Open dataset of AI assistant referrers
One row per hostname, with the evidence and the date. 16 rows, version 1.0.0, CC BY 4.0.
Last checked · How we check
On this page
Download
CSV
18 columns, header row, UTF-8.
JSON
Same rows plus version, license and count.
How to cite
AnswerTraffic (https://answertraffic.com/referrers), "AI assistant referrers", version 1.0.0, last verified 2026-10-19. CC BY 4.0.
Fields
| Field | Meaning |
|---|---|
| assistant_id, assistant, operator | Short id, display name and company |
| hostname, surface | Referrer hostname as it appears in analytics, and the surface (web, app) it was observed on |
| referrer_behavior | sent, partial, stripped or unknown |
| utm_source_value | Value the assistant is documented to add to links; empty when none is documented |
| evidence_type, evidence_url, evidence_date | own_test, vendor_doc, third_party_test, observed_in_logs or unverified; the source; the month the source was published (for tests, the date of the write-up, not necessarily the day the test ran) |
| first_seen, last_verified | First report we know of (where there is one); the day we last compared the row with its source |
| ga4_default | named, one_source, not_named or unknown (two sources compared) |
| plausible, matomo, posthog | named, not_named or unknown, from each tool's documentation |
| ambiguous | true when the hostname also serves non-AI traffic and cannot be separated; never used in generated filters |
| note | What the evidence says and what it does not |
Changelog
- 1.0.0 (2026-10-19): First release: 16 rows, 10 with a test or a vendor source, the rest marked unverified. Not yet re-tested by us.
Machine-readable changelog: CHANGELOG.md. The same entries are in the Atom feed.
Using the data
The data is licensed under CC BY 4.0: you may copy, change and use it, including commercially, if you credit AnswerTraffic with a link. Please keep the evidence_url and last_verified columns with the rows, because they are the point of the list. If you find an error, write to hello@answertraffic.com and name the row.