TL;DR: We checked 30 well-known SaaS companies for llms.txt, the proposed standard file that tells AI crawlers what a site actually is. 15 of 26 conclusive checks (57.7%) had something at that path. Only 6 (23.1%) had a file actually built to the real spec. HubSpot, notably, has none. For comparison, adoption across the top 10,000 sites on the whole web sits at just 5.61%. Having a file is a start. Almost none of these companies are using it to actually get recommended.
This article covers:
- What llms.txt actually is, and who proposed it
- The full results: which of 30 leading SaaS companies have one, and which don't
- Why having a file at that path and having a real, spec-compliant llms.txt are two different things
- How SaaS adoption compares to the broader web
- Whether llms.txt actually moves the needle on getting recommended by AI engines, not just crawled by them
HubSpot doesn't have an llms.txt file.
Salesforce does, sort of.
That's notable because in Arobis AI's own research into which CRM ChatGPT recommends more often, HubSpot came out ahead. The company winning the AI recommendation is the one that hasn't gotten around to the file that's supposed to help AI crawlers understand it.
That contradiction is the reason we stopped reading the dozen "what is llms.txt" explainers already published this year and just checked the file ourselves, across 30 recognizable SaaS companies. Here's exactly what we found.
What Is llms.txt?
llms.txt is a proposed standard for a markdown file placed at a website's root (yourdomain.com/llms.txt) that gives large language models a curated, structured summary of a site: what it is, what its main sections are, and links to the most important pages, without forcing an AI crawler to parse an entire site to figure that out.
It was proposed by Jeremy Howard on September 3, 2024. The spec is simple: an H1 heading with the site or project name (the only required part), a blockquote with a short summary, and then H2-delimited sections listing relevant URLs with brief descriptions. There's also an optional "Optional" section for secondary links that can be skipped if an AI system needs a shorter version.
The idea addresses a real constraint: LLM context windows are limited, and a full website is usually too much to process. llms.txt is meant to hand the model exactly what it needs, in the order it needs it. It's a close cousin to the crawlability fundamentals we cover in our guide to what AI crawlers actually read, but narrower and newer, which is exactly why almost nobody has checked how many real companies have actually implemented it well.
Why We Checked 30 Leading SaaS Companies
Every llms.txt article published so far explains what it is. Almost none of them check who actually has one, and none of them look specifically at SaaS, which is the exact category buyers are increasingly researching through AI before they visit a single vendor site.
So on July 10, 2026, we fetched /llms.txt directly on 30 well-known SaaS companies across CRM, support, project management, communication, and marketing categories, and read what actually came back, not just whether the URL returned a 200 status.
Four of the 30 (Canva, Moz, Mailchimp, and Apollo.io) either blocked automated access or returned data we couldn't verify as reliable. Rather than guess, we excluded them. The 26 remaining results are what follows.
The Results: Who Has llms.txt and Who Doesn't
We split the results into three real categories, because "has a file" and "has a good file" turned out to be very different questions.
Out of the 26 companies we could conclusively check, 15 (57.7%) had some form of dedicated content at /llms.txt. Only 6 (23.1%) matched the actual spec Jeremy Howard proposed: a clean H1, a real summary, and organized H2 sections. The rest of that 57.7% had something real and structured, just not built with llms.txt specifically in mind, more like an auto-generated content index than a deliberate AI-readability decision.
Not All llms.txt Files Are Created Equal
Intercom's file opens with a direct statement of purpose: it names canonical domains, sets retrieval rules by topic, and tells AI systems which source to trust when content conflicts. Slack's file does the same job with more personality, but the same structure: clear sections for product, features, solutions, and resources.
Compare that to what several Tier 2 companies have: a long, flat list of blog titles and product links with no real hierarchy, no summary, and no sense that anyone made a deliberate decision about what an AI system should read first. It's the difference between writing a briefing document for a new employee and just forwarding them your entire inbox.
That distinction matters more than the raw adoption number. A flat content dump at /llms.txt doesn't tell an AI engine what your company does, who it's for, or why it should matter in a buyer's answer. It just gives the crawler more to sort through. AI engines already weigh source clarity and structure heavily when deciding what to trust, and a disorganized llms.txt file doesn't help on that front any more than having none at all.
How This Compares to the Broader Web
Independent research using HTTP Archive's crawl data puts llms.txt adoption at 5.61% across the top 10,000 sites on the entire web as of June 2026, up from 1.04% a year earlier, a jump inflated significantly by one e-commerce platform's automatic rollout to its entire customer base.
Measured the same strict way, spec-compliant adoption among the SaaS companies we checked is 23.1%, roughly four times the general web average. SaaS companies are clearly ahead of the broader internet on this. They're just not nearly as ready as the volume of "what is llms.txt" content being published this year would suggest.
Does llms.txt Actually Help You Get Recommended by AI?
Here's the part most llms.txt guides skip entirely: having the file is not the same as being recommended.
llms.txt helps an AI crawler read your site faster and more accurately. It does not, on its own, build the third-party authority, citations, and category association that AI engines actually use to decide who gets recommended in a buyer's answer. A perfectly formatted llms.txt on a site with no external citations and no clear category positioning is a well-organized filing cabinet that nobody outside the company has any reason to trust.
Think of llms.txt as infrastructure, not strategy. It's necessary. It is not sufficient. The six companies in our Tier 1 group made it easy for AI crawlers to read them accurately. Whether that translates into being recommended more often depends on everything AI Search Demand Generation is actually built to influence: entity clarity, category association, and citation authority, not just crawler access.
Should Your SaaS Company Have an llms.txt File?
Yes, but treat it as a floor, not a finish line. A clean, spec-compliant llms.txt costs very little to build and removes one small barrier between your site and accurate AI understanding. Skip it and you're in the same 11-company group as HubSpot, Zoom, and Figma. Build a disorganized one and you're no better off than the companies that skipped it entirely.
If you want to see where your own company currently stands, not just on llms.txt but across the full picture of how AI engines actually describe and recommend your brand, run a free check with the Arobis AI Visibility Checker. It takes a few minutes and shows you the gap most companies don't know they have.
The Arobis POV
An llms.txt file answers one narrow question: can an AI crawler read your site efficiently. It says nothing about whether that AI engine trusts you, associates you with the right category, or recommends you over a competitor when it matters.
Fifteen of the SaaS companies we checked have taken the first step. Six have taken it well. All of them still have the bigger question in front of them: not "can AI read us," but whether AI is actually choosing them when a buyer asks who's best in their category.
Frequently Asked Questions
What is llms.txt?
llms.txt is a proposed standard markdown file, placed at a website's root, that gives AI systems a curated summary of a site's purpose and most important pages. It was proposed by Jeremy Howard in September 2024 to help language models work within limited context windows.
Does my SaaS company need an llms.txt file?
It's a low-cost, low-risk addition worth having, but it is not a substitute for the harder work of building AI recommendation authority. Of the 26 leading SaaS companies we checked, 57.7% had something at that path, though only 23.1% matched the actual spec.
How many companies actually have a real llms.txt file?
In our July 2026 check of 30 well-known SaaS companies, 6 (23.1% of the 26 we could conclusively assess) had a file that matched the real llms.txt spec: a clear H1, summary, and organized sections. Another 9 had some form of content at that path that didn't follow the spec closely.
Does having an llms.txt file improve AI search rankings or recommendations?
Not directly. llms.txt helps AI crawlers read a site's structure more efficiently, but recommendation depends on separate factors: entity clarity, category association, citation authority, and third-party validation. A well-formed llms.txt is infrastructure that supports those efforts, not a replacement for them.
What's the difference between llms.txt and robots.txt?
robots.txt tells crawlers what they are and aren't allowed to access. llms.txt is the opposite in spirit: it's a curated, structured guide to what a site actually is and which pages matter most, meant to help AI systems understand a site rather than restrict them from it.
Why doesn't HubSpot have an llms.txt file if AI visibility matters so much?
It's a reminder that AI recommendation and technical AI-readiness infrastructure don't always move together. HubSpot's strength in AI recommendations, documented in Arobis AI's CRM visibility research, appears to come from years of content depth, category association, and third-party citations, not from having this specific file in place.
