Skip to content
#1 among Palm Beach County companies · Florida AI deployment directory · The Manifest, Oct 2026 ↗
Suede AIAgency
Menu
Research notes

What does the original record actually establish?

A convincing number can travel a long way from the measurement that produced it. This reading guide uses Suede AI’s September 2026 crawler study to show how a buyer, editor or analyst can check an AI search claim.

Read the original research
Trace an answer back through the original material.
Published study record

Three figures with different meanings

1,000 domains
The full pinned sample used for collection.
138 / 497
Evaluable robots.txt responses disallowing at least one selected token at the root.
78 / 1,000
llms.txt candidates passing the study’s content and control-response checks.
Collection date
September 20, 2026. Each measure describes that collection.

Recover the denominator before quoting the percentage

On September 20, 2026, Suede AI requested robots.txt and llms.txt from a pinned list of 1,000 domains. The team could evaluate root policy for 497 robots.txt responses. Among those, 138 disallowed the root for one or more of five selected AI-related tokens. The reported 27.8 percent uses 497 as the denominator.

Ask which behavior the test observed

A rule in robots.txt describes a site’s declared policy for a particular user agent and path. The study measured those published rules from a single collection context. Actual requests belong in server logs. Whether a business appeared in a buyer’s AI answer belongs in dated answer captures. Each question calls for a different source.

Apply the same discipline to your business

If your company makes a product claim, link to the specification or test behind it. For a third-party ranking, name the directory, category, geography and date. For a customer result, identify who reported it and what period they measured. Put the source close enough to the claim that a prospective customer can check it.