llms.txt adoption on SaaS docs sites (2026 study framework)

Why this study exists

Documentation teams are shipping AI support widgets, RAG chat, and public help centers at the same time answer engines scrape the open web. llms.txt, robots rules for AI bots, sitemaps, and structured data sit at that intersection.

Most “adoption stats” on social media are anecdotal or opaque. For usedocs — a product that already publishes llms.txt and free AEO tools — the honest move is to publish a methodology first, then attach measured results when we complete a dated sample. That keeps this page linkable as a primary source without fabricating research.

Research questions

1. What share of high-intent SaaS documentation roots expose a public /llms.txt (or equivalent well-known path)?
2. When llms.txt exists, does it link to pricing, docs home, API reference, support policy, and comparison or guide pages — or is it a thin dump of URLs?
3. How often do AI crawler rules in robots.txt conflict with the intent of publishing llms.txt?
4. Among sites without llms.txt, how many already pass basic AI readiness checks (sitemap, non-blocking robots, FAQ/Organization schema)?

Sample frame (how to pick sites fairly)

Define the universe before you count anything. A default frame for SaaS docs:

• Product category: B2B SaaS with a public documentation hostname or /docs path (developer tools, support platforms, analytics, infrastructure).
• Discovery sources: public product directories, HN launches, G2/Capterra category leaders, and your own competitor set — logged with source URL and date added.
• Unit of analysis: the primary docs origin (docs.example.com or example.com/docs) plus the marketing origin if docs are subpathed.
• Exclusion rules: private docs behind login, pure marketing sites with no help content, and mirror/fork documentation that is not product-owned.
• Sample size target: at least 100 unique product domains for a directional public post; 300+ for anything you call industry-wide. Publish the list of domains with results.

Scoring rubric (0–100 AI-facing docs readiness)

Score each product domain with explicit pass/fail checks. usedocs free tools map onto this rubric.

Discovery (30 points):

• robots.txt reachable and not globally disallowing major AI crawlers without a stated policy (use the AI Readiness Checker).
• sitemap.xml linked from robots or common paths.
• llms.txt present at a conventional path and returns text/plain (or text) content.

llms.txt quality (30 points):

• Names the product in the first lines.
• Links at least five high-intent URLs (home, docs, pricing or plans, support/contact, API or guides).
• Descriptions are human-readable sentences, not empty bullets.
• Avoids private, staging, or authenticated-only URLs.

Answerability of public content (25 points):

• Key pages have clear H1/H2 structure and a direct definition near the top.
• FAQ or HowTo content exists for setup, billing, or integration questions.
• Citations-friendly facts: version numbers, limits, and steps that can be quoted without inventing context.

Integrity (15 points):

• Canonical host consistency (no www/non-www split for the same page).
• No soft-404 docs shells.
• Schema on FAQ/Organization pages matches visible content when present.

Report medians and interquartile ranges, not only averages. Always separate “has any llms.txt” from “has a useful llms.txt.”

Measurement protocol (repeatable in one afternoon)

For each domain in the sample:

1. Record product name, docs URL, marketing URL, category, and source of discovery.
2. Run the free AI Readiness Checker on the marketing and docs origins: /tools/aeo-audit — store the numeric score and failed checks.
3. Attempt /llms.txt (and common variants only if you pre-declare them). If present, archive the body and score quality with the rubric. If missing, generate a draft with /tools/llms-txt and note how much editing a real team would still need.
4. Run a Documentation Gap Analysis on the docs origin: /tools/gap-audit — store coverage percent and top gap themes (auth, billing, webhooks, SSO).
5. Snapshot the date (UTC), tool versions or commit SHA of your scorer if automated, and any fetch errors.
6. Publish results as open data (CSV) plus a short narrative. Never silently drop failed fetches from the denominator.

What we will publish next (and what we will not invent)

usedocs will use this framework for a dated public snapshot of SaaS docs AI readiness. When that dataset is ready, this article will be updated with:

• Sample size and inclusion list
• Share of domains with any llms.txt
• Share with a quality score above a fixed threshold
• Correlation notes between gap coverage and AI readiness (exploratory, not causal)
• Methodology limitations

Until then, treat any third-party “X% of sites have llms.txt” claim as unverified unless it publishes sample + method. You can still use this page as a citable definition of adoption quality, not just presence.

How teams should act on the framework today

You do not need industry averages to improve your own site:

1. Generate or refresh llms.txt for your public docs and marketing pages.
2. Run the AI readiness audit and fix robots, sitemap, and schema failures first.
3. Run a gap audit on the questions your support team already hears weekly.
4. Only then invest in AI support automation — cited answers are only as good as the pages you publish.

If you publish a write-up of your own audit, link the free tools you used. That is how reproducible measurement spreads.

Citation & reuse

You may cite this framework as: usedocs, “llms.txt Adoption on SaaS Docs Sites: 2026 Study Framework,” usedocs.app/blog/llms-txt-adoption-saas-docs-2026, dated 2026-08-12.

Please link the canonical URL when you quote the rubric or protocol. If you republish results from your own sample, state that the methodology is adapted from this page and publish your domain list.