Do AI crawlers execute JavaScript? A public canary

The most cited measurement on this question is Vercel and MERJ's December 2024 analysis of more than 500 million GPTBot fetches, which observed no JavaScript execution at all. As of September 2026 no public test has replicated it. This page is a small, permanent attempt to keep the answer current.

How it works

Three unique markers live on this page. Each one, if it later shows up in an AI assistant's answer, proves a different thing.

• Marker one is in the served HTML: dic-62d7e3451b34-alpha. It only proves the page was fetched and retained.
• Marker two is never present in any byte this server sends. It is transmitted as an XOR-encoded array with a key that changes on every request, and it is reassembled in the DOM only by running the script below. It appears here: (requires JavaScript execution)
• Marker three is a plain string inside an external file, /e.js, and is never inserted into the page. It separates crawlers that download and read JavaScript source from crawlers that actually run it.

Why the second marker is hidden from the source

GPTBot downloads JavaScript files in about 11.5% of its requests and ClaudeBot in about 23.84%, without executing them. If the marker were a literal in the script, a crawler that merely ingested the file would surface it, and the test would report rendering where there is none.

Verification

Crawl visits are logged server-side with the originating IP and autonomous system, and are checked against the IP ranges published by the operators themselves. A user-agent string on its own proves nothing.