Google Search does not use llms.txt for ranking or AI Overview citations, and says so directly. Google Lighthouse 13.3 checks for it anyway, because that audit is about AI agents browsing your site, not about search. Both positions are correct; they describe different layers.
A new file format called llms.txt has been generating noise in the SEO world for about a year. The idea is simple: place a Markdown-formatted summary of your website at yourdomain.com/llms.txt, and AI systems can read it to understand your site faster.
Whether it matters for your search visibility is another question entirely, and right now, two different Google teams are giving two different answers.
What has Google Search said about llms.txt?
Google’s Search team has been consistent on this for over a year: llms.txt is not something Google uses, recommends, or plans to adopt for search.
John Mueller, Google’s Search Relations lead, compared llms.txt to the keywords meta tag, a file bots don’t actually request and AI services don’t actually use. He called building separate Markdown pages for AI crawlers “a stupid idea.”
In May 2026, Google’s Search Central team published a formal GEO optimization guide listing llms.txt alongside content chunking and AI-specific rewrites as tactics you explicitly don’t need for visibility in AI Overviews or AI Mode. Gary Illyes and Amir Taboul at Search Central Live confirmed the same: Google Search is not pursuing llms.txt.
The message from the Search team has been clear: if you’re adding llms.txt to improve your GEO rankings, you’re solving the wrong problem.
Why does Google Lighthouse 13.3 check for llms.txt?
Here’s where it gets interesting.
On May 7, 2026, days before that Search Central guide published, Google’s Lighthouse team shipped version 13.3, which added a new Agentic Browsing audit category. One of the audits in that category: an llms.txt check.
The Chrome developer documentation for this audit says:
“The llms.txt file is an emerging convention used to provide a machine-readable summary of a website’s content, specifically designed for LLMs and AI agents. Without this file, agents may spend more time crawling the site to understand its high-level structure and primary content.”
Lighthouse 13.3 moved this Agentic Browsing category from experimental into the default config. PageSpeed Insights inherited it within two weeks. Chrome DevTools will follow in the next stable release.
So: Google’s Search team says llms.txt doesn’t matter. Google’s Lighthouse team is now checking whether your site has it by default.
Why are both Google teams right at the same time?
This isn’t a contradiction once you understand the two different contexts.
Google Search’s position is about discovery, whether llms.txt influences how Google finds and ranks your pages in its search index. The answer there is clearly no. llms.txt does not affect rankings, does not influence AI Overview citations, and does not help with traditional GEO.
Google Lighthouse’s position is about functionality, specifically, how well your site works for AI agents that are acting on a user’s behalf: browsing, extracting information, completing tasks. This is the agentic web: the layer where AI agents fetch context from your site to answer questions or take actions, not the layer where Google’s ranking systems decide which pages surface in search.
John Mueller actually articulated this distinction when pressed: llms.txt isn’t for search purposes, but it’s worth separating “discovery” (finding your website via search) from “functionality” (helping users once they’ve found you do what they came to do).
Two different Google product teams. Two different use cases. The guidance is consistent once you know which layer you’re talking about.
What does the Lighthouse llms.txt audit actually flag?
This matters practically. The Lighthouse llms.txt audit does not fail your site for missing the file. A 404 response on /llms.txt marks the audit as Not Applicable, that’s the default state for most sites today.
What it does flag is a server error (5xx response) when Lighthouse tries to retrieve the file. If your server is misconfigured in a way that errors on that path, that gets marked as a problem.
The score impact is currently zero, Agentic Browsing doesn’t factor into your overall Lighthouse score the way Performance or SEO audits do. It’s a data-gathering phase.
But the category is now on by default in PageSpeed Insights. It’s visible to anyone who runs an audit on your site.
What actually moves AI citations, if not llms.txt?
If llms.txt isn’t the lever, what is?
The research and guidance here are consistent: GEO, ranking in AI Overviews, getting cited by ChatGPT, Perplexity, and Gemini, runs almost entirely through the same signals as traditional organic SEO.
Entity schema that says who you are
LocalBusiness markup with the right subtype, RealEstateAgent for a property manager, tells AI systems who you are and what questions you answer. This is the single highest-ROI GEO intervention for local property managers.
Your most important claims in the first third of the page
Research from Princeton and Georgia Tech, published in 2024, found that 44% of AI citations come from the first 30% of page content. Lead with the claim, not the wind-up. This is the same discipline as answering a heading in its first two sentences, applied to the page as a whole.
Organic ranking, because retrieval comes first
AI Overviews and AI Mode pull heavily from pages that already rank. If you are not on page one for the query, you are unlikely to be in the candidate pool at all, which makes every extraction improvement downstream of ordinary ranking work.
Question-shaped content with matching markup
Question-shaped headings answered immediately underneath are the clearest signal that your page addresses a specific query, and FAQPage markup makes that structure explicit. It earns no rich result and never will; the value is entirely in extraction.
A study of 300,000 domains found no statistical correlation between having an llms.txt file and being cited by AI systems. The correlation with organic ranking is far stronger.
Should you add an llms.txt file to your site?
Adding a basic llms.txt file takes about 15 minutes. It’s not a ranking move, but it’s also nearly zero cost, and the agentic web (AI agents acting on behalf of users, not just search crawlers indexing pages) is the direction browsers and AI assistants are clearly moving.
If your server would error on that path, fix it. A misconfigured server is always worth fixing.
If your site is missing entity schema, structured FAQs, and clear content hierarchy, that’s where the ranking gains are. Fix those first, starting with complete LocalBusiness schema.
The llms.txt debate is interesting because it exposes the split between two different futures happening simultaneously: the search ranking future (where the signals haven’t changed as much as the hype suggests) and the agentic browsing future (where a different set of readiness signals is just beginning to matter). Google’s two teams are each pointing at one of those futures.
Both are worth preparing for. Just in the right order.
All GEO findings above draw on Google’s May 2026 official optimization guide, Chrome Developer documentation (Lighthouse 13.3, published May 5, 2026), and Search Engine Journal’s reporting on the documentation gap between Google Search and Chrome teams. A free audit shows how your site is performing on the signals that move AI citations: entity schema, FAQPage, and organic ranking against your real competitors. The audit is the next step.