LH047 search log
Searches run with the harness's WebSearch tool (standard mode) on 8 November 2026; times UTC, approximate (from the session's order of work). A result is a lead, not a source: only pages fetched with scripts/fetch.py and kept under data/reads/ are cited. WebSearch requests are made by the search provider, not by Lighthouse to the sites listed.
Before the brief (part A, sources that touch no outcome)
- about 20:28, "Content Signals Policy robots.txt search ai-input ai-train contentsignals.org": Cloudflare's managed robots.txt documentation and blog post (opened, catalogue-2/002, 003); third-party summaries not opened.
- about 20:28, "IETF aipref vocabulary draft draft-ietf-aipref-vocab 2026": mirror copies of revision 07 and the editor's copies (editor's copies opened, catalogue-2/004, 005); revision numbers checked against LH045's census instead of datatracker (an audited host).
- about 20:28, "Web Bot Auth IETF webbotauth architecture draft HTTP message signatures bots": revision 05 at www.ietf.org (opened, catalogue-2/006); datatracker links not opened (audited host).
- about 20:28, "Cloudflare verified bots policy signed agents documentation": Cloudflare's pages (opened, catalogue-2/007 to 010).
- about 20:28, "Anubis proof-of-work AI scraper protection documentation techaro": anubis.techaro.lol/docs (opened; answered with its own challenge, catalogue-2/011); LWN's article already kept by LH044.
- about 20:28, "Really Simple Licensing RSL standard 1.0 specification robots.txt License directive": rslstandard.org's robots.txt guide (opened, catalogue-2/012).
After the brief (part B)
- about 20:47, "TollBit State of the Bots report AI bots bypassing robots.txt share 2026": tollbit.com reports (opened, evidence-1/001 to 003); Search Engine Journal, PPC Land, Digiday not opened.
- about 20:47, "Cloudflare blog crawlers ignoring robots.txt measurement AI crawler compliance 2026": Cloudflare's robots.txt tab changelog (opened, evidence-1/012); third-party compliance claims not opened (weak sourcing per the summary).
- about 20:47, "arXiv empirical study do AI crawlers respect robots.txt directives scrapers selectively": Kim et al. 2025 (arXiv, an audited host: not opened).
- about 20:47, "ClaudeBot ignored robots.txt complaint website operator Anthropic crawler traffic report": secondary reports of iFixit's 2024 complaint; none opened.
- about 20:47, "Cloudflare Perplexity stealth undeclared crawlers no-crawl directives blog": Cloudflare's post (opened, evidence-1/004).
- about 20:47, "Somesite I Used To Crawl AI crawlers robots.txt efficacy content creators IMC": the authors' PDF at sysnet.ucsd.edu (opened, evidence-1/007); arXiv and datatracker copies not opened (audited hosts).
- about 20:49, "Reddit lawsuit Anthropic bots accessed site more than 100,000 times complaint June 2025": Engadget's report (opened, evidence-1/009); others not opened.
- about 20:49, "Cloudflare crawl-to-referral ratio Anthropic OpenAI blog crawl to refer": Cloudflare's post of 29 August 2025 (opened, evidence-1/005).
- about 20:49, "TollBit State of the Bots H1 2026 report ChatGPT-User 54% Europe robots.txt disallowed": confirmed the figure is in TollBit's own report (already fetched).
- about 20:50, a query for The Verge's iFixit and Reddit articles: no Verge article found; one tried by URL was not sent (its robots.txt disallows anthropic-ai).
- about 20:50, "Scrapers selectively respect robots.txt directives IMC 2025 ACM pdf Wenger Duke": the ACM DOI (requested; answered with a challenge, evidence-1/008).
Part A also read, from operators found without a search: Google's robots meta tag, common crawler and user-triggered fetcher pages, Anthropic's, OpenAI's and Perplexity's bot pages, llmstxt.org, W3C's TDMRep report, Cloudflare's pay-per-crawl page, ai.robots.txt through jsDelivr, PyPI's JSON API documentation (docs.pypi.org), and RFCs 9309, 6585, 9110 and 9421.
From studies/LH047/searchlog.md in the repository, last changed 8 November 2026.