llms.txt adoption by sector: fixed-panel data
Sector rates can describe this panel, but they cannot represent the entire Web. The value lies in a stable denominator and repeatable measurement.
Last updated:
In the 3 August 2026 run, developer tools and SaaS had the highest publication rates in the fixed panel: 68.9% and 66.7%. Media had 0 files among 14 reachable hosts. These are panel observations, not estimates for all sites in each sector.
Key takeaways
- 113 of 218 reachable hosts served a non-HTML root response whose first content line was an H1.
- Sector rates ranged from 0% to 68.9% in the panel.
- A stable denominator supports trends, not Web-wide extrapolation.
Which sectors published the most files?
Developer tools recorded 42 of 61 reachable hosts, or 68.9%. SaaS recorded 32 of 48, or 66.7%. E-commerce had 7 of 13, AI and machine learning 19 of 36, and fintech 8 of 16.
At the lower end, the “Other” group had 4 of 23, documentation platforms 1 of 7, and media 0 of 14. The full machine-readable table is published in adoption-stats.json.
Why use a fixed panel?
A changing list makes trend interpretation impossible: a higher rate could come from adding easy adopters rather than real publication changes. The public panel therefore keeps named hosts and categories stable, tests the same root URL, and excludes network failures from the reachable denominator.
The panel is curated toward relevant technology and digital organizations. That makes it useful for repeated comparison inside the sample, while making it unsuitable for a claim about all domains on the Web.
How to interpret the differences
The proposal is most obviously applicable to structured documentation and agent-facing products. The observed developer-tool rate is consistent with that use case, but the benchmark does not prove why any host published its file.
Avoid causal stories about industry maturity, traffic, or AI visibility. Publication is binary in this measurement. File usefulness varies widely, as the size benchmark and verified examples show.
Sampling limits and next runs
Category samples are small, especially documentation platforms and e-commerce. One host can move those rates materially. Category assignments are curated and cannot represent every business model.
The reliable next step is repetition. Weekly checks can show which fixed hosts appear, disappear, or become unreachable. The adoption page publishes the current snapshot and methodology.
Is 51.8% the Web-wide adoption rate?
No. It is the rate inside this curated fixed panel.
Why exclude unreachable hosts?
A network failure does not reveal whether a file exists. It remains reported separately.
Can sector rates predict agent traffic?
No. This dataset measures publication only.