In May 2026, Google officially confirmed that llms.txt - a file many SEO practitioners had been recommending for improving AI search visibility - is not used by Google AI Search. The statement came as part of Google's official AI optimization guide published on the Google Search Central Blog. For website owners and SEO specialists worldwide: stop wasting time on llms.txt and invest it in what actually influences AI Overviews.
The short version: Google uses the same signals - GoogleBot crawling, robots.txt, content quality, E-E-A-T - for AI Overviews that it uses for traditional search. llms.txt is not part of this system and does not influence whether content appears in AI Overviews or other Google AI responses.
What Google Said: The Official Position on llms.txt
In the official AI Search optimization guide (May 2026, Google Search Central Blog), Google directly addressed the question of llms.txt:
Google Search does not use llms.txt to determine the visibility of content in AI results. The same mechanisms that govern indexing and ranking in traditional search apply to AI Overviews and other AI features.
This has one clear implication: if GoogleBot can crawl and index a page, its content is accessible for AI Overviews. There is no separate "pass" for AI systems in the Google ecosystem. llms.txt simply isn't part of how Google evaluates or surfaces content.
Background: llms.txt is a standard developed by Anthropic (the makers of Claude) in 2024 as a way to provide language models with structured information about a website. Some platforms use it - Perplexity AI reads it in certain configurations, as do some AI agents - but Google does not.
The scale of this statement matters: Google isn't saying "we don't support llms.txt yet." The position is unambiguous - for Google Search and all its AI features, there is one indexing mechanism, and llms.txt plays no role in it.
What llms.txt Is and Why It Became Popular
llms.txt is a plain text file placed at /llms.txt in a website's root directory. It contains a structured description of the site for language models - what the site covers, which sections are most important, how content is organized, and which pages can be used for AI responses.
The analogy is intuitive: llms.txt is to language models what sitemap.xml is to search crawlers. The concept is logical - helping AI systems quickly understand a site's structure and priorities.
Why the idea caught on in SEO circles:
- AI Overviews and AI Mode were cannibalizing organic click-through rates
- Website owners were looking for ways to signal their content to AI systems directly
- Several tools and agencies started marketing llms.txt as essential for "AI SEO"
The reality is more nuanced. Different AI systems handle content retrieval differently. Google - which drives the majority of search traffic for most websites globally - does not use llms.txt.
Who actually uses llms.txt:
- Perplexity AI (under certain search configurations)
- AI agents making direct API calls to language models with browsing enabled
- Specialized B2B AI tools explicitly built to read the file
Who does NOT use llms.txt:
- Google Search - officially confirmed, May 2026
- Google AI Overviews and AI Mode
- Google Gemini in standard search mode
What Actually Influences Visibility in Google AI Overviews
If llms.txt doesn't work for Google, what does determine whether content appears in AI Overviews and other AI-generated responses?
1. Crawling and Indexation by GoogleBot
The foundational requirement: GoogleBot must be able to crawl and index a page. If a page is blocked in robots.txt, excluded via noindex, or simply not indexed - it cannot appear in AI Overviews regardless of content quality. Check your Coverage report in Google Search Console regularly.
2. E-E-A-T Signals
Google Quality Raters evaluate E-E-A-T for both organic ranking and source selection in AI Overviews. Content that demonstrates genuine expertise - author biographies with verifiable credentials, citations from primary sources, original data and observations - receives priority.
One nuance worth understanding: E-E-A-T is evaluated at the domain level, not just the page level. A strong article on a weak domain performs less well than an average article on a domain with an established reputation in its niche.
3. Structured Data
FAQPage Schema, Article Schema, HowTo Schema - these markup types help Google AI systems extract structured answers from pages. Google officially recommends them in the AI optimization guide. This isn't speculation; Schema.org markup creates a direct bridge between your content and AI response formats.
4. Content Format and Structure
- A direct, clear answer to the question in the first 50-100 words of each H2 section
- Logical H2/H3 hierarchy
- Specific data: numbers, dates, percentages
- Accuracy and timeliness - outdated information reduces AI citation likelihood
5. Topical Authority
Sites that systematically cover a niche through clusters of related content receive AI citations significantly more often than sites with isolated popular articles. Google AI prefers sources that demonstrate depth of expertise on a topic, not just breadth.
Should You Still Add llms.txt to Your Site
After Google's official statement, the question remains: should you add llms.txt for other AI systems?
Arguments for adding it:
- Perplexity AI and some AI agents do read llms.txt
- Creating the file takes 30-60 minutes and is technically harmless
- If your audience actively uses non-Google AI search tools - there's real benefit
- The standard is evolving; future broader adoption is possible
Arguments against prioritizing it:
- Google Search - the dominant source of organic traffic for most sites globally - ignores the file
- Time spent on llms.txt is better invested in E-E-A-T, FAQPage Schema, or new content
- No meaningful data yet exists showing significant traffic gains from Perplexity specifically due to llms.txt
Practical recommendation: If llms.txt already exists on your site - leave it, the harm is zero. If you haven't created it - don't let it distract from more impactful work. Focus on the factors in the previous section.
One important note: if you want to block your content from being used to train language models without blocking search indexing, use the Google-Extended directive in robots.txt. That actually does something - it controls whether Google uses your content to train Gemini.
Global SEO Implications: Key Takeaways for 2026
Google's official guidance consolidates something the best SEO practitioners already knew: there is no separate "AI SEO" layer that exists independently of good traditional SEO practice.
What this means in practice:
AI Overviews draw from the same indexed pool as organic results. The sites that rank well and demonstrate strong E-E-A-T are the same sites that get cited in AI responses. This alignment is intentional from Google's perspective - they want AI answers to come from trusted, authoritative sources, and they already have a system for identifying those.
The practical implication for your content strategy:
- Audit your existing content for E-E-A-T signals - author pages, source citations, original data
- Implement FAQPage Schema on informational pages if not already done
- Reformat key H2 sections with "answer first" structure: direct response in the opening sentence, elaboration in subsequent paragraphs
- Build topical authority through content clusters, not isolated articles
- Verify indexation in GSC - pages not indexed cannot appear in AI Overviews
None of this requires llms.txt.
FAQ
Does llms.txt help with Google AI Overviews?
No. Google officially confirmed in May 2026 that llms.txt is not used when generating AI Overviews. Google applies the same indexing and quality evaluation mechanisms for AI results as for traditional search.
What is llms.txt and who uses it?
llms.txt is a text file at the root of a website containing a structured description for language models. Created by Anthropic in 2024. Used by Perplexity AI and some AI agents, but not by Google Search.
What actually determines visibility in Google AI Overviews?
GoogleBot indexation, E-E-A-T signals (expertise, authoritativeness, trustworthiness), structured data (FAQPage, Article, HowTo Schema), content quality and structure, topical authority in the niche.
Should I add llms.txt to my site given Google's statement?
For Perplexity and other AI search tools - yes, if your audience uses them (30-60 minutes work). For Google AI - technically harmless but not a priority. Better to invest the time in FAQPage Schema and E-E-A-T improvements.
What is the Google-Extended directive in robots.txt?
A directive that blocks Google from using your content to train Gemini, without blocking search indexation. Add User-agent: Google-Extended and Disallow: / in robots.txt if you want to prevent training use while keeping search visibility.
Conclusion
Google's official statement removes one of the more persistent myths of 2025-2026 SEO: llms.txt is not a shortcut to AI visibility in Google. The factors that actually work are the same principles that have always driven good SEO: technical accessibility for crawlers, demonstrable E-E-A-T, structured data, and topical authority.
For SEO practitioners and website owners globally - this is straightforward confirmation of direction. Invest in FAQPage Schema, strengthen E-E-A-T signals, build content clusters. The path to Google AI visibility runs through the same practices that produce good organic rankings.
Related articles: Google's Official AI Search Guide: AEO and GEO Are Still SEO | How to Optimize for Google AI Overviews
