Yes, your SaaS website needs an llms.txt file, and it will probably do more for you than it does for a blog. An llms.txt file is a short, structured text file at the root of your site that tells AI systems which of your pages matter most, in plain language they can parse. For a SaaS product, that means your pricing, your feature pages, your docs, and your changelog get surfaced to AI assistants in a clean order instead of being guessed from a crawl. This guide shows exactly what to put in a SaaS llms.txt, how to write it, and how to publish it in under an hour.
SaaS sites are a special case because the pages that sell the product are not the pages a crawler finds first. A typical SaaS site has a marketing homepage, a pricing page, a dozen feature pages, a help center with hundreds of articles, API reference docs, a changelog, a status page, and a blog. A general-purpose crawler wanders through all of it and builds its own picture of what the product does, often from the wrong pages. An llms.txt file hands the AI a table of contents written by you, so the model reads the pricing page before the 2023 blog archive and the getting-started guide before a deprecated API page.
Why SaaS sites get more from llms.txt than blogs do
Blogs have one job: publish articles. A SaaS site has several jobs at once, and the AI assistant needs to know which pages serve which job. When someone asks an AI assistant to compare project management tools, the assistant needs your pricing, your feature list, and your limits. When someone asks how to set up your product, it needs your docs. When someone asks whether your product had an outage last week, it needs your status page. Without llms.txt, the assistant guesses from search results. With it, you point it at the right pages first.
There is also a trust angle. SaaS buyers ask AI assistants questions that used to go to a sales rep: does it integrate with Slack, is there a free plan, what happens to my data if I cancel. If the assistant answers from an outdated blog post or a third-party review, you lose the deal before you ever hear about it. An llms.txt file that links to current pricing, current integrations, and current terms gives the assistant a better source than a review site. It does not guarantee the assistant will use it, but it stacks the odds in your favor.
Be honest about the limits, though. Log studies published in 2026 found that AI crawlers almost never fetch llms.txt today. One 36-day study counted 1,687 AI-crawler visits and zero llms.txt requests. The file is a cheap bet, not a traffic channel. It costs an hour to write and nothing to host, and it positions your product for the day assistants do start reading it. Think of it as a signpost, not a sales funnel.
What belongs in a SaaS llms.txt file
The llms.txt format is simple: a title, a short description in a blockquote, then sections with headings and markdown links, each with a one-line description. For a SaaS product, five sections cover almost everything an AI assistant needs.
Product: your homepage, pricing, feature overview, and comparison pages. These answer buying questions.
Documentation: getting started, core guides, and API reference. These answer setup and how-to questions.
Integrations: the tools your product connects to. These answer the second-most-common buying question after pricing.
Company and trust: security page, privacy policy, terms, and status page. These answer the questions a cautious buyer asks before entering a card number.
Optional: changelog, blog, and community links. Mark these clearly as optional so the assistant reads them last.
Here is a realistic example for a fictional project management tool. Adapt the structure, not the wording:
# Acme — Project management software for small teams > Acme helps teams plan work, track time, and ship on schedule. Plans start at $9 per user per month, with a free plan for up to 3 users. ## Product - [Pricing](https://example.com/pricing): Plans, limits, and what each tier includes. - [Features](https://example.com/features): Task boards, time tracking, reporting, and automations. - [Acme vs Asana](https://example.com/compare/acme-vs-asana): Side-by-side comparison of features and pricing. ## Documentation - [Getting started](https://example.com/docs/getting-started): Set up your first workspace in 10 minutes. - [API reference](https://example.com/docs/api): REST API endpoints, authentication, and rate limits. - [Import your data](https://example.com/docs/import): Migrate from Trello, Asana, or a spreadsheet. ## Integrations - [Slack integration](https://example.com/integrations/slack): Get task updates and create tasks from Slack. - [Google Workspace](https://example.com/integrations/google): Calendar sync and Drive attachments. ## Company and trust - [Security](https://example.com/security): SOC 2 Type II, encryption, and data residency. - [Privacy policy](https://example.com/privacy): How customer data is stored and used. - [Status](https://status.example.com): Live service status and incident history. ## Optional - [Changelog](https://example.com/changelog): New features and fixes, newest first. - [Blog](https://example.com/blog): Product news and team workflows.
Notice what the file does not contain: no marketing slogans, no keyword stuffing, no duplicate links, and no pages that are already out of date. Every link has a plain description of what the reader will find. That is the whole format. If a line does not help an AI answer a real customer question, cut it.
The pages AI assistants actually need from a SaaS site
Not every page on your site deserves a place in llms.txt. The file should be a shortlist, not a sitemap. Prioritize the pages that answer the questions buyers and users actually ask AI assistants.
Pricing first. Pricing is the most-asked SaaS question in AI chat, and it is also the page most likely to be wrong in an AI answer, because pricing changes and old pages linger in search results. Put your pricing page near the top, and make sure the linked page itself is current. If your pricing page hides numbers behind a sales call, say so in the description. An assistant that knows pricing is gated will say so instead of inventing a number.
Feature and comparison pages second. Comparison pages are gold for AI answers because assistants love structured comparisons. If you have honest comparison pages against your main competitors, list them. If you do not have them, a feature overview page is the next best thing. Avoid linking to comparison pages that are really just ads for your product with no real data. Assistants and readers both discount those fast.
Docs and API reference third. Setup questions are the most common post-purchase use of AI assistants. A getting-started guide and an API reference cover most of them. If your docs are versioned, link to the current version only. Linking to three old versions of the same guide confuses the assistant and wastes its context.
Security, privacy, and status pages. These rarely get linked in llms.txt examples, but they matter for SaaS more than for any other kind of site. A buyer evaluating a tool for their company will ask the assistant about SOC 2, data residency, and uptime. If those pages are in your file, the assistant can answer from your words instead of from a review site's guess.
llms.txt vs llms-full.txt for documentation-heavy products
The standard llms.txt is a shortlist of links. The companion file, llms-full.txt, is the full text of the linked documentation in one file. For SaaS products with large help centers, the difference matters.
Use llms.txt for every SaaS site. It is small, easy to maintain, and it points at your canonical pages. Use llms-full.txt only if your documentation is genuinely useful as a single document, for example a developer API guide or a self-contained onboarding manual. A 200-page help center dumped into one file is not helpful. It is slow to fetch, expensive for the assistant to read, and it goes stale the moment you edit one article.
If you do publish an llms-full.txt, generate it from your docs automatically, not by hand. Most docs platforms can export Markdown. Keep it to your core guides, not your entire knowledge base, and regenerate it on every docs release. A stale llms-full.txt is worse than none at all, because the assistant will quote the old version with confidence.
How to create and publish your SaaS llms.txt in 5 steps
Step 1: List the 10 to 20 pages that matter. Open your analytics and your support inbox. The pages that get the most visits and the questions your support team answers most often are your shortlist. If a page is not in the top 20 by either measure, it probably does not belong in the file.
Step 2: Write one-line descriptions. Each link needs a short, factual description. Write what the page is, not what you wish it were. Bad: Revolutionary AI-powered workflows that transform your team. Good: Automations that move tasks between boards when a status changes.
Step 3: Generate the file. You can write it by hand in any text editor, since the format is plain Markdown. If you want a head start, use a free generator to build the skeleton and then edit the descriptions yourself. Our free llms.txt generator builds a valid file from your URL list with no signup.
Step 4: Publish it at your domain root. The file must live at https://yourdomain.com/llms.txt, exactly that path. On most SaaS stacks this means adding a static file to your web root or a route in your app. If your marketing site and your app live on different subdomains, put a file on each one that needs it, since llms.txt applies per host.
Step 5: Validate it. Open the URL in a browser and check that it loads as plain text. Click every link and confirm none of them 404. Then check that the file is not blocked by your robots.txt. A llms.txt file that crawlers cannot fetch is decoration. Our guide to testing and validating your llms.txt file walks through the full check.
Common mistakes SaaS teams make with llms.txt
Listing every page. An llms.txt with 200 links is a sitemap with worse formatting. The file works because it is curated. If your file is longer than a screen or two, you have listed too much.
Writing ad copy instead of descriptions. AI assistants are trained to discount marketing language. Best-in-class, world-leading, and game-changing tell the assistant nothing. Factual descriptions tell it everything.
Linking to gated pages. If a page requires login, an AI crawler cannot read it, and linking to it in llms.txt just wastes the assistant's time. Link to the public version of the information, or leave the page out.
Forgetting the file exists. Pricing changes, features ship, integrations get added. If nobody owns the llms.txt file, it rots within a quarter. Assign it to the same person who owns the pricing page, and review it whenever pricing or packaging changes.
Treating it as an SEO tool. There is no solid evidence that llms.txt improves search rankings. Google has said no AI system currently uses it for search. Its value is for AI assistants and answer engines, not for Googlebot. Anyone selling llms.txt as an SEO ranking factor is overselling it.
How to measure whether your llms.txt is working
Honest measurement starts with server logs. Check whether any crawler has requested /llms.txt in the last 30 days. In most cases the answer will be none, and that is normal in 2026. The log studies that tracked this found near-zero llms.txt requests across thousands of sites. Do not panic, and do not delete the file. The bet is on where AI systems are heading, not where they are today.
The second check is qualitative. Ask ChatGPT, Claude, Gemini, and Perplexity about your product and see what they say about pricing and features. If their answers match your current pages, your public information is already clear enough. If they quote old pricing or a competitor's feature list, your llms.txt plus cleaner public pages are the fix, but give it time. AI answers change slowly, and a new file does not rewrite a model's memory overnight.
Finally, keep the file fresh and move on. The best llms.txt strategy for a SaaS team is to write a good file once, review it quarterly, and spend the rest of your energy on the things that actually move AI visibility: clear pricing pages, honest comparison pages, and documentation that answers real questions. The file is a signpost. The pages it points to are the destination.
FAQ
Do SaaS sites need an llms.txt file?
Not strictly, but SaaS sites benefit more than most sites because their key information, pricing, features, docs, and integrations, is spread across many page types. An llms.txt file gives AI assistants a curated shortlist instead of leaving them to guess from a crawl.
Should I include pricing in my llms.txt?
Yes. Pricing is the most-asked SaaS question in AI chat. Link to your live pricing page with a factual description. If pricing is gated behind sales, say so in the description so the assistant does not invent a number.
How long should a SaaS llms.txt be?
Aim for 10 to 20 links across 4 to 6 sections. If the file is longer than one or two screens, you have listed too much. Curation is the whole point of the format.
Does llms.txt help SaaS SEO?
There is no solid evidence that it improves search rankings, and Google has said no AI system currently uses llms.txt for search. Its value is for AI assistants and answer engines, not for Googlebot.
How often should I update my SaaS llms.txt?
Review it whenever pricing, packaging, or core features change, and at least once a quarter. Assign one owner, ideally the person who owns the pricing page, so the file does not go stale.
Do I need llms-full.txt too?
Only if your documentation works well as a single document, such as an API guide or onboarding manual. For most SaaS sites, the short llms.txt shortlist is enough, and a giant llms-full.txt goes stale too fast to be worth maintaining by hand.
Ready to build yours? Try the free llms.txt generator, and read our llms.txt vs robots.txt explainer if you are still deciding how the two files work together. For a real-world look at how rarely bots read the file today, see our breakdown of the log studies on whether AI bots read llms.txt.
