BrandAnchor Research ? 2026 ? v1.0
A file is not the same as commercial truth.
We audited 28,015 domains to see what AI-readable files actually contain?and whether commercial facts agree across official sources.
THE COHORT
28,015
audited domains: 27,403 B2B SaaS and 612 agencies, distilled from 434,873 candidate domains.
Unless stated otherwise, percentages use the B2B SaaS cohort or the explicitly named subset.
Research by Yair Zilberman ? BrandAnchor.io
Zilberman, Y. (2026). 28,015-Domain B2B SaaS Brand Truth Benchmark (v1.0). BrandAnchor.io. https://brandanchor.io/research/28k-saas-brand-truth
01 / Validate the response
A 200 response can be an HTML page in disguise.
We checked both HTTP status and response content. A homepage redirect or single-page application shell returning 200 does not count as a valid Markdown file.
24.20%
Valid /llms.txt
6,631 / 27,403 SaaS domains served validated plain text or Markdown.
2.69%
HTML fallback at /llms.txt
736 / 27,403 domains; 9.99% of the 7,367 raw HTTP 200 responses.
24.37%
Across all audited domains
6,826 / 28,015 SaaS and agency domains served valid /llms.txt.
02 / Read inside the file
Presence tells only part of the story.
Among the 6,631 validated SaaS /llms.txt files, the commercial detail available to a reader varies substantially.
80.26%
No numeric plan prices
5,322 / 6,631 files are link lists or boilerplate with zero numeric plan prices.
56.58%
Live prices omitted
1,514 / 2,676 valid files associated with a self-serve numeric /pricing page omit all of those prices.
13.18%
Substantive pricing
874 / 6,631 files contain plan names, numeric prices and billing unit or cadence: 3.19% of all SaaS domains.
26.54%
Markdown versus pricing
232 / 874 in the aggregate?s price-bearing-file denominator have a live Markdown price contradicting /pricing. The source describes Markdown endpoints collectively, so this is not limited to /llms.txt.
Across the SaaS cohort, 442 / 27,403 (1.61%) domains have verified cross-page commercial conflicts involving numeric prices, seats or free-tier limits. These are detected conflicts in this audit, not a claim that all other pages agree.
03 / How files are generated
Generator signals in the validated files.
The aggregate classifies generator signals among 6,631 validated SaaS /llms.txt files. Custom / Other includes files without an identified named generator; it should not be read as proof that every file was hand-written.
| Generator | Files | Share |
|---|---|---|
| Yoast SEO | 191 | 2.88% |
| Mintlify | 39 | 0.59% |
| GitBook | 9 | 0.14% |
| AIOSEO | 35 | 0.53% |
| ReadMe | 33 | 0.50% |
| Rank Math | 46 | 0.69% |
| Docusaurus | 59 | 0.89% |
| Custom / Other | 6,219 | 93.79% |
04 / Context & method
Different studies answer different questions.
A focused SaaS cohort
This benchmark starts with a deduplicated, selected B2B SaaS and agency cohort. It validates response content and examines commercial information and conflicts. It is not a random sample of the web or a measurement of AI citation impact.
A broader web perspective
Surfer?s 37,703-site general-web study offers a complementary, broader lens. Cohort selection, endpoint validation, and measurement goals can differ; adoption figures should not be treated as directly interchangeable or as evidence that either population represents the other.
Put the benchmark to work
Check what your own official sources say.
The $99 Brand Truth Check covers up to 5 public official URLs and 3 commercial facts for one English-language B2B SaaS product, plus two buyer observations and a post-fix re-check.