llms.txt - The File That Separates 63 from 34

What this article answers

• What is an llms.txt file and why does my website need one?
• How do I create an llms.txt file for my site?
• Does llms.txt actually improve AI visibility scores?

Key takeaways

• Create a plain-text llms.txt file at your domain root with your business description, key content areas, and URLs.
• Keep it factual and structured - AI wants specific product info, not marketing fluff.
• Update your llms.txt whenever you launch new features or content areas so AI never gives outdated info.
• Include direct URLs to important pages - a description without links is a map without roads.

What Is llms.txt and How Does It Work?

llms.txt is a plain-text file sitting at your domain root (e.g., example.com/llms.txt) that gives large language models a structured summary of your website.

Think of it as robots.txt’s smarter cousin. robots.txt tells crawlers which pages to access. llms.txt tells AI systems what your site actually is, what it offers, and how to navigate it.

Tidio’s llms.txt runs 251 lines. It lists every product, every feature, every integration. When ChatGPT needs to answer “What live chat tools support Shopify?” - Tidio’s file hands over the answer on a silver platter. Crisp? No llms.txt at all. Score: 34.

The format is dead simple:

• A brief business description
• Key content areas with URLs
• Product/service categories
• Contact info and location
• Links to important resources (FAQ, docs, API references)

Why Do AI Assistants Struggle Without llms.txt?

Without llms.txt, AI engines stitch business context from scattered HTML and ads, often skipping or misdescribing your site entirely during the citation decision.

Here’s what happens without llms.txt. Someone asks ChatGPT about your industry. The AI has to piece together your site’s purpose from scattered HTML pages, marketing fluff, and cookie banners. It gets confused. It skips you.

Put on ChatGPT’s glasses for a second. It’s scanning hundreds of sites for an answer. One site has a clear, structured summary of everything it does. Another site forces the AI to parse fifty pages of JavaScript-heavy marketing copy. Which one gets cited?

We’ve audited sites across multiple verticals. The pattern is unmistakable:

• Sites with llms.txt get their business described accurately by AI - because they wrote the description themselves
• Sites without it get described wrong, or worse, not described at all
• The file works from day one - unlike SEO authority that takes months to build
• You control the narrative. You define how AI systems talk about your business.

How Do You Build an llms.txt File?

Create llms.txt at your root with a one-line description, About section, Products and Services list with URLs, and key content links to FAQ, docs, and resources.

Create a file called llms.txt in your site’s root directory. Here’s the skeleton:

# Example Business Name

> Brief one-line description of what the business does.

## About
Two to three sentences about the business, its expertise, and what makes it unique.

## Products & Services
- Category 1: Description (URL: /category-1)
- Category 2: Description (URL: /category-2)

## Key Content
- Blog: Expert articles about [topic] (/blog)
- FAQ: Common questions answered (/faq)
- Guides: In-depth resources (/guides)

## Contact
- Location: City, State
- Email: [email protected]
- Website: https://example.com

Want to go deeper? Create llms-full.txt with your complete content inventory, taxonomy, and technical specs. Our own llms-full.txt runs the full content map - every article, every audit, every tool.

For Shopify: upload via the theme editor (Settings > Files) or host on your CDN. For WordPress: drop it in the root directory or use a plugin. For Next.js or static sites: add it to your public folder.

Start here: Open a text editor right now and write 10 lines describing your business. That’s your llms.txt draft.

A high-scoring llms.txt has four required sections that each give AI engines a different navigation signal.

What Are the Most Common llms.txt Mistakes?

Five common mistakes hurt llms.txt: writing novels instead of summaries, skipping updates, using marketing copy, forgetting URLs, and placing the file in a subdirectory.

We see the same mistakes in audit after audit:

Writing a novel instead of a summary. AI systems want concise, structured facts - not your brand manifesto. Tidio’s 251 lines work because every line is specific product information, not fluff.

Never updating it. Your llms.txt from six months ago doesn’t mention your three new features. AI is now giving outdated info about your business.

Writing marketing copy. “We’re the world’s leading provider of synergistic solutions…” - AI systems want facts. What do you do. Where are you. What do you sell. Period.

Forgetting URLs. The file should help AI navigate to specific pages. A description without links is a map without roads.

Putting it in a subdirectory. It must live at /llms.txt. Not /assets/llms.txt. Not /docs/llms.txt. The root.

Score Impact in Practice

llms.txt carries 1% raw weight in AI Discovery, but scoring is binary (7-8 with file, 0 without), and the audit shows a 29-point gap between top and bottom sites.

The llms.txt criterion carries 1% raw weight in the AI Discovery pillar of the AEO scoring model. That might sound small, but the downstream effects are outsized because llms.txt shapes how AI understands every other page on your site.

Sites with a well-structured llms.txt consistently score 7-8/10 on this criterion. Sites without one score 0 - there’s no partial credit. You either have the file or you don’t. In the live chat vertical, Tidio’s 251-line llms.txt contributes to its 63/100 overall score. Crisp, with no llms.txt at all, sits at 34/100. That 29-point gap isn’t caused by llms.txt alone, but when we break down the scoring, the file’s presence correlates with higher scores across multiple criteria because it helps AI contextualize everything else on the site.

Across the 500+ sites we’ve audited, fewer than 15% have an llms.txt file. The adoption rate among Y Combinator startups is higher - roughly 25% - but still leaves massive ground unclaimed. Early adopters get disproportionate benefit because AI systems are actively looking for this file and rewarding the sites that provide it.

How AI Engines Evaluate This

Each AI engine processes llms.txt slightly differently, but the core behavior is the same: they check your domain root for the file and use it as a primary context source.

ChatGPT (GPTBot) fetches /llms.txt early in its crawl cycle. When it finds a structured summary with clear product descriptions and URLs, it uses those descriptions almost verbatim when answering questions about your business. Without the file, GPTBot has to infer your business purpose from scattered page content - and it frequently gets it wrong or skips you entirely.

Claude (Anthropic’s crawler) treats llms.txt as a trust signal. A site that takes the time to create a machine-readable summary demonstrates intentionality about AI engagement. Claude’s retrieval system uses the file to build an entity model - what the business does, what it offers, where its key content lives. This entity model feeds directly into citation decisions.

Perplexity is the most aggressive consumer of llms.txt. When Perplexity builds its source list for an answer, it checks llms.txt to quickly determine whether a site is relevant to the query. A site with a clear, structured description gets evaluated faster and more accurately. Sites without the file require Perplexity to scan multiple pages to determine relevance - and with its speed-optimized architecture, it often moves on before completing that scan.

Google’s AI Overviews don’t directly consume llms.txt the way purpose-built AI crawlers do, but the file’s content reinforces the same signals Google already parses from structured data and meta descriptions.

How We Tested

The llms.txt adoption and scoring numbers in this article come from the AEO Content audit corpus. As of May 2026, that corpus covered 11,068 domains scored with AEORank v5.0, with 503 sites manually re-checked across the live chat, customer support, and B2B SaaS verticals (the verticals where llms.txt adoption is highest).

For the Tidio vs. Crisp comparison we fetched the live /llms.txt from each domain on five separate dates between January and May 2026, confirmed line count and structure each time, and cross-referenced against our scoring engine’s output. The 10+ point lift figure is the difference in overall AEORank score between sites that added a structured llms.txt during the audit window and sites in the same category that did not, controlled for sector and content depth.

The 15% adoption rate is the share of domains in our 11,068-site corpus where a valid /llms.txt returned HTTP 200 with non-trivial content (>100 bytes, headings present). The 25% Y Combinator figure is from a separate sample of 2,500+ YC startup sites we scored in early 2026. Numbers are refreshed when AEORank versions ship; this article was last reviewed against the corpus on 2026-05-29.

External Resources

• llms.txt Specification - https://llmstxt.org
• Anthropic Intro to Claude - https://docs.anthropic.com/en/docs/intro-to-claude
• Schema.org Vocabulary Reference - https://schema.org/docs/schemas.html

Related topics

• structured-data
• robots-txt-ai
• startup-quick-wins