There Is No Magic File That Makes AI Cite You. Do This Instead.

There is no magic file

Somewhere you read that if you drop an llms.txt or an author.txt file on your site, or add the new content-signal line to your robots file, AI tools will start citing you. So you added them, felt productive, and waited. Nothing happened, because the file was never the lever.

On July 6, 2026, Google's John Mueller answered this directly. He said Google does not use llms.txt or author.txt, and that he knows of no other crawler or LLM confirming they use them. On the content-signal directive he was blunter, saying it has no effect whatsoever for any crawler or LLM. The shortcut everyone is chasing is a placebo. While you edit text files, the thing that actually earns AI citations sits untouched.

The file myth versus the entity

SEO trained us to believe visibility is a config setting. Add the tag, flip the switch, get found. AI citation does not work like that. A model cites what it can confirm is a real, consistent thing across the web. You do not configure your way in. You prove your way in.

Here is the concrete version. Two companies sell the same thing. Company A spends the afternoon adding llms.txt and tweaking robots directives. Company B spends it adding clean Organization schema, linking its LinkedIn, Crunchbase, and YouTube in a sameAs list, and landing one podcast mention. Ask ChatGPT about the category next month, and it names Company B. Not because B found a trick, but because B is the only one the model can actually verify.

Why AI trusts entities, not files

An AI model will not stake an answer on a name it can only find in one place. It looks for the same brand, described the same way, across sources it already trusts: your site, your schema, your LinkedIn, a directory, a mention on a page you do not control.

That web of matching signals is what being an entity means. It is how the model tells you apart from the other companies with a similar name, and how it decides you are safe to name in an answer. A file you wrote about yourself proves nothing. A footprint the web agrees on proves you exist. That is the whole game.

The play this week: become a verifiable entity

You do not need a file. You need proof you are real. Pick your homepage and spend about twenty minutes on three things.

• Add Organization schema to your homepage with your name, url, logo, and a sameAs array that links every profile you own, from LinkedIn and Crunchbase to YouTube and your other socials. Google's own documentation says sameAs is how it disambiguates you from everyone else with your name.
• Make your name, URL, and logo identical everywhere. Same spelling, same domain, same mark on your site, your LinkedIn, and any directory. A mismatch tells a model you might be two different companies, so it trusts none of them.
• Get named on one page you do not own this month. A podcast, a guest post, one interview. Mentions on sites AI already trusts are what it checks you against.

One prompt to run today

Stop guessing what AI can and cannot verify about you. Paste the prompt below into ChatGPT or Perplexity and swap in your brand and URL. It tells the model to describe what you do and who you are for, using only what it can verify on the web, list every source it would cite, and flag where it is unsure or confusing you with a different company.

If it cannot describe you, cannot cite a source, or mixes you up with someone else, that is not a content problem. It is an entity problem, and the three steps above are the fix.