Hardly any GEO guide goes without two tips: create an llms.txt and add structured data for AI. Both sound technical and quick to do. The question is whether they achieve anything.
Here you get Google’s own answer, an honest assessment of both formats, a decision aid and what helps instead: an entity map.
Key takeaways
- Google says clearly: Google Search doesn’t use llms.txt, and AI features need no special schema.
- Structured data still makes sense: for rich results and as a clear statement of who is behind a page.
- llms.txt is a proposal by Jeremy Howard from 2024, not an official standard.
- What really counts: crawlable, indexed content and a brand that is described the same way everywhere.
- Effort vs. benefit: An llms.txt costs little. But don’t expect any visibility in Google from it.
- Better than any file: an entity map. It defines which terms your brand owns, how you define them and where the evidence is.
On this page:
- What Google says
- Structured data: what it is good for
- llms.txt: origin and structure
- Do I need an llms.txt?
- What helps instead: the entity map
- FAQ
- Sources
What Google says
In July 2026, Google published its own guide to AI features in Search. It clears up several common recommendations. On llms.txt and similar files it says briefly: “Google Search itself doesn’t use them.”
| Recommendation from GEO guides | What Google says |
|---|---|
| Create an llms.txt | Not needed Google Search doesn’t use it. |
| Special schema for AI | Not needed There is no special markup for AI features. |
| Chop texts into small sections | Not needed Google recognizes several topics on one page. |
| Markdown versions of pages | Not needed Neither helps nor harms visibility in search. |
| Maintain structured data | Still useful as regular SEO, e.g. for rich results. |
For Google there is no AI trick, only good SEO.
In short: AI Overviews and AI Mode have the same requirements as regular search. The page must be crawled, indexed and eligible to show with a snippet.
Structured data: what it is good for
Structured data based on schema.org describes in machine-readable form what is on a page: an article, a product, an organization, a person. Google uses it for rich results. According to Google, AI features need no additional markup.
ORGANIZATION: Example Ltd ├─ sameAs → LinkedIn ├─ sameAs → Wikidata └─ sameAs → commercial registerone name, many profiles
I still see a benefit: structured data makes it unambiguous who is behind a website and which profiles belong to the same entity. The sameAs field in the Organization type connects the website with LinkedIn, Wikipedia or the commercial register.
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Example Ltd",
"url": "https://www.example.com/",
"logo": "https://www.example.com/logo.png",
"sameAs": [
"https://www.linkedin.com/company/example",
"https://en.wikipedia.org/wiki/Example_Ltd"
]
}
</script>On bernard.digital I use two more types: DefinedTerm for every term in the GEO glossary and Dataset for the studies, such as the keyword domain study. Both describe the content more precisely; neither is a ranking promise.
llms.txt: origin and structure
Jeremy Howard proposed the llms.txt on 3 September 2024 at llmstxt.org. The idea: a small text file in the root directory that tells a language model in Markdown what matters on the website and links to clean Markdown versions of the pages. It is meant to complement the sitemap.xml and robots.txt, not replace them.
# Name of the website required, nothing else > Short summary optional ## Sections with links - [Page](url): note optional ## Optional links that may be skipped
# Example Ltd
> Tax advice for founders in Berlin.
## Services
- [Company formation](https://www.example.com/formation.md): process and costs
- [Bookkeeping](https://www.example.com/bookkeeping.md): packages and prices
## Optional
- [About us](https://www.example.com/about.md)Watch out: in my research I found no solid evidence that ChatGPT, Claude or Perplexity use the llms.txt for their search answers. Google has said explicitly that its search doesn’t use it.
Do I need an llms.txt?
DOCUMENTATION FOR DEVELOPERS? yes → create llms.txt coding assistants often read it MARKETING WEBSITE? optional, costs little does nothing in Google
My recommendation: create an llms.txt if it costs you less than an hour, and treat it as an experiment. Your GEO time is better spent on an entity map, on content that gets cited and on mentions on other sites.
From my practice
For clients, I have used llms.txt ever since I learned it exists. For my own projects, I decided against it: it costs time and probably achieves nothing. For several clients I connected the server logs via Peec so we can track visits from LLMs. On top of that there is LLM referral tracking in Analytics. It only works moderately well, but it is still useful as an indicator of growth and trends. I see the bigger lever in the entity map. Models probably don’t read it directly either. But it creates clarity about the entities, on the client’s website and in their content, and that is exactly what LLMs benefit from.
What helps instead: the entity map
An llms.txt tells an AI which pages exist. It knows that from search anyway. What it lacks is something else: which terms your brand owns, how you define them and where the evidence is. That is exactly what an entity map captures.
LLMS.TXT which pages exist a list, no connections ENTITY MAP GEO ─depends on─▶ SEO ├─ achieves ─▶ AI visibility └─ evidence: passage + URL what your brand knows
How to build your own entity map, which free tools help and what my map for bernard.digital looks like is covered in the chapter Entity map: your brand as a map.
The map is the work, the file is a by-product.
FAQ
What is an llms.txt?
A text file in a website’s root directory that gives language models an overview of the most important content in Markdown. It was proposed by Jeremy Howard in 2024 and is not an official standard.
Does Google use the llms.txt?
No. According to Google’s guide to AI features, Google Search doesn’t use llms.txt or similar files. They neither help nor harm visibility in search.
Do I need schema.org for AI Overviews?
No. Google says there is no special markup for AI features. Structured data remains useful as regular SEO, e.g. for rich results.
Where do I put the llms.txt?
In the root directory of the domain, so that it can be reached at yourdomain.com/llms.txt.
What is an entity map?
An overview of the terms a brand owns, with definitions, evidence from its own website and the relations between the terms. It makes content and signals consistent, whether or not you publish it as a file.
Next in the handbook: That completes the technical foundation. Part 3 is about the content itself and why AI understands brands as entities. Until those chapters are out, Measuring AI visibility shows how to check whether your work is paying off.
Sources
- Google Search Central: AI features and your website, July 2026
- Jeremy Howard: The /llms.txt file, September 2024
- Schema.org: Organization
- EntityMap: Specification v1.0, Fred Laurent and Dixon Jones
