Short answer
Content gets cited in AI-generated answers when it is technically easy to read, answers the question directly within the first hundred words, contains specific numbers with a source, and is mentioned on other credible sites. Neither llms.txt nor extra schema markup is required — Google has said so explicitly.
There is a lot of vendor-driven noise around GEO right now. This article separates what has support in published data from what is wishful thinking sold as method.
What it takes to get cited
Four attributes recur across studies of which sources generative systems actually surface: the page can be read without JavaScript, the answer appears early and self-contained, the text contains specific numbers with a stated source, and the site is mentioned in places the system already trusts.
What you can ignore
Google published its first dedicated AI search guidance in May 2026 and called both llms.txt and special schema markup unnecessary for AI Overviews. A study the same month found no measurable lift in citations when JSON-LD was added to pages already appearing there.
The conclusion is not to stop using structured data. It still drives classic rich results and helps entity resolution. The conclusion is not to call it a GEO strategy.
What to do instead
- Render statically so the content exists in the first response
- Write a self-contained 40–60 word direct answer at the top
- Attribute named expertise with real credentials
- Use specific numbers and link the source
- Keep content current and show the date
Common questions
- Do I need llms.txt?
- No major search engine requires it and Google has called it unnecessary. It also costs almost nothing to generate and some agent frameworks read it, so publishing it is fine — just do not build a strategy on it.
- Is GEO different from SEO?
- Google itself describes GEO and AEO as "still SEO". The fundamentals are the same; what differs is that the goal is a citation rather than a click.
Websites, products and brands — built in weeks
Torn Studio builds digital products with AI tooling. You get the same quality in a third of the time, and you always talk directly to the person building it.