A critical survey asks what GEO research actually proves

What happened?

A new survey posted to arXiv on July 15 reviews 45 studies of generative engine optimization published since the field was named in late 2023. The author, Olivier Martinez, finds that the two most reliable levers are topical relevance and position, that generic tricks transfer poorly between engines and topics, and that rewrites aimed at winning citations can even make a page less likely to be retrieved in the first place. The survey is a preprint and has not yet been peer reviewed.

Why does it matter?

The survey puts the most quoted number in GEO into context. The up to 40 percent visibility gain from the founding 2024 study describes what happens after a page is already among the handful of sources an engine is reading, not the odds of being found in the first place. It also reports that no reviewed technique shows a stable, cross-platform effect on discoverability or traffic. The levers that do hold up are unglamorous: answer the question directly, provide verifiable facts, and make sure your page can be crawled and retrieved at all.

Should I do anything?

Treat this as permission to ignore anyone selling a trick that guarantees placement in ChatGPT. Instead, pick your single most important page and check the two things the research says matter most: does it answer its core question directly near the top, and can AI crawlers actually reach it? Our guide on whether AI tools can see your website walks through the second check in a few minutes. And when you test your own visibility, repeat each question a few times, because answers vary from run to run.

Learn more

• The survey on arXiv
• Our guide: can AI tools see my website?
• Our guide: what makes content citation-worthy