llms.txt and Schema for AI: The Technical GEO Playbook

By Abhiijay Vinayak, co-founder of Beamcite. Last updated June 23, 2026.

TL;DR

llms.txt is a plain Markdown file you place at the root of your site (yourdomain.com/llms.txt) that gives AI tools a clean, curated map of your most important pages, proposed by Answer.AI co-founder Jeremy Howard in September 2024. Schema markup is structured data (usually JSON-LD) that labels what your content actually is, so machines read an entity instead of guessing at text. Here is the honest split the breathless guides skip: schema is already used by Google's and Microsoft's AI features to understand and verify content, while Google has said it does not use llms.txt and is not planning to. So the technical GEO setup that earns AI citations is schema done right, plus an llms.txt that helps in the narrow cases where it helps, plus the visible, well-sourced content that engines actually quote. Beamcite is a done-for-you GEO service that handles this technical setup and the content together for SaaS and Mac-app companies, and the rest of this playbook is exactly what that setup involves.

Most "technical GEO" advice online is one of two extremes: add an llms.txt file and watch the AI citations roll in, or schema markup is dead, just write good content. Both are wrong. The technical layer is real and worth getting right, but only if you know what each piece does and what the evidence says about it. This playbook covers what llms.txt is, the actual state of its adoption, how to write one, what schema markup does for AI search, and where the technical setup ends and the content work begins. Beamcite handles this layer for the companies it works with, and the method below is the same one it uses.

Quick navigation:

• Want the non-technical version of getting cited by AI? Read the full GEO method
• Just need to ship an llms.txt file? Jump to "How to write an llms.txt file" below.
• Comparing the whole technical setup? You are in the right place. Keep reading.

What is llms.txt?

llms.txt is a proposed standard: a single Markdown file at your site's root that gives large language models a structured, low-noise index of your site. Instead of forcing a model to crawl and parse hundreds of HTML pages to work out what your business does, the file hands it a curated summary, who you are, what the site covers, and which pages carry the most signal, in clean Markdown a model reads easily.

The convention was proposed by Jeremy Howard, co-founder of Answer.AI and fast.ai, on September 3, 2024, and the specification lives at llmstxt.org. The proposal actually defines two files. The first, /llms.txt, is the navigation aid: an H1 title, a short description, and lists of key pages with one-line descriptions and links. The second, /llms-full.txt, is an optional companion that holds the full text of your important pages in one Markdown document, so a model can ingest everything in a single request.

The idea borrows from robots.txt and sitemap.xml: a predictable file at a predictable location that machines can find without guessing. The difference is what it is for. robots.txt tells crawlers what they may access; llms.txt tries to tell language models what matters and hand it to them in a format that wastes no tokens on navigation, ads, and markup.

The honest state of llms.txt adoption

This is where most guides stop being useful, so this section is blunt. As of mid-2026, no major AI or search engine has committed to using llms.txt for ranking or content discovery in its production systems.

Google has been the most direct. At Search Central Live in July 2025, Gary Illyes said Google does not support llms.txt and is not planning to, and John Mueller compared the file to the old meta keywords tag, the tag search engines stopped trusting decades ago because site owners controlled it and gamed it. Mueller's core point was practical: AI crawlers already download the full page, so serving them a separate, self-declared version of your content is both redundant and easy to abuse. A site could show one thing in llms.txt and another to readers, which is exactly the cloaking problem that killed meta keywords.

It is worth being precise about what that does and does not mean. Several large companies do publish an llms.txt, including Anthropic, Stripe, Cursor, Mintlify, and Zapier. But in nearly every case the file sits on developer documentation, and it serves coding assistants: it gives an AI tool a clean entry point into API docs so it can answer technical questions without scraping the whole doc site. That is a genuine use. It is not the same as "add llms.txt and AI will recommend your business," and anyone selling the second claim is ahead of the evidence.

So the file is cheap to add, harmless, and occasionally useful, and it is not a citation lever for most businesses today. Treat it as low-cost hygiene with a real but narrow payoff, not as the thing that gets you named in an answer.

When llms.txt is actually worth adding

The file makes the most sense for a specific kind of site. It is worth adding if any of these describe you:

• You run documentation-heavy or developer-facing content. API docs, SDK references, and technical guides are exactly what coding assistants parse, and a clean llms.txt or llms-full.txt genuinely helps those tools answer questions about your product accurately.
• Your pages are heavy with navigation, scripts, and interactive elements that bury the actual content. A Markdown summary gives a model a cleaner read than your rendered HTML.
• You already maintain content in Markdown and can generate the file automatically, so it costs you almost nothing to keep current.

It is not worth obsessing over if you run a standard marketing site or a local business site, where the payoff today is speculative. If you add it, automate it. WordPress users can generate the file with plugins (Yoast SEO and Rank Math both added native llms.txt generation in 2025), and several standalone generators will build one from your sitemap. A stale, hand-maintained llms.txt that drifts out of sync with your real pages is worse than none, because the whole point is an accurate map.

How to write an llms.txt file

The format is deliberately simple. It is Markdown, it lives at yourdomain.com/llms.txt with no subdirectory, and it follows the structure the spec lays out. A minimal valid file looks like this:

# Acme Analytics

> Acme Analytics is a privacy-first product analytics tool for SaaS teams. This file maps the most useful pages for AI tools.

## Core pages

- [Product overview](https://acme.example/product): what Acme does and who it is for
- [Pricing](https://acme.example/pricing): plans, limits, and what is included
- [Documentation](https://acme.example/docs): setup, SDKs, and the API reference

## Guides

- [Getting started](https://acme.example/docs/start): install and send your first event
- [Self-hosting](https://acme.example/docs/self-host): run Acme on your own infrastructure

## Optional

- [Changelog](https://acme.example/changelog): recent releases

The rules that matter: one H1 with the site or product name, an optional blockquote description right under it, then H2 sections grouping your links, each link followed by a short plain-language note on what the page covers. Keep an "Optional" section for pages a model can skip when it is short on context. If you publish llms-full.txt as well, it holds the actual page content in one file rather than links. Validate that every URL resolves and that the file is served as plain text, then keep it in sync with your real site whenever pages change.

Schema markup for AI: what actually works

Schema markup is the part of the technical setup with real, current evidence behind it, and it is where most of your attention should go. Schema (the vocabulary lives at schema.org) is structured data you embed in a page, normally as a JSON-LD block, that labels your content explicitly: this is an Organization, this is its name and logo, this is a BlogPosting by this author on this date, these are the questions and answers in the FAQ.

Unlike llms.txt, structured data is actively used by AI search features. Google's own documentation explains that structured data helps it understand the content of a page, and at Search Central Live, Google engineers described machine-readable markup as fundamental to AI-powered search because language models need precise, structured signals to verify entities and avoid hallucinating. Industry reporting on the same event noted that Google and Microsoft both confirmed using schema markup in their generative AI features, where it acts as context for answer generation and entity verification.

One honest caveat, because it gets oversold the other direction too: there is no special "AI" schema type, and no markup is a shortcut that forces an engine to cite you. Schema does not write your answer; it helps the machine correctly read the answer you wrote. It is infrastructure, not magic. The visible content is still what gets quoted. Schema makes sure the engine attributes that content to the right entity, with the right author, the right dates, and the right relationships.

The schema types that matter for AI search

You do not need every type schema.org defines. A focused set covers most businesses:

Schema type	What it labels	Why it matters for AI
Organization	Your business as an entity: name, logo, URL, social profiles	Lets engines build a confident entity for your brand and attribute mentions to it
BlogPosting / Article	A post: headline, author, publish and modified dates	Ties content to a named author and a date, both trust signals AI features weigh
FAQPage	Question-and-answer pairs on the page	Maps directly to how assistants extract short, quotable answers
Product	A product: name, description, key attributes	Feeds precise product facts into comparison and recommendation answers
BreadcrumbList	The page's position in your site hierarchy	Helps engines understand structure and the relationship between pages

Implement these as JSON-LD in the page head, keep every value consistent with what the page actually shows, and validate the markup before you ship it. Schema that claims something the visible page does not say is the structured-data version of cloaking, and it erodes the trust the markup is supposed to build.

llms.txt vs schema: which does what

The two are not competitors, they do different jobs, and the evidence behind them is very different. This is the comparison to keep in mind before you spend time on either.

	llms.txt	Schema markup
What it is	A Markdown index of your site at /llms.txt	Structured data (JSON-LD) embedded in each page
Used by major AI engines today	Not confirmed; Google says it does not use it	Yes; Google and Microsoft confirm using it in AI features
Main real-world payoff	Cleaner input for coding assistants and doc tools	Entity understanding and verification across AI search
Effort to maintain	Low if automated, risky if hand-kept	Moderate; per-page, must match visible content
Verdict	Optional hygiene, narrow payoff	Worth doing well for most sites

The short read: if you only have time for one, do schema properly. Add llms.txt if it is cheap to automate and your content is documentation-heavy. Neither one, on its own, gets you cited.

What actually earns AI citations

The technical setup is necessary but not sufficient, and it is important to be clear about that so you do not spend a month on markup and wonder why nothing moved. The single largest study on this question, a Princeton-led paper presented at the KDD 2024 conference, tested optimization strategies across 10,000 real queries in 25 domains and found that what lifts a page's visibility in AI answers is content quality: adding authoritative citations raised visibility by about 30%, statistics by about 32%, and expert quotations by about 41%, for up to a 40% overall gain. Keyword stuffing did nothing, and the study did not find a markup trick that substituted for substance.

So the technical layer earns its place by making sure engines can read and attribute your content correctly, and the citations come from the content itself: an answer-shaped opening, original data, self-contained paragraphs, comparison tables, and a real FAQ, published under a named human author and corroborated by genuine presence on the platforms your buyers use. That is the full method, and the technical pieces here are one layer of it. The complete playbook is in the guide to getting your business cited and recommended by AI.

Doing the technical setup yourself vs having it done

The technical pieces are not hard in isolation. The work is in doing all of it, correctly, on every post, and keeping it in sync as the site grows. Here is how the realistic options compare.

Approach	Who maintains the schema and files	Consistency across the whole site	Best for
Do it yourself	You, by hand or with plugins	Depends on discipline; drift is common	Teams with technical capacity and spare time
Raw AI tool	Nobody; markup is generated and shipped unread	Often inconsistent or invalid	Teams who will validate every output themselves
Beamcite	A human-reviewed pipeline, every post	Built in, checked before publish	SaaS and Mac-app teams with no time to run it

The honest trade-off: a one-time schema setup is a weekend project, but GEO is not a one-time job. Every new post needs correct markup, an accurate author entity, and current dates, and a site that adds content weekly will drift out of sync fast without a process. Beamcite is built for that recurring case: AI drafts each post for speed, a human editor reviews and fact-checks it, and the technical setup (schema, internal links, and llms.txt where it helps) ships correct with every post rather than as a one-off task someone forgets to repeat. Content is AI-drafted and human-edited, and Beamcite says so plainly, because the human-review layer is the point.

Which technical GEO setup does your site need?

• Standard marketing or local-business site: Get schema right (Organization, Article, FAQPage), skip the time-sink on llms.txt, and put your effort into citable content.
• Documentation-heavy or developer-facing product: Do schema, and add an automated llms.txt or llms-full.txt; coding assistants genuinely use it.
• Site that publishes content weekly: The bottleneck is consistency, not setup. Use a process (or a service) that ships correct markup on every post, not a one-time configuration.
• No time or team to run any of it: A done-for-you GEO service like Beamcite handles the technical layer and the content together, reviewed by a human before each post ships.

Conclusion

The technical GEO playbook comes down to one clear priority order. Schema markup is real, current, and confirmed in use by Google's and Microsoft's AI features, so do it well: label your organization, your posts, your authors, and your FAQs, and keep the markup honest to the visible page. llms.txt is cheap, harmless, and useful mainly for documentation and coding assistants, so add it if you can automate it and skip the hype that says it gets you cited. Above both sits the content itself, where the Princeton evidence says citations are actually won, through original data, quotable structure, and a named human author.

Different sites need different amounts of this. A local business should fix its schema and move on. A developer tool should do schema and llms.txt. A company publishing weekly needs a process that ships correct markup every time, which is exactly the recurring work that gets dropped when one person is doing everything. SaaS and Mac-app teams that want the technical layer and the content handled together, reviewed by a human before each post ships, are who Beamcite is built for. Start with a free strategy call to map out what your site actually needs.

Frequently asked questions

What is llms.txt in simple terms? llms.txt is a plain Markdown file you put at your site's root (yourdomain.com/llms.txt) that gives AI tools a curated list of your most important pages with short descriptions. It was proposed by Answer.AI's Jeremy Howard in September 2024 as a way to hand language models a clean map of your site instead of making them parse all your HTML.

Does llms.txt help you rank or get cited by AI? Not on current evidence. Google has said it does not use llms.txt and is not planning to, and no major AI engine has confirmed using it for ranking or content discovery in production. Its real payoff today is giving coding assistants and documentation tools a clean entry point into developer docs, which is why companies like Anthropic and Stripe publish one for their docs.

Is schema markup worth it for AI search? Yes. Unlike llms.txt, structured data is actively used: Google and Microsoft have confirmed their AI features use schema markup to understand content and verify entities. There is no special "AI" schema type and no markup shortcut to a citation, but correct Organization, Article, FAQPage, and Product markup helps engines read and attribute your content accurately.

What schema types should a business add first? Start with Organization (your brand as an entity), Article or BlogPosting (with a named author and dates) on every post, and FAQPage where you have real question-and-answer content. Add Product markup if you sell products and BreadcrumbList for site structure. Implement them as JSON-LD and make sure every value matches what the page visibly shows.

Do I need both llms.txt and schema? Most sites should prioritize schema, because it is the one with current evidence of use in AI search. Add llms.txt if your content is documentation-heavy or you can generate it automatically at near-zero cost. Neither file gets you cited on its own; they help engines read and attribute the content that does.

Can a done-for-you service handle the technical GEO setup? Yes. Beamcite sets up and maintains schema markup, internal linking, and llms.txt where it helps, and pairs that technical layer with human-reviewed GEO content on each post, so the markup stays correct and in sync as the site grows rather than being a one-time configuration that drifts.

---

Written by Abhiijay Vinayak, co-founder of Beamcite, the done-for-you GEO service that gets SaaS and Mac-app businesses cited and recommended by AI. Start with a free strategy call.