Can AI assistants read harvard.edu?

Measured on 2026-09-19 by asking the site 5 times — once as a browser, once as each of the 4 crawlers that feed ChatGPT, Claude, Perplexity, Gemini, Meta AI, Apple Intelligence and Doubao — and comparing what came back. The AI-training opt-out tokens that never crawl are read from robots.txt instead. Tranco rank #313.

What each crawler got back

What to change, in order

1. Let in the 1 crawler your own robots.txt already allows+1 crawler
   ClaudeBot is turned away before reading the page (answered with an HTTP error), while robots.txt permits it — so this block is not written in your site. No CDN signature was found in the response headers, so the refusal comes from the origin server itself or from a WAF this index does not recognise. Fixing it takes this domain from 3 to 4 of 4 crawlers.
2. Publish llms.txt
   llms.txt is missing, so an assistant has to discover the site by following links. llms.txt is the emerging convention for telling an assistant which pages actually matter.

Get told if this changes

One email only when a measured crawler flips on harvard.edu, served to refused or back. No schedule, no newsletter; double opt-in, one-click stop.

The checks

• Answers AI agents like it answers people80%
• Readable without JavaScript100%
• robots.txt lets the crawlers in100%
• Facts in JSON-LD67%
• Content reachable, not buried60%
• Publishes a map of itself45%
• Plain structure97%

History

2026-09-13: 84 · 2026-09-14: 84 · 2026-09-15: 80 · 2026-09-16: 85 · 2026-09-17: 85 · 2026-09-18: 85 · 2026-09-19: 80

Sites with a similar score

Browse the whole index →

Embed this score

Put the badge on harvard.edu — it links back here, and re-measures every time this index re-crawls.

<a href="https://aivis.lumnika.com/ai-readiness/harvard.edu"><img src="https://aivis.lumnika.com/ai-readiness/harvard.edu/badge.svg" alt="AI readability"></a>