.png)
Last updated: July 2026
.webp)
llms.txt is a proposed standard: a plain text (markdown) file you place at [yourdomain.com/llms.txt](<https://yourdomain.com/llms.txt>) that summarizes your site for AI models and points them to your most useful pages. Think of it as a robots.txt for language models, written for readers rather than crawlers. It is quick to write, costs nothing to host, and does no harm even if adoption stays partial.
Let us cover the format, how to write one, and an honest read on how much it actually does today.
What Is llms.txt
.jpeg)
The idea, proposed by Jeremy Howard in 2024, is that AI models struggle to process full HTML pages efficiently: navigation, ads, scripts, and boilerplate all compete with the actual content for the model's limited context window. llms.txt gets around that by giving models a clean, curated, markdown-formatted summary instead: what your site does, who it is for, and direct links to the pages that matter most.
It lives at a fixed, predictable location: /llms.txt at your domain root, the same convention robots.txt and sitemap.xml use. This makes it easy for both humans and machines to find without you having to submit it anywhere.
The Format
The proposed spec is deliberately simple markdown. A typical structure looks like this:
Markdown
# Company Name
> One or two sentence summary of what you do and who you serve.
## Docs
- [Getting Started](<https://example.com/docs/start>): Setup guide
- [API Reference](<https://example.com/docs/api>): Full endpoint list
## Optional
- [Blog](<https://example.com/blog>): Product updates and guides
The H1 is your site or company name. The blockquote underneath is your one-line summary, the part most likely to get lifted directly into an AI answer. Each H2 groups related links, and each bullet is a link plus a short description of what is behind it. An "Optional" section signals lower-priority pages a model can skip if it is short on space.
How to Write One
.png)
- List your five to fifteen most important pages. Not everything, just the pages you would want an AI assistant to read first if it could only read a handful: your homepage, your core product or feature pages, your pricing, and your best explainer content.
- Write a one-sentence company summary. Plain language, no jargon, no marketing fluff. This is the line most likely to get paraphrased back to a user.
- Group links by category. Docs, product, pricing, guides, whatever fits how your site is organized.
- Add a one-line description per link. Tell the model what it will find there before it clicks through.
- Save it as plain text at
/llms.txt. No special hosting needed, it is just a static file. - Keep it current. Update it when you publish something genuinely important, the same discipline you would apply to a sitemap.
For a deeper look at making your actual pages easier for AI systems to trust and cite, not just discover, see our guide on how to build a website that AI wants to cite.
An Honest Note on Support
Here is the part most articles skip: as of mid-2026, no major AI vendor, not OpenAI, not Anthropic, not Google, not Perplexity, has publicly confirmed that their models read or prioritize llms.txt files during training or retrieval. It is a community-proposed convention, not an official standard, and adoption has been driven mostly by developer tools and documentation platforms rather than mandated by the model providers themselves.
That does not mean it is pointless. It costs almost nothing to add, it forces you to write a clean one-line summary of your business (a useful exercise on its own), and if support does expand, you are already there. Treat it as a low-cost bet, not a guaranteed ranking lever. It is one piece of a broader generative engine optimization approach, where emotional design and layout hierarchy still carry significant weight, not a replacement for foundational technical SEO.
Why Cromojo Ships One
We publish our own llms.txt because we ask customers to do the same for their sites, and we think tools should follow the advice they give. It is a five-minute file. There is no reason not to have one, and if it turns out to matter more than expected, you will already be ahead rather than scrambling to catch up. Being fast to adopt small, low-cost changes like this is also just good practice for staying findable, the same instinct behind getting indexed quickly through automated indexing rather than waiting weeks for a crawler to notice you.
Does llms.txt Help With Google AI Overviews?
Not directly. AI Overviews are built from Google's own index and crawling, not from a separate llms.txt-style feed. What actually influences whether you show up there is indexability, structured data, and clear, direct answers on the page itself.
If you are troubleshooting why your content isn't appearing in search or generative answer engines, start with our article on why your pages aren't getting indexed or use a Google index checker tool to verify your page status. For non-Google engines powering AI chat tools, our Bing Webmaster Tools 2026 setup guide covers how to ensure complete visibility across Microsoft and OpenAI's crawling networks.
Where This Fits
llms.txt is a small, sensible addition to a much bigger picture. Getting cited by AI systems depends far more on having clear, well-structured, genuinely useful content than on any single file. For a complete blueprint on discovery and crawl efficiency, read our complete guide to Google indexing.
Once you have both in place, the next question is whether your site is even getting indexed fast enough for AI systems and search engines to find that content in the first place, which is where Cromojo's Indexing module comes in: it pushes new and updated pages to Google and Bing the moment you publish, instead of waiting for a crawl.
.png)
