WHAT IS LLMS.TXT? WHAT THE FILE DOES, WHAT IT DOES NOT, AND WHERE IT FITS IN AI SEO

LLMS.TXT IS HOUSEKEEPING. AI answers are won elsewhere.

llms.txt is a short Markdown file that offers AI tools a map of your website. Google says its search ignores the file and no major AI company has confirmed using it, so AB Media Co gives it an hour of tidy work and spends your effort on what AI answers do read. Across 100+ brands and ₹3,300 Cr+ in revenue generated for clients, our team has learned to put the hours where the evidence points.

GOOGLE, 10 JULY 2026LLMS.TXT
Google’s guide says an llms.txt file neither harms nor helps your visibility in Google Search, which ignores it. No major AI company has confirmed using one either, so we add it on request and never sell it as AI SEO.
EFFECT ON GOOGLE SEARCH VISIBILITY0
GOOGLE SEARCH CENTRAL · UPDATED 10 JULY 2026
THE BRANDS THAT TRUSTED US 100+ brands, 8+ years
Sonalika / Solis
Air India SATS
Adobe
TATA Tele Business Services
ICRA
Birlasoft
SBI Life
Radico Khaitan
Lovely Professional University
Hero
TalentGum
ISKCON
Niramaya Soul
ISTD
Juhi Nanda
Soni Cosmetics
Hotel Sunstar Group
United Coffee House
Republic of Zoofari
Onyx Neo
OCP Medical Center
Azza Fencing
Maska Mein Chaska
Gustora
Arohul
JIET
AI SEO01 TECHNICAL AI SEO02 LOCAL AI SEO03 AI SEO SERVICES04 ECOMMERCE AI SEO05 AI SEO CONTENT06 AI SEO AUDITS07 AEO AND GEO08
7-FIG AI search grew from zero sales in 2024 to a seven-figure sales channel ECOMMERCE BRAND
3x Domain authority tripled with AI SEO, paid media and PR run together TALENTGUM
500+ PR links from trusted publications, then AI SEO on the website SENSES ELECTRONICS
40x Customer acquisition in 18 months with AI SEO in the mix TALENTGUM
Akshit BhasinFOUNDER, AB MEDIA CO LINKEDIN ↗

WHY LLMS.TXT KEEPS REACHING THE BOARDROOM, and where we put the effort instead.

llms.txt tends to arrive on a leadership agenda as a quick fix: add one file and the AI assistants will finally understand your brand. The appeal is easy to see. It is cheap, it sounds technical enough to matter, and it promises some control over answers you cannot otherwise see.

The file itself is modest. Jeremy Howard of Answer.AI proposed it on 3 September 2024 as a Markdown file at yoursite.com/llms.txt, with your site’s name as a heading, an optional summary and lists of links to the pages that matter. The proposal is now in its second version, and it describes the file as something an agent reads on demand while it is helping someone.

The evidence on whether it moves AI answers points one way. Google’s guide to generative AI search says Google Search ignores llms.txt files and that creating one will neither harm nor help your visibility there. Google’s John Mueller wrote in June 2025 that server logs showed chatbots requesting pages rather than the file. And SE Ranking, across about 300,000 domains, found no measurable link between having one and being cited by AI tools.

So we treat llms.txt as housekeeping, not strategy. We will add one if you want it, keep it accurate and confirm that it loads, but it never appears as a deliverable in your plan and never comes with a promise of citations.

Before any tactic takes your money, we weigh the evidence behind it. On llms.txt that analysis is short. On crawler access, readable pages and what other websites say about you, it runs long, and that is where the effort goes.

Of the 100+ brands our team has worked with, one ecommerce brand, the first case further down, found its growth there too: a rebuilt store and tracking, buying guides that answer first, and sales from AI assistants read by source every month. In 2024, AI assistants sent it zero sales. Today AI search is a seven-figure sales channel for it.

If your real question is why AI tools leave your brand out, the file will not answer it and the free mini audit will. One of our senior people checks your crawler access, your key pages and what AI answers say about your market, personally, and sends you the five to ten fixes that would grow your revenue first.

What changes the plan

FOUR FACTS ON LLMS.TXT,
and what each one means for the file in your plan.

Two come from Google, one from a 300,000-domain study and one from Chrome’s own audit tool. Together they set how much of your plan AB Media Co gives the file, which is very little.

01LLMS.TXT: WHAT THE FILE IS, WHAT GOOGLE AND THE EVIDENCE SAY, AND WHY AB MEDIA CO TREATS IT AS HOUSEKEEPING INSIDE AI SEO

Google Search ignores the file

Google’s guide to generative AI search, updated on 10 July 2026, says you do not need AI text files, markup or Markdown to appear in Google Search, its AI features included, because Google Search does not use them. It adds that making llms.txt for other systems is fine and will neither harm nor help your visibility. So your AI Overviews and AI Mode plan stays on indexable pages, snippet eligibility and answers worth quoting.

SOURCE: Google Search Central: Optimizing for generative AI search, July 2026

02LLMS.TXT: WHAT THE FILE IS, WHAT GOOGLE AND THE EVIDENCE SAY, AND WHY AB MEDIA CO TREATS IT AS HOUSEKEEPING INSIDE AI SEO

Crawler logs show almost nobody reads it

Google’s John Mueller wrote in June 2025 that server logs show consumer chatbots fetching pages for training and grounding but not llms.txt, and the one public counter-example, a marketer’s logs in July 2025, showed OpenAI’s crawler requesting it on a few sites. Logs are evidence you already own. On any account where the file exists, we read the requests to /llms.txt before anyone spends another hour on it.

SOURCE: Search Engine Roundtable, June and July 2025

03LLMS.TXT: WHAT THE FILE IS, WHAT GOOGLE AND THE EVIDENCE SAY, AND WHY AB MEDIA CO TREATS IT AS HOUSEKEEPING INSIDE AI SEO

300,000 domains showed no link to citations

SE Ranking analyzed about 300,000 domains, found llms.txt on 10.13% of them, and found no measurable connection between having the file and how often AI tools cited a domain. In its model, removing the llms.txt variable made the predictions more accurate. So the file takes no share of the AI SEO plan we build for you. Mentions on cited websites and answer-first pages take it instead.

SOURCE: Search Engine Journal on SE Ranking’s study, 20 November 2025

04LLMS.TXT: WHAT THE FILE IS, WHAT GOOGLE AND THE EVIDENCE SAY, AND WHY AB MEDIA CO TREATS IT AS HOUSEKEEPING INSIDE AI SEO

A broken file fails Chrome’s audit, a missing one does not

Chrome’s Lighthouse includes an llms.txt check among its agentic browsing audits. A server error when fetching the file fails it, while a missing file is marked not applicable. So if you publish one, it has to return cleanly, and our technical checks fetch /llms.txt like any other URL on your site.

SOURCE: Chrome for Developers: Lighthouse llms.txt audit

WHAT YOUR TEAM SHOULD CHECK BEFORE LLMS.TXT, and how we deal with each one.

THE FULL AI SEO PRACTICE →
  1. 01

    Crawler access in robots.txt and at the CDN

    A text file cannot open a door that robots.txt or a firewall rule has closed. We check the rules for Googlebot and for OpenAI’s and Anthropic’s crawlers by name, then test at the CDN, because a security setting can block a bot that robots.txt allows.

  2. 02

    Pages that keep their answers without JavaScript

    We fetch your key pages with scripts switched off. If the price, the specification or the answer disappears, a crawler that does not run scripts never sees it, so we fix the template inside your own CMS.

  3. 03

    Openings that answer the question

    AI answers quote passages from pages, not summaries from files. Our writers put the answer in the first lines of each revenue page and the proof straight after, with your URLs unchanged and every page approved as a draft.

  4. 04

    The file itself, kept small and accurate

    If you want an llms.txt, we write it from your sitemap: your name, a one-line summary and links to the pages that matter. Every link is checked by hand, so no staging page, admin path or private document slips into a public list.

  5. 05

    Facts that agree everywhere

    Prices, locations and terms in the file, on your pages, in your profiles and in the press have to match. A stale llms.txt states the wrong fact with confidence, so it sits on the same update list as everything else.

  6. 06

    Logs and a monthly read

    We check server logs for requests to /llms.txt and for the AI crawlers that do visit, and read GA4 sessions from AI assistants beside our prompt runs. Your Share of Decision sits above it all, so the file is judged by what it changes, like every other piece of work.

WHAT MOVES AI ANSWERS INSTEAD, as our 100+ brands have shown.

The ecommerce brand on the first card, one of the 100+ brands AB Media Co has worked with, shows what AI answers respond to. We rebuilt its store, funnel and tracking, then its buying guides, each one written from the questions customers ask before they buy and each opening with the answer. Sales from AI assistants were read in GA4 by source every month. In 2024, AI assistants sent it zero sales. Today AI search is a seven-figure sales channel for it. The same audience now makes its ads work harder, with Performance Max returns at about 27x the industry norm. That came from years of AI SEO warming up an audience the ads could retarget, and not from day one. Beside it, TalentGum multiplied customer acquisition 40x in 18 months by running paid media, AI SEO and PR as a single plan.

Ecommerce brand AI SearchPaid MediaFunnel

Case study: Ecommerce brand

Ecommerce brand +48%

In 2024, AI assistants sent this ecommerce brand zero sales. Today, AI search is a seven-figure sales channel for it.

  • Online revenue in 2026 already 48% above all of 2024.
  • Ad returns about 27x the norm on PMax and 36x on Meta, from a warm audience years of AI SEO built.
Read the full case
TalentGum AI SearchPaid MediaDigital PR

Case study: TalentGum, Kids’ edtech

Kids’ edtech 40x

A kids’ learning brand went from one market on Meta to a global acquisition engine. Customer acquisition up 40x in 18 months.

  • Paid media, AI SEO and PR run as one system.
  • Domain authority tripled. India, the US, the Middle East.
  • From zero to 500+ first-page keywords.

WHAT LEADERSHIP TEAMS ASK US ABOUT LLMS.TXT, and where the evidence leaves each one.

REAL QUESTIONS, ANSWERED PROPERLY

What is llms.txt used for?

llms.txt is a Markdown file at the root of a website that gives AI tools a short summary and a list of links to the pages that matter. Its proposal says it is meant for agents reading on demand while they help someone with a task.

Jeremy Howard of Answer.AI proposed it in September 2024, and the format is light. Only one part is required, a heading with the site’s name. A summary, a few notes and sections of links with a line on each are optional, and a section called Optional holds links an agent can skip.

Today it is used mostly for developer documentation. According to llmstxt.org, OpenAI, Anthropic and Google’s Gemini team publish llms.txt files for their own docs, so a coding assistant can find the right page quickly.

For a business website its use is narrower than it is often sold. It does not change Google Search, and no major AI company has confirmed that its assistant reads other websites’ files when it answers a buyer.

So at AB Media Co we write one when it serves a clear purpose, such as public product or API documentation, and keep it short and accurate. Everything else goes into the pages and sources AI answers do read.

Does Google use llms.txt?

No. Google’s guide to generative AI search says Google Search ignores llms.txt files and that creating one will neither harm nor help your visibility or rankings, AI Overviews and AI Mode included. Google is not alone either: no major AI company has confirmed using the file.

The guide is precise on the point. It says you do not need machine-readable files, AI text files, markup or Markdown to appear in Google Search, because Google Search does not use them. It also notes that Google may crawl many kinds of files without treating them specially, so seeing Googlebot fetch the file proves nothing.

For a leadership team the meaning is practical. Any plan that promises better AI Overview visibility from llms.txt is promising something Google has ruled out.

What Google does use is where we spend your effort. Google only treats a page as a possible supporting link in AI Overviews or AI Mode once it is indexed and allowed to show a snippet, so our team checks indexing, snippet controls and robots rules on every revenue page. Then we rewrite openings to answer first and track the result in Search Console’s Generative AI performance report, page by page.

If you want the file for other tools, we add it. It never carries the Google plan.

Does ChatGPT use llms.txt?

There is no public confirmation that it does. OpenAI publishes an llms.txt file for its own developer documentation, according to llmstxt.org, but publishing a file is not the same as ChatGPT reading yours when it answers a buyer.

The best public evidence points the other way. In June 2025, Google’s John Mueller described server logs that showed consumer chatbots fetching pages for training and grounding, but not llms.txt. SE Ranking’s study of about 300,000 domains found no measurable link between having the file and being cited by AI tools.

What ChatGPT does rely on shows up in its answers. When it cites websites, the citations reveal whose word it took. In Ahrefs’ 15,000-prompt study, only around 12% of the pages that ChatGPT, Gemini and Copilot cited were also in Google’s top 10 for that prompt, which shows how much of the picture is assembled from other people’s websites.

So for ChatGPT, AB Media Co works on three things. OpenAI’s crawlers are allowed in robots.txt and at the CDN, your key pages read cleanly without JavaScript, and our digital PR team works on the sites its answers cite in your category. Those are the levers we can watch moving, in prompt runs every month and in GA4 sessions from chatgpt.com.

Is llms.txt worth it?

For most business websites it is worth an hour, not a line in the plan. It is housekeeping: harmless when accurate, useful to a few tools, and unproven as a route into AI answers.

It earns more than an hour in one case. If you publish product documentation, an API or a large help center that developers read through coding assistants, a clean llms.txt with links to Markdown versions of those pages can save the assistant time and cut wrong answers about your product.

For a business whose buyers ask AI tools which brand to choose, the file is unlikely to decide that answer. Google says its search ignores llms.txt, no major AI company has confirmed using it, and SE Ranking found no link between the file and citations across about 300,000 domains.

So our advice to leadership teams is about proportion. We will add one if your team wants it and check that it loads cleanly, but we do not bill it as a deliverable or count it as progress.

The return sits elsewhere: crawler access, pages that answer in their first lines, facts that agree everywhere, and mentions on the websites AI answers cite. That is where AB Media Co spends the hours, and where our guarantee sits: a 30 to 50% rise in AI search impressions within 90 days, measured in your own Search Console, with mentions across ChatGPT, Gemini and Perplexity tracked beside it.

Is llms.txt a standard?

Not in the formal sense. llms.txt is a proposal published on llmstxt.org, now in its second version, rather than a specification from a standards body such as the W3C or the IETF.

That matters for how much weight to give it. A proposal becomes useful when the systems you care about adopt it, and on the reading side that has not been shown. Adoption on the publishing side is real. llmstxt.org lists documentation platforms and plugins that generate the file, Chrome’s Lighthouse now checks for it in its agentic browsing audits, and SE Ranking found it on about 10% of the 300,000 domains it studied.

Publishing a file is not proof that anyone reads it, though. Google says its search ignores llms.txt, and Google’s John Mueller noted in June 2025 that the server logs he had seen showed chatbots fetching pages, not llms.txt.

So we handle it the way AB Media Co handles any emerging convention. We follow it, we test it where it might matter to you, and we keep it out of your plan until the evidence changes. If a major AI company confirms that its assistant reads llms.txt when answering buyers, we will tell you and adjust the plan. Until then, your money goes to work whose effect we can measure in your own accounts.

llms.txt vs robots.txt: what is the difference?

robots.txt controls access: it tells crawlers what they may fetch. llms.txt controls nothing: it offers a summary and links to tools that choose to read it. Of the two, robots.txt is the one that decides your AI visibility.

The difference has real consequences. If robots.txt or a firewall rule blocks Googlebot or OpenAI’s crawlers, no llms.txt file brings them back, because they never reach the pages it points to. And because llms.txt grants and withholds nothing, it cannot stop anyone using your content either.

In our audits the mix-up usually looks like this: a site blocks AI crawlers through a security or CDN setting nobody remembers switching on, then adds an llms.txt and wonders why nothing changes.

So our technical team always starts with access. We read robots.txt for each crawler by name, test the CDN and firewall rules in front of it, and check server logs to see which bots arrived and what response they received. Only once access is right do we discuss the optional file.

There is one business decision inside this. If you publish premium content, you may choose to block some AI crawlers on purpose. That is a commercial call for your leadership team, and we set robots.txt to match whatever you decide.

Does llms.txt help GEO?

There is no evidence that it does. GEO, generative engine optimization, is about being named when ChatGPT, Gemini or Perplexity recommends a brand, and those recommendations lean on what other websites say about you, which a file on your own server cannot supply.

The research on GEO points away from your own site. Across 75,000 brands, Ahrefs put the correlation between branded web mentions and AI Overview visibility at 0.664, and the one for backlinks at just 0.218. Google’s own guide notes that its AI features reflect what people say about products in blogs, videos and forum threads. Both describe evidence that lives on other people’s websites.

An llms.txt file describes you in your own words, to tools that may never open it. Even when one does, it gives a model a list of your pages, not a second opinion.

So the GEO side of our work at AB Media Co starts with a citation source map. We run your buyers’ questions through the assistants, log which websites they cite and mark the ones that have never mentioned you. Our digital PR team then works on those sites, with earned stories when there is news and sponsored features labeled as sponsored. We measure the result with monthly prompt runs and report how often you are named, as a range.

llms.txt vs schema: which matters more?

Schema, by a distance, though neither gets you named on its own. Schema describes the visible page to Google in a vocabulary it supports, while llms.txt is a summary Google Search ignores.

Google’s guide is clear that its AI features need no special schema, so markup is not a ticket into AI Overviews either. What structured data does well is remove doubt. When it matches the visible text, it tells Google what your organization is, what a product costs and which page covers what, and it keeps pages eligible for the rich results that still exist.

The two can also disagree, and that is where the real risk sits. A price in your schema, a different price on the page and an old one in an llms.txt file hand every system three versions of you.

So at AB Media Co we treat both as accuracy work, not visibility work. Our technical team checks that schema matches the page, removes markup that no longer does anything and, if you have an llms.txt, keeps it in step with the same facts. The visibility work then happens where answers actually move: openings that answer first, facts in plain HTML, and mentions on the sites AI tools cite.

Are AI search, AI SEO, AEO and GEO the same or different?

They are related parts of one job, and none of them depends on llms.txt. AI search is wherever your buyer asks, from Google’s AI Overviews and AI Mode to ChatGPT, Gemini and Perplexity.

AI SEO covers the full practice of being found in all of those places and in Google’s ordinary results. AEO, answer engine optimization, works on one page so that a passage from it is quoted as the answer. GEO, generative engine optimization, works on your reputation so that assistants name your brand when they recommend.

Each one depends on things a crawler can read and a buyer can trust. AEO needs indexable pages that answer in their first lines. GEO needs other websites describing you accurately. AI SEO needs the technical access that lets either happen. An llms.txt file sits to one side, as an optional summary that Google ignores and no major AI company has confirmed using.

At AB Media Co the four run as one plan, so you pay for the shared foundations once. The measurement is where they part: Search Console’s Generative AI report for Google’s AI features, our monthly prompt runs for the assistants, and your Share of Decision, how often the buyer, or the machine buying for them, picks you, over all of it.

How do you check an llms.txt file?

Open it in a browser, click every link and confirm the server returns it cleanly. A short file needs human eyes more than a tool, because the real risks are stale facts and links that should never have been public.

Here is what our team checks. The file loads at the root of your domain without a server error, since Chrome’s Lighthouse fails its llms.txt audit on a server error and only marks a missing file as not applicable. It opens with a heading carrying your site’s name, the one required element in the proposal. Every link resolves to a live page you are happy for anyone to read, with no staging URLs, admin paths or private documents swept in by a generator. And every fact in it matches your live pages, your profiles and your schema.

Then we look at whether anyone reads it. Server logs show requests to /llms.txt and the user agents behind them, so after a few months you can see whether the file has an audience at all.

At AB Media Co the file sits on the same update list as your prices and terms, so it is never the one place an old fact survives.

THE OTHER three.

ONE SYSTEM · ONE NUMBER

This is the decade the machines started choosing. Make sure they choose you.