Do I need an llms.txt file?

If your website has more than a handful of pages and you care how AI assistants read it, yes — an llms.txt file costs minutes and gives language models a curated map of your canonical content. If your site is small or your content is blocked to AI crawlers, fix crawlability first. No major AI provider has publicly guaranteed it consumes llms.txt, so treat the file as inexpensive hygiene rather than a ranking lever.

What is the difference between llms.txt and llms-full.txt?

llms.txt is a curated index: a Markdown file of links with one-line descriptions, served at the site root. llms-full.txt is a companion file that inlines the full text of key pages into a single document, so an AI system can load the content in one request. Most sites should start with llms.txt and add llms-full.txt only if their documentation is a primary product surface.

Does llms.txt improve Google rankings?

No. llms.txt targets large language models and AI agents, not Google's ranking systems, and Google has not adopted it as a ranking input. Its value is in guiding AI assistants and agentic browsers to your canonical pages — a different channel from classic search rankings.

Where does the llms.txt file go?

The llms.txt file lives at the root of your domain — for example, https://example.com/llms.txt — served as plain UTF-8 Markdown. It requires one H1 with the site name, and typically includes a blockquote summary followed by H2 sections of annotated links. No registration, no tooling, no build step.

What Is an llms.txt File (and Do You Need One)?

llms.txt is a plain-text file at the root of your domain that hands AI systems a curated map of what matters on your site. The file takes ten minutes to write. Deciding what belongs in it is the actual work.

What is an llms.txt file?

Updated July 2026.

An llms.txt file is a Markdown document served at the root of a website (example.com/llms.txt) that gives large language models a curated, machine-readable summary of the site's most important content. It was proposed by Jeremy Howard of Answer.AI in September 2024. Unlike robots.txt, which controls crawler access, llms.txt guides comprehension: it tells an AI system what the site is, which pages are canonical, and what each one covers.

That is the whole idea. One file, one purpose. Everything else is curation.

The llms.txt spec, in brief

The llms.txt spec is deliberately small. One H1 with the site or project name — the only required element. An optional blockquote summary directly beneath it. Then H2-delimited sections containing link lists in the form `- Page name: one-line description`. A final section titled "Optional" holds links a model may skip when context is short.

There is a companion convention, llms-full.txt, which inlines the full text of your key pages into a single file for systems that want everything in one request.

Markdown, because language models parse it natively. No XML, no schema registry, no tooling. If your file needs a build step, you have misread the spec.

A copyable llms.txt example

Here is a spec-accurate llms.txt example. Swap in your own pages.

``` # Blankpage > Strategic technology practice in Albuquerque, New Mexico. AI models, hyperdata systems, e-intelligence. ## Services - Generative Engine Optimization: Engineering brand citations inside AI answers. - AI Reputation Management: Auditing and correcting what AI systems say about a brand. ## Insights - What Is GEO?: Definition and mechanics of generative engine optimization. - What Is an llms.txt File?: The llms.txt spec, an example, and when it matters. ## Optional - Contact: Talk to an operator. ```

Notice what the file forces: every line is a claim about priority, stated as one factual sentence. That is the part no generator can do for you.

Do you need an llms.txt file? A terse decision

Yes, if any of these hold: your product has documentation an assistant might read; your site has more pages than a stranger could triage in a minute; you are investing in generative engine optimization and want your retrieval plumbing clean; you care about AI agents completing tasks on your site.

Low priority, if: your site is five pages of brochure, or your fundamentals are broken — text rendered only by JavaScript, no heading hierarchy, AI crawlers blocked. Fix those first. An llms.txt file on an uncrawlable site is a map to a locked building.

One honest caveat. As of mid-2026, no major AI provider has publicly committed to consuming llms.txt in production, and Google's ranking systems do not use it. Treat it as low-cost hygiene with asymmetric upside, not a ranking lever. Ten minutes of work. Do it. Then do the real work.

Why one-click llms.txt generators miss the point

A generator dumps your sitemap into Markdown and calls it done. That is inventory, not curation. An uncurated llms.txt tells a model nothing about priority, which is the only signal the file exists to carry.

Curation is editorial work: choosing the twenty URLs out of four hundred that deserve model attention, writing a one-line factual summary that survives being quoted, deciding what is genuinely Optional. It requires knowing which questions buyers ask and which of your pages answers each one.

The file is trivial. The curation strategy is not.

llms.txt vs robots.txt vs sitemap.xml

| File | Audience | Question it answers | | --- | --- | --- | | robots.txt | Crawlers | May you fetch this? | | sitemap.xml | Search indexers | What exists here? | | llms.txt | Language models and agents | What matters here, and what does it mean? |

The three files stack; none replaces another. robots.txt is permission, sitemap.xml is discovery, llms.txt is comprehension. A complete retrieval setup carries all three, plus server-rendered content and clean heading hierarchy.

Where llms.txt fits in a GEO program

llms.txt is one item of retrieval plumbing among several: explicit crawler allowances for GPTBot, ClaudeBot, and PerplexityBot; server-rendered text; extractable answer-first passages; structured data; third-party corroboration. The file helps models find your best pages. It does not make those pages worth citing — that is what generative engine optimization is for.

In a two-week Diagnostic Sprint we write the llms.txt file in the first week, because it is fast and it is hygiene. The remaining time goes to the parts that actually move citations: passage structure, corroboration, and answer engine optimization across ChatGPT, Perplexity, and Google AI Overviews.

We will not sell you a file you could write before lunch. We will tell you what to put in it, and why.

Frequently asked questions

Do I need an llms.txt file?: If your website has more than a handful of pages and you care how AI assistants read it, yes — an llms.txt file costs minutes and gives language models a curated map of your canonical content. If your site is small or your content is blocked to AI crawlers, fix crawlability first. No major AI provider has publicly guaranteed it consumes llms.txt, so treat the file as inexpensive hygiene rather than a ranking lever.
What is the difference between llms.txt and llms-full.txt?: llms.txt is a curated index: a Markdown file of links with one-line descriptions, served at the site root. llms-full.txt is a companion file that inlines the full text of key pages into a single document, so an AI system can load the content in one request. Most sites should start with llms.txt and add llms-full.txt only if their documentation is a primary product surface.
Does llms.txt improve Google rankings?: No. llms.txt targets large language models and AI agents, not Google's ranking systems, and Google has not adopted it as a ranking input. Its value is in guiding AI assistants and agentic browsers to your canonical pages — a different channel from classic search rankings.
Where does the llms.txt file go?: The llms.txt file lives at the root of your domain — for example, https://example.com/llms.txt — served as plain UTF-8 Markdown. It requires one H1 with the site name, and typically includes a blockquote summary followed by H2 sections of annotated links. No registration, no tooling, no build step.