Generative engine optimization statistics: our original research
This page collects our own data studies on generative engine optimization and AI search. Each study measures one thing, such as how many top sites block AI crawlers or which sources AI answers cite, and publishes its method, sample size and test dates alongside the results. We do not repeat other people's numbers here; every figure on this hub comes from a test we ran ourselves.
Updated October 21, 2026
What the study program measures
The program focuses on questions where we could not find reliable, current public data. Each topic becomes its own study, and some will be re-run so changes over time are visible.
| Topic | The question it answers |
|---|---|
| AI crawler access | How many leading websites allow or block AI crawlers such as GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and the Google-Extended token in robots.txt, and whether they treat training bots differently from search bots |
| Schema adoption | How often sites use structured data such as Organization, Product, FAQPage and Article, and whether it matches the visible page |
| llms.txt adoption | How many sites publish an llms.txt file, and how many of those follow the format at llmstxt.org |
| Cited sources | Which kinds of sources ChatGPT, Perplexity, Gemini and Google AI Overviews cite for buyer questions: brand sites, review sites, forums, news or reference sites |
| AI Overview presence | How often Google shows an AI Overview by query type, such as informational, commercial, local and branded searches |
If you want the background on why these things matter, start with our guide to generative engine optimization or the overview of Google AI Overviews.
How the studies are run
Every study on this hub follows the same rules:
- We say exactly what was tested. Which sites, which prompts or queries, which AI engines, which country and language, and the dates the data was collected.
- Sample sizes are reported per study. You will see how many domains, pages, prompts or queries were included, and how the sample was chosen.
- The method is published with the results. Each study has a methodology section explaining how data was collected and classified, and what was excluded.
- Limits are stated plainly. AI answers vary between runs and change over time. Where that affects a result, the study says so.
- No borrowed numbers. If a study mentions someone else's figure for context, it is clearly labeled and linked to its source, and kept separate from our findings.
- Corrections are dated. If we find an error, we fix it and note the change and date on the study page.
The same checks behind these studies are used in our free AI visibility checker, so you can run them on your own site.
How to cite our research
Journalists, researchers and bloggers are welcome to use and quote the findings.
- Cite the individual study page, not this hub, so readers can see the method and sample behind the number.
- Include the test dates where you can. AI search changes quickly, and a figure from one quarter may not hold the next.
- Suggested format: Study title, aisearchoptimization.net, month and year, followed by the study URL.
- Charts and tables may be reused with a link back to the study.
- Questions about method or raw data: contact us through the site and we will answer what we can.
Who runs the studies
The research is run by the team behind our AI search optimization services. We run the studies because the same questions come up in client work: which bots to allow, which schema matters, whether llms.txt is worth adding, and where AI engines get their answers.
Studies are listed below as they are published. If you are here because you want those fixes made on your own site, see how our AI SEO agency works, or read AI visibility for a plain-English primer.
Published studies
The first studies are in progress. To get them as they publish, email hello@aisearchoptimization.net.
Frequently asked questions
Where do the statistics on this page come from?
Every figure in the studies listed here comes from tests we ran ourselves. Each study page explains what was tested, the sample size, the dates and the method.
What does the study program measure?
It covers AI crawler access in robots.txt across leading sites, schema adoption, llms.txt adoption, which sources AI answers cite, and how often Google shows AI Overviews by query type.
Can I use your data in an article?
Yes. Please cite and link the individual study page and include the test dates, since AI search results change over time.
How often are the studies updated?
Some studies will be re-run so changes can be tracked over time. Each study shows its collection dates and notes any updates or corrections with the date they were made.
Why don't you include statistics from other sources?
This hub is for our own original research, so readers can see the method behind every number. Where a study refers to outside data for context, it is labeled and linked to its source.
See where AI search leaves you out
Run the free AI visibility audit. If you want the fixes done for you, the plan is a flat $299/month.