AI-friendly websites · What Google says
Google's rankings are the most valuable trade secret on the internet, and Google guards them the way any business guards the thing it sells. Not to be difficult. Because that is the business. So every statement is worded to give nothing away, and the only reliable tell is what Google builds for itself. Below: the rule proved three times over, then run on one file, llms.txt, which Google Search says it does not use and nine Google properties publish anyway.
What Google Search says: one careful line, about one product.
What Google builds: the file, on nine of its own sites.
Why Google talks this way
Rankings pay for everything at Google. The recipe is the product, and a company does not hand its product to the people trying to game it. So every public statement about ranking is written by people who know that, and the statements have a shape: careful, literally true of the thing they name, and silent about a system inside Google with a different name that does the job the question was about.
That is not a theory. Here are three times it was settled, by sworn testimony or by Google's own documents, and in every case the public line had been true of the word Google chose.
01
02
03
So the rule. When Google says a thing does not affect Search, believe it about Search. When Google's own properties do the thing anyway, do it too, because those are the hours Google chose to spend when nobody was asking, and Google does not spend engineering hours on decoration.
The rest of this page runs the rule on one file: llms.txt, the plain-text account of a site written for the machines that read it. Google Search says it does not use the file. Nine Google properties publish it. You can guess what we build.
In Google's words
This is the official line, from the guide Google Search Central published on 15 May 2026 and last updated in July 2026. It is the clearest thing Google has said about the file, and it was written by people who know exactly what a sentence from Google is worth.
Read it the way it was written. The subject is Google Search, three times in one sentence. The claim is that Google Search does not use the file. The same page then says it is completely fine to maintain one for other systems that do. That is not a company dismissing the file. That is a company being precise about one product and saying nothing at all about the rest of itself, which is the rule at the top of this page, in action.
Read the guide“You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them.”
In order
Every quote below is verbatim from the linked source, with its date. Nothing has been sharpened, because it does not need to be. Watch two things: how the scope never once widens past Search, and how the one time Google was asked about its own files, the answer was the best argument anyone has made for a business publishing one.
17 April 2025
John Mueller, Google Search
“To me, it's comparable to the keywords meta tag”
The opening position, and a fair one for a search engine: a file where a site owner says what the site is about is a claim, and Google checks claims against the site. He added that server logs showed no AI service asking for the file yet. Yet is the word doing the work.
20 January 2026
John Mueller, Google Search
Bluesky
“I'm tempted to say something snarky since this has come up so often, but to be direct, no.”
Asked point blank whether the llms.txt files on Google's own properties were an endorsement. Read the shape of the answer: it tells you what the Search team thinks, and the files stayed up. The Search team removed the one on its own documentation. Nobody else at Google did.
15 May 2026
Google Search Central
The guide to optimizing for generative AI features on Google Search
“You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them.”
The official line, written down. Count the times Google Search appears in one sentence. The same page says it is completely fine to create and maintain the files for other services or systems that use them, which is Google leaving the door open while telling you it has nothing to do with the room behind it.
20 May 2026
John Mueller, Google Search
Bluesky, answering Lily Ray
“The short answer is that it's not done for search. There's more to websites than just SEO.”
The one time Google was asked why its own sites publish the file, and the answer is the whole case for a business having one. Coding assistants produce better work when they can read a plain account of the thing, so Google's teams give them one. He called it a crutch that saves some tokens and said non-developer sites did not need it. Hold on to the first half: a machine reads you better from a plain file. Your customer's assistant is a machine.
2 June 2026
John Mueller, Google Search
“When an AI platform that brings you clients complains that it needs the file for your site, then I'd recommend taking the time to create one.”
Google's own rule for when to publish: the day a platform that sends you customers wants it. The trouble with that rule is that the platform never calls. It quotes whoever it can read and moves on. We would rather be ready before the complaint than after it.
15 June 2026
Google Search Central
The documentation changelog
“while these files aren't needed for Google Search (and won't negatively or positively impact your visibility or rankings), it's fine if you want to maintain these files for other services or systems that use them.”
The clarification, added because people kept asking. It is also the best news on this page for anyone weighing the file: there is no penalty. Google has said in writing that publishing one costs you nothing in Search. The guide was last updated on 10 July 2026 and still says so.
What Google does
On 18 September 2026 we fetched /llms.txt on every Google domain we could think of and read what came back. These nine answered with a real file: an H1 with the name, a summary, and links, generated from the property's own documentation. Two publish a long companion at /llms-full.txt as well, which is the same split this site uses, because we copied it.
We were strict about it. Five more Google domains answer 200 at the address with a web page, which is the shape our checker fails a business for, so they are not in the table. Neither is anything that answered 404, three of which are worth reading below.
| Property | Address | First line | Size |
|---|---|---|---|
| Google DeepMindThe research lab's own site | deepmind.google/llms.txtand llms-full.txt | # Google DeepMind | 7 KB |
| Gemini APIDeveloper documentation | ai.google.dev/gemini-api/docs/llms.txt | # Gemini API | 34 KB |
| FirebaseDeveloper documentation | firebase.google.com/docs/llms.txt | # Documentation | 1.8 MB |
| Firebase StudioDeveloper documentation | firebase.google.com/docs/studio/llms.txt | # Firebase Studio | 9 KB |
| AndroidDeveloper documentation | developer.android.com/develop/llms.txt | # Develop for Android | 284 KB |
| AngularThe framework's site | angular.dev/llms.txt | # Angular | 7 KB |
| GenkitDeveloper documentation, on developers.google.com and genkit.dev | developers.google.com/genkit/llms.txt | # Genkit | 7 KB |
| Agent Development KitDeveloper documentation | google.github.io/adk-docs/llms.txtand llms-full.txt | # Agent Development Kit (ADK) | 17 KB |
| StitchA design tool from Google | stitch.withgoogle.com/llms.txt | # Stitch with Google | 3 KB |
Fetched with a plain GET on 18 September 2026, following redirects, and read by content type and first line rather than by status. Firebase and Android are far past the 50 KB we hold a business file to, because they are whole documentation sets for a coding agent. A business file has a smaller job and does it in a page: the name, the number, the place, the prices.
Lighthouse 13.3, the page-quality tool built into Chrome, added a category called Agentic Browsing to its default run on 7 May 2026. It reached every Chrome user's developer tools with Chrome 150, stable on 30 June 2026. One of its audits is the file the Search team says to ignore. Google built a grader for it. Read that twice as well.
To be exact about our own claim: we are not saying Google Search secretly reads llms.txt. Every sign is that it does not, and we take the Search team at its word. We are saying Google is being Google: exact about Search, quiet about the rest, and busy building the thing anyway. The company that will never tell you what ranks has told you the one way it cannot help telling you, through what it builds for itself.
What we do about it
The two are about different things, which is why both fit in one sentence, and why every site we ship does both.
The file will not move a Google ranking, and it will not hurt one. Anyone selling an llms.txt as a way up Google's results is selling against Google's own documentation, and you should walk. We do not sell it that way. We sell it the way Google builds it.
Nine Google properties publish a plain account of themselves for the machines that read them, generated from their own content, never typed twice. Every site we ship does the same on launch day: one route, written from the site's own pages, so the phone number and the prices an assistant quotes are the ones on the site that morning.
The file is one of nine checks a stranger can run on any address, and the other eight are things Google Search does read: the crawlers admitted by name, structured data that says what the page says, answers first, a fast page underneath. Nobody in the 281 trades businesses we measured passed all nine. Every site we build does.
For the record, since a good pitch survives the footnote: Ahrefs read the server logs of 137,210 domains and published on 15 June 2026. 28% had a file and 97% of those files were not requested in May 2026. That is what a new address looks like before the readers arrive, and the readers who did arrive were the assistants and coding agents Google publishes for. Being ready costs a route. Being absent costs a customer you never hear about. Ahrefs.
Straight answers
Sources
Every quote and figure on this page, where it can be read. We would rather you checked. The file list is our own fetch on 18 September 2026; repeat it with any HTTP client and read the first line of each body.
The free checker fetches your home page, robots.txt and llms.txt and scores the nine checks, each with what it found. About ten seconds, nothing stored. Google will never tell you what ranks. We will tell you what your site says to everything that reads it, and what we would build.
Or skip the form entirely. hello@pinelandtech.com reaches a person, not a queue.