Research question and hypothesis

Question: Do verified AI-category requests reach two adjacent, equally linked read-only Observatory resources in different formats?

Both equally linked resources may receive verified AI-category requests, with potentially different aggregate counts.

Rationale: Existing trail observations include some verified AI-category exact-path requests, making a fixed paired-resource measurement worthwhile.

Previous evidence: Follows the trail and llms.txt observations by testing aggregate requests to deliberately paired isolated resources.

Setup

Publish one HTML resource and one JSON resource with the same short factual content, linked equally from the Observatory format-study landing page.

Intervention: Publish one HTML resource and one JSON resource with the same short factual content, linked equally from the Observatory format-study landing page.

Measurement protocol

Registered 2026-09-22. Original measurement window: 2026-10-15 UTC for 7 complete UTC days. Collection and future experiment execution are paused.

Primary metric: Successful verified AI-category exact-path requests to each resource during seven complete UTC days.

Exact paths: /observatory/format-study/html/, /observatory/format-study/data.json

Analysis rule: Count successful HTTP 200 requests in verified AI Crawler and AI Search categories separately for each exact path, then sum over the fixed window.

Confounders: Different URL suffixes, caching, independent crawls and Cloudflare classification can affect counts; requests cannot be paired by client.

Invalidation criteria: Any missing complete day, unqueried exact path, changed resource content or unequal landing-page links invalidates a comparative finding.

Result and interpretation

The fixed window has not produced a validated report.

Limitations

Aggregate requests cannot identify a crawler, prove reading, or establish a preference for one format.

Evidence and data

This record as JSON →