If you’ve been following SEO over the past two years, you’ve probably heard that your website needs an llms.txt file to appear in ChatGPT, AI Overviews, or Perplexity. Many agencies have even created services around it, and some content management systems now offer one-click options to add it.
But Google has repeatedly said that you don’t need llms.txt for visibility in Google Search or its AI features.
This isn’t a one-time comment. Google has consistently maintained this position through its Search Central documentation and statements from Search Advocate John Mueller and Analyst Gary Illyes. If you’re wondering whether you should spend time creating and maintaining this file, this article explains what Google has said, why it takes this position, and what actually helps improve visibility in AI-powered search.
What Is llms.txt, Anyway?
Before looking at Google’s position, it helps to understand what llms.txt is meant to do. Proposed in 2024, llms.txt is a plain-text Markdown file placed at the root of a website, similar to robots.txt. The idea is that website owners can use it to give AI models a short, organized summary of their most important content. This would help AI systems understand a website without crawling every page.
On paper, the idea makes sense. Robots.txt tells crawlers what they can access, while sitemap.xml shows search engines which pages exist. So why not create a file that tells AI models what a website is about?
The problem is that, unlike robots.txt or sitemap.xml, llms.txt has not been adopted as a technical standard by major search engines or AI companies. It is only a voluntary proposal, and it matters only if the systems it targets actually use it.
What Google Has Actually Said
Google’s position has been remarkably stable, and it comes from multiple people on the Search team over an extended period.
John Mueller compared it to the keywords meta tag. This is the most telling comparison Google has made. The keywords meta tag was, for years, a field where site owners could stuff whatever terms they wanted their page to rank for. Search engines stopped using it more than a decade ago for one simple reason: it’s entirely self-reported. A site can claim to be about anything, regardless of whether that claim is true. Mueller’s point is that llms.txt has the exact same flaw. It’s a text file where a website tells an AI system how great and relevant it is, and there’s no mechanism to verify any of it. Every site can claim to be the best, the most authoritative, or the most comprehensive resource on a topic, which means the claim carries no real signal value at all.
Mueller stated plainly that no AI system was fetching the file. In one of his more direct comments, Mueller noted that when he checked server logs, consumer AI chatbots and crawlers were fetching and reading ordinary web pages for training and grounding, but none of them were requesting llms.txt. If AI systems aren’t reading a file, then optimizing it can’t influence how those systems respond, no matter how well-crafted the content inside it is.
Gary Illyes confirmed Google isn’t building support for it. At Search Central Live events, Illyes reiterated that Google does not support llms.txt and has no plans to. He emphasized that ranking in AI Overviews and AI Mode depends on the same fundamentals as ranking in traditional Search, not on a separate AI-specific signal file.
Google’s own documentation addressed it directly. Search Central’s help content includes guidance clarifying that publishers do not need special AI files, llms.txt included, to appear in AI-powered search experiences like AI Overviews or AI Mode. This wasn’t a passing mention. It was Google actively pushing back against a piece of SEO advice that had become widespread enough to warrant an official response.
There’s also an interesting footnote to this story: llms.txt files briefly appeared on some of Google’s own properties, which some in the SEO community took as a quiet endorsement. Mueller later clarified this wasn’t the case. The files existed because Google’s internal content management system happened to support them, not because the Search team had decided the file mattered. Some teams simply hadn’t gotten around to removing it.
The Data Backs Up Google's Position
It’s one thing for a Google spokesperson to say a file doesn’t matter. It’s another to see that confirmed independently, and multiple third-party studies have done exactly that.
One widely cited analysis tracked AI bot requests across tens of thousands of domains over a 90-day period. Out of more than 62,000 recorded AI bot requests, only a tiny fraction, well under one percent, ever touched the llms.txt file. For context, that’s a lower hit rate than an average, unremarkable page buried deep in a site’s folder structure. Separate reporting on crawler activity found that the requests llms.txt files did receive were mostly from SEO audit tools and technology-detection bots checking whether the file existed, not from AI systems actually retrieving it to answer a query.
In other words, the file isn’t being ignored by Google alone. The consumer-facing AI platforms that SEOs are actually trying to get traffic from, the tools people use to search and get answers, are not requesting llms.txt in any meaningful volume either. OpenAI, Anthropic, and Perplexity have not documented llms.txt as a citation signal or a visibility requirement.
Why This Advice Spread Anyway
If the evidence is this consistent, it’s worth asking why so many sites implemented llms.txt in the first place. Part of the answer is straightforward marketing incentive. “Google secretly reads this one file” makes for a far more compelling headline than “keep doing normal SEO.” Vendor blog posts and tool vendors had a natural incentive to promote llms.txt as a must-have, since it created a new service or feature to sell. Between the file’s introduction and mid-2026, hundreds of thousands of sites added one, largely on the strength of these claims rather than any confirmed benefit from an actual AI platform.
There’s also a more understandable psychological pull here. AI search feels new and unfamiliar, and when something feels unfamiliar, a simple checklist item like “add this file” feels like control. Compare that to the alternative, which is the unglamorous, ongoing work of building genuinely authoritative content, that’s a much harder sell.
So What Should You Actually Focus On?
Google’s guidance isn’t “don’t do anything for AI search.” It’s “the fundamentals you already know still apply.” A few things worth prioritizing instead:
- Standard technical SEO. Crawlability, clean site architecture, fast load times, and mobile usability all still matter, because AI Overviews and AI Mode pull from the same underlying Google index as regular Search results. There is no separate index or ranking system exclusively for AI answers.
- Semantic HTML and structured data. Clear heading hierarchies, descriptive alt text, and schema markup help both traditional crawlers and AI systems understand what a page is actually about, without relying on a self-reported summary.
- Genuine E-E-A-T signals. Experience, expertise, authoritativeness, and trustworthiness remain the foundation Google’s ranking systems rely on. These are demonstrated through real content quality, citations, and reputation, not through a manifest file that simply asserts them.
- Solid content that answers real questions. AI systems, including Google’s own, are trying to synthesize the most useful, accurate answer to a query. Content that thoroughly and clearly answers a question is more likely to get pulled into an AI-generated answer than content optimized around a niche file format almost nothing reads.
- Keep an eye on server logs. If you want to know whether any AI crawler is requesting files from your site, including llms.txt, your own server logs will tell you far more reliably than any third-party claim. If you ever see a specific AI platform that sends you real referral traffic actively requesting the file, that’s a legitimate, evidence-based reason to create one. Until then, there’s no signal to chase.
The Bottom Line
Google’s position on llms.txt has not wavered: it is not a ranking factor, not a requirement for AI Overviews or AI Mode visibility, and not something any major AI platform has confirmed using to decide what to cite. John Mueller’s comparison to the long-dead keywords meta tag is the clearest way to understand why. A file where a website simply declares its own value, with no way to verify the claim, is exactly the kind of signal search engines learned to discount years ago.
None of this means llms.txt is harmful. Adding one won’t get a site penalized, and Chrome’s Lighthouse tool treats a missing file as “not applicable” rather than a failure. But time spent maintaining a Markdown summary that almost nothing reads is time not spent on the technical SEO, structured data, and content quality work that both traditional Search and AI search experiences are actually built on. If a real, traffic-sending AI platform ever asks for it directly, that’s the moment to reconsider. Until then, the fundamentals remain the fundamentals.