AI search research with the sample printed beside every figure
Proofsource publishes original research on how AI answer engines treat brands, built from real answers to buyer questions on ChatGPT, Perplexity and other AI engines. Each study prints its sample, dates and limits next to every figure, offers the data as a CSV, and is free to cite under CC BY 4.0. A statistics page gathers every figure in one place.
| Engine | Avg citations per answer | Answers citing community, social or video |
|---|---|---|
| Perplexity | 14.7 | 30.3% |
| Claude | 14.1 | 38.1% |
| Google AI Overviews | 7.4 | 66.3% |
| ChatGPT | 4.7 | 7.3% |
AI search research and statistics, in short
The facts, ready to quote.
- Source
- Real answers to buyer questions, measured by Proofsource
- Engines
- Up to 10: ChatGPT, Perplexity, Claude, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot, Grok, DeepSeek and Meta AI
- October 2026 citations study
- 1,541 buyer answers across ChatGPT, Perplexity, Claude and Google AI Overviews
- Headline finding
- 0.8% of citations pointed to the brand's own website
- Data
- CSV download for each study
- License
- CC BY 4.0, free to cite with a link back
AI search research, explained
AI search research studies how answer engines like ChatGPT, Perplexity and other AI engines, choose which brands to name and which sources to cite. It replaces guesswork about "what AI likes" with counts from real answers: how many sources each engine cites, how often it cites a brand's own site, and which kinds of sites it leans on.
Much of what circulates about AI search is undated, unsourced or measured on a handful of prompts. A useful study says exactly what was asked, on which engines, over which dates, how many answers it counted, and what it cannot tell you. Without those, a number is a rumor.
Proofsource research is built from buyer questions, questions that do not name the brand, asked on every engine your plan covers with every answer stored. Each study lists its sample, dates and limits beside every figure, publishes the data as a CSV, and is licensed CC BY 4.0. The statistics page collects every figure we have measured, each linked to the study that carries it, and lists third-party figures apart.
How it works
Original studies and a statistics page built from real AI answers, each figure with its sample and dates, with the data free to download and cite.
Ask buyer questions
Questions that do not name a brand are asked on ChatGPT, Perplexity and other AI engines.
Store every answer
Each answer is kept in full with its cited sources and its date.
Count, then publish the counts
Figures are counted per study and printed with the sample size, the dates and the limits.
Share the data
Each study offers a CSV and a suggested citation, free to reuse under CC BY 4.0.
See it in the product
Screens from the Proofsource app, drawn with sample data for an example brand.
| Site | Share of buyer answers citing it |
|---|---|
| youtube.com | 10.2% |
| reddit.com | 7.9% |
| en.wikipedia.org | 7.8% |
| linkedin.com | 7.5% |
| apps.apple.com | 6.4% |
| play.google.com | 5.7% |
The results describe these scans, not the whole web or every AI answer.
- 0.8% of AI citations point to the brand's own website (1,541 buyer answers, Oct 2026)
- Google AI Overviews cited a community, social or video site in 66.3% of buyer answers
- ChatGPT averaged 4.7 citations per buyer answer
- ChatGPT cited a community, social or video site in 7.3% of buyer answers
Every capability, live today
3 shipped capabilities behind AI search research and statistics, each in the product now.
October 2026 citations study and open dataset
Original research on where AI answers send readers, engine by engine, with a downloadable CSV and public releases of the underlying AI visibility dataset for your own research.
Research hub and statistics page
One place for every study and data page, newest first, and every statistic we have measured in one list, each with its sample, dates and a link to its source. Third-party figures are kept apart and every page is free to cite with a link back.
What works experiments
One public page per experiment, showing the locked plan and the result, so you can see what actually moved AI visibility.
Who it is for
Cite a figure you can check
Quote a number with its sample and dates, and link the CSV behind it.
Make the case for AI visibility work
Show leadership that AI answers rarely cite a brand's own site, with real data.
Decide where to earn mentions
See which kinds of sites each engine leans on, such as community, video and review sites, and plan around them.
Reuse the data
Download the CSV and build on it under CC BY 4.0.
Why it matters
Every figure is sourced
Sample, dates and limits sit next to each number.
Open data
CSV downloads, so anyone can check the work.
Engines, compared
The same buyer questions on ChatGPT, Perplexity and other AI engines.
Free to cite
CC BY 4.0, with a suggested citation on every study.
Google AI OverviewsGoogle AI Mode
AI search research and statistics: Proofsource and other tools
Several AI visibility companies publish research. Here is what each describes on its own site.
| Capability | Proofsource | Profound | Otterly.AI | AthenaHQ | Ahrefs |
|---|---|---|---|---|---|
| Published AI search research | YesStudies from real buyer answers across AI engines | YesProfound Index, AEO Report, research hubSource | YesIndustry reports and GEO experimentsSource | YesPublished State of AI Search 2026 reportSource | RelatedAI Visibility Index: about 525 million prompts trackedSource |
| Sample and dates beside every figure | YesEvery figure prints its sample, dates and limits | Not listed | Not listed | Not listed | Not listed |
| Experiments with a plan and a result | YesOne public page per experiment, with the locked plan and the result | Not listed | YesPublished researchSource | Not listed | Not listed |
| Free report to download | YesOpen dataset releases and a CSV for each study | Not listed | Not listed | YesReport available to downloadSource | Not listed |
Where each of them shines
A recognized research program with the Profound Index and an AEO Report. Compare in full
Regular industry reports and published GEO experiments. Compare in full
A detailed State of AI Search report for download.
An AI Visibility Index of about 525 million prompts across many engines, with no setup. Compare in full
Questions about AI search research and statistics
Where does the data come from?
From answers Proofsource collected by asking buyer questions, questions that do not name a brand, on ChatGPT, Perplexity and other AI engines, with every answer and its sources stored.
How often do AI answers cite the brand's own website?
In our October 2026 study of 1,541 buyer answers across ChatGPT, Perplexity, Claude and Google AI Overviews, 0.8% of citations pointed to the brand's own website. Most citations went to other sites, such as community, video, review and reference sites.
Which engine cites the most sources?
In the October 2026 study, Perplexity averaged 14.7 citations per buyer answer and Claude 14.1, while Google AI Overviews averaged 7.4 and ChatGPT 4.7.
Can I use the figures in my article or deck?
Yes. The research is free to cite and reuse under CC BY 4.0, with a link to the study. Each study gives a suggested citation.
Can I download the raw numbers?
Yes. Each study offers its data as a CSV.
Do the results describe all of AI search?
No. They describe the scans in each study: its brands, questions, engines and dates. Each study states its limits so the figures are not read beyond them.
Can I see the same measurement for my own brand?
Yes. Proofsource runs the same measurement on your own brand and buyer questions, on every engine your plan covers, and shows which sites cite you and which leave you out.
Keep exploring
Try it on your own brand.
Proofsource is live: scan your brand and see AI search research and statistics at work with your own answers, competitors and sources.
Google AI OverviewsGoogle AI Mode