How might a question be understood differently across contexts? AI can simulate diverse respondent perspectives, flagging potential misinterpretation before you’re in the field. This is one of a few areas where GeoPoll uses synthetic data.

Identifying confounding variables. What factors might influence the outcomes you’re measuring that aren’t captured in your design? AI can generate lists of potential confounds worth considering.

Checking logical consistency. Does the research design actually answer the research questions? It’s surprisingly easy for these to drift apart. AI can serve as a check, mapping questions to design elements and flagging gaps.

What AI can’t do in Research Design

It would be easy to overstate AI’s role here, so let’s be clear about the limits.

AI can’t define what matters. The strategic decisions, such as what questions are worth answering, what tradeoffs are acceptable, and what the research should ultimately accomplish, remain human judgments. AI can inform these decisions; it can’t make them.

AI doesn’t understand context the way practitioners do. A model doesn’t necessarily know that a particular region has experienced recent political upheaval that will affect response patterns, or that a certain phrasing carries unintended connotations in local dialect. Contextual knowledge is irreplaceable.

AI can’t navigate relationships. Research design is often negotiated with clients, partners, communities, and institutions. The interpersonal work of aligning stakeholders, building trust, and managing expectations is entirely human.

AI outputs require judgment. Everything AI produces in the design phase needs evaluation by experienced researchers. The model doesn’t know when it’s wrong. Humans have to.

How to Integrate AI into Research Design

The most effective use of AI in research design follows a consistent pattern:

1. Human defines the problem and constraints. The client’s need, the project parameters, and the contextual factors come from people.
2. AI powers exploration. Literature synthesis, question generation, methodology options, risk identification, and AI compresses what would otherwise take days into hours.
3. Human evaluates and decides. Every AI output gets filtered through research expertise. What’s useful gets kept; what’s off-base gets discarded.
4. The cycle repeats. Design is iterative. AI can be brought back in at each stage to pressure-test, expand options, or check consistency.

This is not AI replacing researchers at the research stage. This is actually one of the areas where human experts are critical because it can make or break research. It’s AI amplifying what good researchers already do – asking better questions, considering more angles, anticipating more problems- at a pace that matches real-world project timelines.

Questionnaire Development

Research design ultimately culminates in the instruments you will use to collect data: the questionnaire, discussion guide, or observation protocol. AI has significant applications here as well, from drafting and iteration to translation and cognitive testing.

We’ll cover questionnaire development in depth later in this series. For now, the key point is that stronger upstream design – clearer questions, better understanding of context, more thoroughly considered methodology – makes instrument development faster and more effective.

Looking Ahead

Thinking about the climate adaptation project we started with, with AI assistance, the research team can move from a vague brief to a detailed design proposal in a fraction of the time it once required. The proposal is sharper because more options were considered. The methodology is stronger because more risks were anticipated. The questions are better because more assumptions were tested.

None of this guarantees good research. That still depends on execution, judgment, and the irreplaceable expertise of people who understand what they’re studying. But the foundation is stronger.

Working on a research design challenge? We’d welcome the conversation. Contact GeoPoll to discuss how we approach complex projects across diverse contexts.

The post AI in Research: Design and Problem Definition appeared first on GeoPoll.

AI has become one of those words that’s everywhere, a buzzword in boardrooms, a curiosity in most conversations, professional or social, and increasingly, a quiet presence in how work actually gets done. According to Google’s Our Life with AI Report, 48% people globally now use AI at work at least a few times a year, with writing and editing tools among the most common applications. Among content professionals, the numbers are even higher: over 70% use AI for outlining and ideation, and more than half use it to draft content.

The adoption curve is real. But so is the uncertainty. In Stack Overflow’s 2025 developer survey, 84% of respondents use or plan to use AI tools, yet 46% say they don’t trust the accuracy of the output. People are using AI. They’re just not sure how much to believe it.

For researchers, this tension is especially acute. Our work demands rigor. It requires accuracy, nuance, and accountability, qualities that don’t pair naturally with tools known for confident-sounding hallucinations. And yet the potential is hard to ignore: faster questionnaire development, smarter quality assurance, analysis at scales that weren’t previously practical.

So where does that leave us? Adoption. For all the attention it receives, much of the conversation remains polarized. On one end is hype: claims that AI will “replace research as we know it.” On the other is skepticism: a belief that AI is fundamentally incompatible with rigorous, ethical, human-centered inquiry.

The reality sits somewhere in between.

As our CEO, Nicholas Becker wrote in this article, AI is not changing why research is conducted. It is changing how it is conducted, and in doing so, it is forcing the research community to revisit long-held assumptions about quality, speed, scale, and responsibility.

This post and the series that follows aim to fill that gap. We will share what we have learned about where AI genuinely adds value in research, where it falls short, and how to think about integration in ways that strengthen rather than complicate your work.

The Current Landscape

AI adoption in research is uneven, and for understandable reasons.

Some organizations, such as GeoPoll, are experimenting aggressively and automating significant portions of their analysis workflows. Others are watching and waiting, uncertain whether the tools are mature enough to trust with work that demands rigor.

Both positions are reasonable. The gap between what AI can do in controlled demonstrations and what it reliably does under field conditions is real. A tool that performs impressively on clean, English-language data may struggle with the realities of multilingual surveys, low-connectivity environments, or the cultural nuance required to interpret responses from communities the model has never encountered.

This is particularly true for research in emerging markets and complex settings, exactly the contexts where good data is most needed and hardest to collect. The assumptions baked into many AI tools often reflect their training environments: high-resource languages, stable infrastructure, Western cultural frameworks. When those assumptions don’t hold, performance degrades in ways that aren’t always obvious.

None of this means AI isn’t useful. It means we need to be specific about where it works, honest about where it doesn’t, and thoughtful about how we integrate it.

Where AI Genuinely Adds Value

Let’s start with what’s working. These are applications where the technology is mature enough to deliver consistent value, and where we have seen real improvements in efficiency, quality, or both.

1. Research Design and Problem Definition

Early-stage research design has always been one of the most human-dependent phases of the process. Defining the right question, aligning objectives, and translating abstract goals into measurable constructs requires judgment, domain knowledge, and contextual awareness.

AI can support this stage by synthesizing large volumes of background material, identifying recurring themes across prior studies and stress-testing logic, assumptions and consistency in objectives.

This is one of the very few places where GeoPoll uses synthetic data – to simulate real-world possibilities and tighten the research design.

However, AI cannot determine what matters. It can help refine how a question is phrased, but it cannot decide whether the question is meaningful, relevant, or appropriate for a given context. That responsibility remains firmly human.

2. Questionnaire Development and Translation

In relation to the research design above, AI has also become a genuine accelerator in the early stages of instrument design. AI can generate initial question drafts, identify ambiguous phrasing, suggest alternative wording, and flag potential sources of bias. They are particularly useful for cognitive pretesting, helping you anticipate how respondents might misinterpret questions before you’re in the field.

Translation and back-translation workflows have also improved significantly. While human review remains essential, AI can produce working drafts faster and more consistently than traditional approaches, freeing skilled translators to focus on nuance rather than first passes.

This has been particularly useful to us as we conduct several multicountry and multilingual surveys. Using thousands of our past translated questionnaires, we have trained our own models to produce translations that are close to fine, which makes the work a lot easier and more efficient for our translation teams to only review.

3. Quality Assurance and Data Cleaning

Quality control is where AI’s pattern recognition capabilities shine. Real-time monitoring during data collection can flag anomalies. Interviews completed suspiciously fast, response patterns that suggest straightlining or satisficing, geographic inconsistencies, or interviewer behaviors that warrant review.

The value here isn’t replacing human judgment but directing it more efficiently. Instead of reviewing random samples, quality teams can focus attention where it’s most needed. Fraud detection, in particular, has become significantly more sophisticated with machine learning approaches that identify coordinated fabrication patterns humans might miss.

4. Analysis and Insight Generation

Anyone who has manually coded thousands of open-ended responses understands the appeal of automation. Natural language processing, again, with well-trained models such as the one GeoPoll Senselytic uses, can now handle initial coding, theme extraction, and sentiment analysis at scale. Work that previously consumed enormous time and introduced its own inconsistencies.

The keyword is “initial.” AI-generated codes require human review, and the categories need refinement based on contextual understanding the model might lack. But as a first pass that analysts then validate and adjust, the efficiency gains are substantial. Also, analysis is not insight. AI can surface patterns, but it may not fully understand causality, significance, or implication in the way decision-makers require. Without human interpretation, there is a real risk of over-fitting narratives to statistically convenient patterns.

Then feed the results back into the model and continuously improve its capabilities for next time.

5. Reporting, Visualization, and Storytelling

Beyond analysis, AI streamlines the communication of findings: drafting report sections, generating visualization options, summarizing results for different audiences, and adapting technical findings into plain narratives.

For organizations producing high volumes of research, this represents significant time savings. First drafts that once took days can be generated in hours, freeing researchers to focus on refinement, interpretation, and strategic recommendations.

6. Operational Efficiency

Beyond the research process itself, AI streamlines the operational work that surrounds it: drafting reports, cleaning and restructuring data, generating documentation, and summarizing findings for different audiences. These applications are less glamorous but often deliver the most immediate time savings.

But Human Judgment Remains Essential

Listing AI’s capabilities without acknowledging its limitations would be both incomplete and misleading. There are aspects of research where human judgment isn’t just preferable, it’s irreplaceable.

1. The Foundation

Deciding to conduct research does not begin at the research design stage. It starts with a real problem an organization needs to solve. AI can help refine questions, but it can’t tell you which questions matter. The strategic decisions that shape a study – what to measure, why it matters, how findings will be used – require understanding of context, stakeholders, and objectives that models don’t possess. This is where research value is created or lost, and it remains fundamentally human work.

2. Contextual Interpretation

Data doesn’t interpret itself. Understanding what a response pattern means requires knowledge of local context – political dynamics, cultural norms, recent events, historical relationships – that AI tools lack. A model might identify that responses in a particular region differ from the national average; understanding why they differ, and what that implies for the research question, requires human insight.

This is especially critical in cross-cultural research, where the same words can carry different meanings, and where what’s left unsaid is often as important as what’s captured in the data.

3. Ethical Judgment

Research involves ongoing ethical decisions: how to handle sensitive disclosures, when informed consent requires additional explanation, how to protect vulnerable respondents, whether certain questions should be asked at all in particular contexts. These judgments require moral reasoning, empathy, and accountability that can’t be delegated to algorithms.

4. Stakeholder Relationships

Research happens within relationships – with communities, partners, clients, and institutions. Building trust, navigating sensitive topics, communicating findings in ways that lead to action rather than defensiveness: these are human skills that no AI will replicate. The credibility of research ultimately rests on the people behind it.

5. Final Analytical Decisions

AI can surface patterns and generate hypotheses, but the final interpretive decisions – what the data means, how confident we should be, what recommendations follow – belong to researchers. The stakes of getting this wrong are too high, and the accountability too important, to outsource.

The Integration Question

Based on all this, the question isn’t whether to use AI but how to integrate it without breaking what already works.

The most sustainable approach treats AI as an augmentation rather than a replacement. The goal isn’t to automate researchers out of the process but to free them from tasks where their judgment adds less value, so they can focus where it adds more. AI handles the volume while humans handle the judgment.

This requires what’s often called “human-in-the-loop” workflows: processes designed so that AI outputs are reviewed, validated, and refined by people before they influence decisions. It’s slower than full automation, but it’s also more reliable and more accountable.

It also requires building internal capacity. Organizations that outsource AI entirely to vendors risk losing understanding of how their research is actually being conducted. The teams that will use AI most effectively are those that understand it well enough to know when it’s helping and when it’s not.

In our work at GeoPoll, we see AI as a tool that strengthens research when it is embedded thoughtfully, not when it is layered on top as a shortcut. The most effective applications combine automation with clear methodological guardrails and continuous human oversight.

What This Series Will Cover

This article sets the foundation for a deeper exploration of AI across the research lifecycle. In the coming pieces, we will go into each stage in detail, looking closely at what works, what doesn’t, and what responsible use looks like in practice:

• Research design and questionnaire development: From hypothesis to instrument
• Sampling and recruitment: Reaching the right respondents
• Data collection: Fieldwork in the age of AI
• Quality assurance: Detection, monitoring, and validation
• Analysis and interpretation: From data to insight
• Reporting and visualization: Communicating findings effectively
• Ethics and limitations: What AI can’t do, and why it matters

Each post will be practical and specific, drawing on real-world applications and our experience rather than theoretical possibilities.

GeoPoll’s Perspective

At GeoPoll, we have spent over a decade conducting research in some of the world’s most challenging environments—conflict zones, low-connectivity regions, rapidly evolving political contexts. We complete millions of interviews annually across more than 100 countries, in dozens of languages, using mobile-first methodologies designed for conditions where traditional approaches don’t work.

That experience has shaped how we think about and work with AI. We have seen what works when assumptions break down, when infrastructure isn’t reliable, and when the cultural context is unfamiliar to the models. We have learned through iteration, testing tools in the field, finding their limits, and building workflows that account for them. As a technology research company, we have built AI platforms and processes into our research and are actively employing AI to make our work easier and deliver greater value to our clients and partners.

This is the knowledge we are sharing in this series.

If you are thinking about how AI might strengthen your research, we would welcome the conversation. Contact us to discuss what’s working, what’s not, and where the opportunities might be.

The post AI in Research Series: Where we are and where it actually works (or not) appeared first on GeoPoll.

Research clients are feeling it. Actually, a few have reached out to us at GeoPoll recently to say that other panel providers delivered datasets full of questionable responses. As an example, we audited a dataset from one of these projects and found respondents claiming to work for companies that, after cross-checking, did not exist. That is not a minor quality issue, but a failure of the most basic layer of respondent verification.

The problem is not isolated. It is becoming pervasive, and it threatens the trustworthiness of survey research if left unchecked.

In this article, we break down what is happening, why it is happening, and, most importantly, what the industry must do about it.

Why online sampling is under pressure

The challenges the industry is experiencing step from pressures on

• The explosion of bots and automated respondents – Fraudulent actors can now generate large volumes of convincing survey completions using tools that simulate human behaviour, including normalised click paths, varied timing, and even device switching. The barrier to entry is low, the incentives are high, and the fraudsters are increasingly sophisticated.
• AI-generated open-ended responses – One of the downsides of generative AI to the industry is that it has introduced a new challenge: artificial open-ended responses that sound perfectly human but contain no personal context. This is especially dangerous because open-ended questions were once reliable indicators of quality. Today, AI models can produce responses that are linguistically rich yet completely unauthentic, which makes manual review far more difficult.
• Panel fatigue and low engagement – A third pressure point is panel fatigue. In many markets, respondents are oversurveyed and under-engaged. As genuine participation declines, some panel providers fill quotas through loosely vetted traffic sources, unverified accounts, or third-party supplies whose quality mechanisms are opaque. This is often where “junk” data enters the chain, responses that look complete but crumble under scrutiny.
• Nonexistent profiles and artificial identities – Beyond fake companies, we are now seeing invented educational histories, geographic misrepresentation through VPNs, and household profiles that defy demographic reality. Incentive-driven fraud compounds this by enabling entire online communities to trade survey links, completion codes, and tips for bypassing checks.

The result is a landscape where bad data can be gathered at scale, faster than many traditional panels can detect it, compounded by technology.

Even from our own tests using the GeoPoll AI Engine, AI models can now generate human-like narratives, differentiated “voices”, realistic demographic profiles, and varied completion speeds. The reality is that as long as incentives exist, fraudulent responders will continue to innovate.

Meanwhile, many panel providers rely on legacy systems built for a world where fraud meant speeding or straight-lining. They were not designed to detect AI paraphrasing, synthetic behavioural fingerprints, cross-platform identity laundering, and real-time pattern anomalies

This mismatch creates structural vulnerability.

What this means for researchers and clients

Poor-quality sample data has obvious consequences, the immediate of which include:

• Misleading insights
• Incorrect targeting
• Wasted budgets
• Incorrect strategic decisions
• Damaged credibility

But the deeper consequence is even more serious: If the industry does not rebuild trust in online sampling, brands and organizations will hesitate to rely on survey research at all. When decision-makers cannot trust the integrity of respondent data, they begin to question the value of surveys as a method. This is the real risk—an industry-wide credibility problem.

A reliable respondent ecosystem rests on three foundations: identity, location, and behaviour.

Respondents must be tied to real, verifiable identities. Their location must reflect where they actually are, not where their VPN says they are. And their behaviour must reflect natural human variation—not the automated consistency of scripts, bots, or artificially generated text.

These are basic principles, but in an era of synthetic identities and AI-driven fraud, they require much more rigorous systems to uphold.

How the industry should respond

Online sampling is not going away; if anything, demand will increase. But the industry must adapt. Fraud is evolving faster than legacy panel systems can respond, and researchers cannot afford to rely on outdated assumptions about respondent authenticity.

The future belongs to providers who treat data quality as a core capability, and not a back-office function. Those who invest in verification, diversify sampling modes, apply advanced fraud detection, and communicate transparently will set the new standard. The rest will continue to generate “junk” data and erode trust in research.

Rebuilding trust in online sampling will require a combination of technology, methodological discipline, and transparency.

• Strengthen Identity Verification: Email-based registration is no longer sufficient. Providers need to move toward systems grounded in SIM-based verification, mobile operator partnerships, two-factor authentication, and device-level identity checks. Emerging markets with national SIM registration frameworks have a distinct advantage here.
• Detect Fraud Behaviourally: Quality control must evolve beyond speeding and straight-lining. Modern systems should detect unusual device patterns, inconsistent browser fingerprints, abnormal timing sequences, proxy use, and other signs of automation. This has to happen pre-survey, not only during data cleaning.
• Use AI to Fight AI: Just as AI can generate deceptive responses, AI can also detect them. Linguistic analysis, stylometric fingerprints, and semantic anomaly detection are becoming essential tools for flagging artificial or copy-pasted open-ended text.
• Apply Human Oversight on High-Stakes Work: For sensitive audiences or high-value projects, manual review remains indispensable. Calling back a sample of respondents, checking claims when relevant, or auditing open-ended text can act as guardrails against fraud that slips through automated systems.
• Reduce Reliance on Third-Party Traffic: Panels built on first-party respondent networks, such as mobile communities, app-based samples, and telco-linked panels, are inherently more secure than those that rely on opaque third-party supply. Direct relationships create accountability and allow for deeper verification.
• Blend Modes When Necessary: Some populations or markets simply cannot be reliably captured through online traffic alone. Combining online surveys with CATI, SMS, WhatsApp, in-person intercepts, or panel phone lists reduces exposure to any single failure mode and strengthens representativeness. This why, at GeoPoll, we live for multimodal approaches to research.
• Be Transparent With Clients: Clear reporting on quality checks, verification processes, and exclusion rates builds trust. As fraud grows more sophisticated, transparency becomes a competitive advantage.

How GeoPoll approaches online sampling to reduce these risks

These issues are increasingly common, but they are avoidable with the right systems. GeoPoll’s platforms and processes are deliberately designed to protect data integrity and put the voice of real humans first. Our model was built for the types of environments where online sampling is now struggling most. Our respondent network is anchored in mobile-first infrastructure, with SIM-linked verification and direct partnerships that ensure respondents are real people, reachable through real devices.

We complement this with multi-mode data collection – CATI, mobile web, SMS, WhatsApp, app-based sampling, and in-person CAPI – so no single sampling method carries the full burden of quality. Our now AI-powered fraud detection systems track behavioural anomalies, detect AI-like response patterns, and monitor unusual activity across surveys. And for complex or high-stakes studies, our teams perform human review of suspicious profiles or open-ended answers.

Contact us to learn more about how we make sure your data collection is valid.

The post The Online Sampling Crisis: Why Bad Data is Rising and how to Stop it appeared first on GeoPoll.