What Is llms.txt and Do I Need It?

If you have spent any time researching AI visibility in 2026, you have probably come across llms.txt. It is one of those files that sounds technical and easy to ignore, but it has become one of the simplest high-impact moves you can make for AI visibility. This guide explains what llms.txt actually is, who genuinely needs it, who can wait, and how to create one that works.

If you want the deeper background on what AI visibility means and why files like this matter, our explainer on what AI visibility is and why it matters sets the context. For the in-depth technical guide to llms.txt specifically, see LLMs.txt: the new file every website needs in 2026.

What llms.txt Actually Is

llms.txt is a plain-text file you host at the root of your website (yoursite.com/llms.txt) that gives AI engines a structured, human-readable summary of your business in markdown format. Think of it as the AI equivalent of robots.txt, except instead of telling crawlers what they can and cannot access, it tells them in your own words what your business is, what you offer, and where to find your most important content.

The format is intentionally simple. A typical llms.txt has a top-level title, a brief description, and then sections grouping links to your most important pages with short descriptions of what each link contains. It is markdown, so it renders cleanly both as plain text and when read by an AI engine that is trying to understand your business.

The file emerged from a proposal originally aimed at giving large language models a structured entry point for understanding any website. The llmstxt.org spec remains the canonical reference for the format, and the spec is short enough to read in a few minutes.

How AI Engines Actually Use llms.txt

The honest answer is that adoption varies across AI engines and is still evolving. ChatGPT, Perplexity, Gemini, and Grok all retrieve and use llms.txt when they find it, but the weight they place on it differs. Perplexity tends to use llms.txt content directly in source citations. ChatGPT uses it as one input among many. Gemini and Grok use it more sparingly but still parse it when present.

What llms.txt does consistently is give AI engines a high-quality summary of your business that they can rely on when they want to introduce you or pick the right page to cite. Without llms.txt, the AI has to infer your business identity from your homepage, your title tags, and whatever schema markup you have implemented. With a good llms.txt, the AI has a clean, structured description in your own words.

In our analysis of how AI search engines decide who to recommend, llms.txt comes up as a useful trust and structure signal, particularly for businesses whose homepages are heavy on marketing copy and light on actual information.

Who Genuinely Needs llms.txt Right Now

llms.txt has the biggest impact for businesses that:

• Have a marketing-heavy homepage that does not clearly explain what the business does in the first paragraph. If a customer would have to scroll or click to learn what you offer, your homepage is hurting your AI visibility. llms.txt provides the plain-language summary that the homepage is missing.
• Operate in a niche or specialty area where AI engines might miscategorise the business. If you sell handmade ceramics, run a recruitment agency for tech roles, or offer mobile pet grooming, AI engines often default to generic categories without explicit guidance. llms.txt corrects this.
• Have multiple distinct service lines or product categories that need their own context. The llms.txt format with grouped sections makes it easy to describe a multi-service business in a way the AI can navigate.
• Publish technical or specialised content where the structure of your site matters. Documentation sites, SaaS products with API references, and content libraries all benefit because llms.txt acts as a curated index of what is most important.
• Are competing for AI visibility in a crowded niche where every small signal matters.

One client we worked with, an e-commerce store selling artisan kitchenware, saw measurable improvements in AI engine descriptions after adding llms.txt. Before, ChatGPT described them generically as a UK kitchenware retailer. After llms.txt, ChatGPT described them accurately as a UK-based artisan kitchenware brand focused on hand-forged knives and traditional cookware made by independent makers. The improvement came from the AI having a clean, branded description to draw from rather than having to construct one from product listings.

Who Can Probably Wait

llms.txt is not equally valuable for every business. You can deprioritise it if:

• Your homepage is already clear and informative. If a customer landing on your homepage understands what you do, where you operate, and what you offer within five seconds of reading, llms.txt adds less because the AI gets the same information from your homepage directly.
• Your AI visibility score is already above 70. At high scores, the marginal gain from llms.txt is small. Better to focus remaining effort on the larger weighted factors like content depth and review signals.
• You have not yet implemented basic schema markup. Schema and llms.txt are complementary but schema should come first because it carries more weight. Adding llms.txt without schema is like writing a great description for a building that has no foundation.
• Your business is hyper-local with a single service. If you are a solo plumber serving one town, your homepage probably already says exactly what AI engines need to know. llms.txt is still nice to have, but it is rarely the highest-leverage move.

The honest priority order for most businesses is: complete schema first, AI crawler access second, trust signals third, content structure fourth, and llms.txt fifth. It is genuinely useful, but not at the top of the list for most situations.

How to Create a Good llms.txt

The structure follows the official spec from llmstxt.org. At minimum, your file should have:

1. A top-level title (H1) with your business name
2. A blockquote with a one-line description of what you do
3. A short paragraph or two giving more context about your business
4. Sections (H2) grouping your most important pages by category, with each link followed by a brief description of what that page contains

A simplified example of what a service business llms.txt might look like in markdown form: the title would be a level-one heading with the business name, followed by a blockquote summary like "Family law firm based in Maidstone, serving Kent and the South East since 2003." Then a short paragraph adding context. Then headed sections like "Services" listing the three or four core service pages with one-line descriptions, and an "About" section linking to the team and contact pages. Plain factual descriptions of each link beat marketing copy every time.

The format is approachable and self-documenting. The key qualities of a good llms.txt are: clear single description, accurate categorisation, well-chosen links to your highest-value content, and natural language descriptions of each link.

Common Mistakes to Avoid

From our audits of llms.txt files in the wild, the common mistakes are:

• Treating it like a sitemap by listing every URL on the site. llms.txt is curated, not exhaustive. Pick your ten to twenty most important pages, not every page you have ever published.
• Marketing-style descriptions that read like ad copy. "Award-winning, industry-leading solutions for forward-thinking businesses" tells the AI nothing useful. Plain factual descriptions work better.
• Outdated information because llms.txt is text and easy to forget. If your services or pricing change, update llms.txt at the same time. Stale llms.txt can hurt more than no llms.txt.
• Missing the description blockquote at the top. The blockquote is one of the most important parts because some AI engines pull from it directly when introducing your business.
• Wrong file location. The file must be at yoursite.com/llms.txt, not /docs/llms.txt or /assets/llms.txt. Hosting it anywhere else means AI engines will not find it.
• Forgetting to allow crawler access. If your robots.txt explicitly blocks certain user agents, make sure those agents can access llms.txt. Some servers also need a content-type header set to text/plain or text/markdown.

How llms.txt Fits With Other AI Visibility Work

llms.txt sits in a specific place in the AI visibility stack. It works best when paired with:

• Complete schema markup so the structured data on your pages reinforces what llms.txt says. The two should agree about what your business does, where you operate, and what you offer. For the decision tree on which schemas you need, see what types of schema does my business need.
• Open AI crawler access in robots.txt so the AI engines that read llms.txt can also retrieve and parse your linked pages.
• Real third-party trust signals on Trustpilot, Google Business Profile, Reddit, and industry platforms. llms.txt describes your business but trust signals verify the claims.
• Active content publishing because llms.txt can point to your most current content. A blog or news section with regular updates pairs well.

The combination of all four is what we work through in how to increase your AI visibility score. llms.txt alone moves the needle a few points. llms.txt combined with the rest moves it by tens of points.

Real Implementation Examples From Client Work

A recruitment agency we work with implemented llms.txt as part of a broader AI visibility programme. Before any of the work, their AI visibility score was 18. We added complete schema, opened crawler access, and added llms.txt with a clean business description and links to their sector pages. Within three months their score was 62 and they began appearing in answers from all four major AI engines. The llms.txt contribution to that gain was small in isolation (probably 2 to 3 points) but it amplified the work done elsewhere because AI engines had a clean source to cite when describing the agency.

Another client, a Shopify jewellery store, had no schema and no llms.txt when we started. We added eight schema types and an llms.txt that grouped their product categories (rings, bracelets, necklaces, custom commissions) with descriptions of each. Their score rose from 31 to 84, and ChatGPT began describing them accurately for queries about sustainable jewellery brands. Without llms.txt, ChatGPT had been describing them as a UK jewellery seller. After llms.txt, it described them as an independent UK jewellery brand specialising in handmade sustainable pieces with custom commission work available. The accuracy gain came from giving the AI a curated description to use instead of one generated from product listings. More before-and-after detail is on the FlinnSchema results page.

FlinnSchema itself publishes an llms.txt file. You can visit flinnschema.com/llms.txt to see ours. It uses the same structure recommended above with a clear description, our key sections (services, blog, audit tool), and brief explanations of each link.

The Practical Action Plan

If you have decided llms.txt is worth doing, here is the practical sequence:

1. Write the file in plain markdown using the structure above. Aim for one to two pages of well-organised content rather than a sprawling document.
2. Host it at yoursite.com/llms.txt. On WordPress, you can add it via a plugin or by uploading the file to your server root. On Shopify, you may need a developer's help because the platform does not allow root-level files natively (though there are workarounds via app proxies).
3. Verify it loads by visiting the URL directly in a browser. The file should render as plain text or markdown without authentication or redirects.
4. Reference it from robots.txt as an allow rule, particularly if you have strict crawler rules elsewhere.
5. Update it quarterly or whenever your services, pricing, or major content changes. Staleness defeats the purpose.

For a deeper technical walkthrough including the exact markdown structure and validation steps, see our LLMs.txt: the new file every website needs in 2026 post. For the broader AI visibility roadmap that llms.txt fits into, see how to increase your AI visibility score.

The Short Answer

Yes, you probably need llms.txt, but it is not the most urgent thing on your AI visibility list. Add it as one part of a complete AI visibility programme alongside schema markup, crawler access, trust signals, and active content. Add it after you have the foundations in place. Skip it entirely if your homepage already clearly explains your business and your AI visibility score is above 70.

If you are not sure where you stand, our free audit includes an llms.txt check as one of the 26 factors. You will see exactly whether you need it and how much it would affect your score. For implementation support, our Premium plan covers the work, or you can book a free 15-minute walkthrough to talk through your specific situation.

The honest summary: llms.txt is a useful low-effort addition for most businesses doing AI visibility work, but it is not magic on its own. Pair it with the rest of the stack and it earns its place. Treat it as a standalone fix and you will be disappointed. For the deeper context on customer behaviour driving this whole shift, see our piece on whether customers actually use ChatGPT to find businesses.