Adding llms.txt Won't Get You Found in AI Search. Here Is What Does.

If you run a small business, there is a fair chance someone has recently told you to add a file called llms.txt to your website so that ChatGPT, Gemini and Claude will start recommending you. It sounds plausible, it is quick to do, and a growing number of agencies are charging for it. This week, Google's own search team poured cold water on the idea.

On Search Off the Record, the podcast run by Google's Search Relations team, John Mueller explained why a file like llms.txt cannot help an AI model decide which website to surface for a given question. His reasoning is simple, and worth understanding before you spend a cent on it.

The short version: llms.txt is self-reported. It is you telling the AI how good your own site is, and a self-reported claim is exactly what an AI cannot lean on when it is choosing between you and a competitor making the very same claim.

What llms.txt actually is

llms.txt is a proposed standard, first floated in September 2024, for a plain text file you place at the root of your site (so it would sit at yoursite.com.au/llms.txt). In it you list and summarise your most important pages in a tidy, machine-readable form, the idea being that an AI model can read that instead of crawling your whole site. On paper it sounds like a sitemap for the AI era, and as a way of laying out your content neatly, it is harmless enough.

The trouble is the gap between what it is being sold to do and what it can do. Plenty of businesses are being pitched llms.txt as a way to get discovered or ranked inside AI assistants. That is the part Google says does not hold up.

Why Google says it won't get you found

Mueller's point is that a file like this is essentially you telling the AI systems that you have the best website around. There is nothing inside an llms.txt file that lets a model pick your site over a competitor's, because every competitor can write the identical thing about themselves. An AI works out who to trust from signals it cannot fake on your behalf, not from a description you wrote about your own business.

This matches what Google has said repeatedly elsewhere: it does not use llms.txt as a ranking signal for Search, and no major AI search engine has adopted it for discovery either. Mueller also noted that the broader question of how AI agents navigate websites is still unsettled, with several competing proposals on the table and, by his estimate, six months to a year or more before anything becomes a genuine standard. In other words, this is not a settled playbook you are missing out on. It is an argument that has not finished.

There is nothing inside a file you write about yourself that tells an AI to trust you over the business next door. That trust is earned off the page, not declared on it.

Where llms.txt might still earn its keep

Mueller did leave one narrow door open, and it is worth being fair about it. Once an AI agent is already on your site, say it has arrived to complete a task like making a booking or a purchase, a clear, structured map of your pages may help it find its way around. That is a navigation aid for an agent that already chose you, not a way to get chosen in the first place. So if adding the file is quick and free, there is little harm in it. The mistake is paying real money for it as an AI ranking tactic, or believing it replaces the work that actually moves the needle.

What actually gets you found in AI search

The uncomfortable truth is that getting recommended by an AI assistant runs on much the same fundamentals as ranking on Google, because most AI answers still lean on web search underneath. We have written a fuller guide to AI search visibility, or GEO, if you want the whole picture. Here is the short, practical list.

• Get cited by sources the AI already trusts. Mentions on reputable local sites, industry directories and press do the heavy lifting, precisely because they are signals you cannot write about yourself.
• State your facts in plain, readable text. Your services, locations, hours, prices where appropriate and contact details belong in clear text on the page, not locked inside an image or a PDF, so an AI can quote them accurately.
• Answer the questions customers actually ask. Pages that clearly answer best plumber in your suburb or product A versus product B are what get pulled into AI answers.
• Keep your Google and Brave rankings strong. Assistants still draw heavily on traditional search results, so the SEO you already understand feeds straight into your AI visibility. We covered how Claude leans on search rankings in a recent post.
• Treat it as ongoing, not a one-off. AI tools change who they cite constantly, and a file you set once and forget was never going to keep pace with that anyway.

By all means add an llms.txt file if it is quick and costs you nothing. Just do not let anyone sell it to you as the thing that gets you found in AI search, and do not let it distract from the work that genuinely does.

This is exactly the noise we cut through at NextAura. We keep watch on what actually moves AI search visibility and what is just this month's myth, then build and maintain the SEO and AI search foundations that get Australian businesses named inside the tools their customers are already asking. If you would rather hand the optimising and the keeping-up to people who track this daily, get in touch and we will carry it while you focus on running the business.