Amzn-User
Amazon's user-triggered fetcher. Amazon states it may not follow all robots.txt directives.
Reference reviewed by Sona
| Operator | Amazon |
|---|---|
| Powers | Live page fetches when Alexa and Amazon AI act for a user |
| Purpose | User-triggered fetch |
| User-agent token | Amzn-User |
| Respects robots.txt | No (user-triggered) |
Amzn-User fires when a user action requires fetching current information from the web - an Alexa question whose answer depends on a live page, for instance. It is Amazon's counterpart to ChatGPT-User and Claude-User: on-demand, human-initiated fetches rather than scheduled crawling.
Amazon's language on compliance is qualified, and the qualification is the point: it states that Amzn-User "may not follow all robots.txt directives" because the traffic is user-initiated. This is the same reasoning OpenAI, Perplexity, Google, and Meta apply to their equivalents, and the same practical conclusion follows - a Disallow here is a preference rather than a control.
Amazon does confirm the useful half: Amzn-User is not used for generative AI model training. So allowing it means live retrieval to answer someone's question, not contribution to a training corpus. Only Amazonbot carries the training role.
Verification is well supported. Amazon publishes a live IP address list for this token, distinct from the Amazonbot and Amzn-SearchBot lists - which is what makes deliberate edge enforcement possible if you need the block to hold.
How Amzn-User behaves
- Amazon states it may not follow all robots.txt directives, on the basis that fetches are user-initiated.
- Documented as not used for generative AI model training.
- Single-page, on-demand fetches driven by user questions rather than a crawl schedule.
- Has its own live IP address list, distinct from Amazon's other two crawler lists.
Full user-agent string
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Amzn-User/0.1)
Allow Amzn-User
Users asking Alexa or Amazon AI about your content get answers from the live page, with no training use involved.
User-agent: Amzn-User Allow: /
Block Amzn-User
You want to prevent all programmatic Amazon access including user-initiated fetches - enforce it at the WAF, since robots.txt may not be followed.
User-agent: Amzn-User Disallow: /
Heads up: Amzn-User does not reliably honor robots.txt. To enforce a block, combine the rule above with server- or firewall-level filtering of the user-agent.
How to verify Amzn-User
Published IP range feed
Amazon publishes a live IP address list for Amzn-User, separate from the Amazonbot and Amzn-SearchBot files. That separation is what makes enforcement practical here: because robots.txt may not be honored, blocking real Amazon fetches deliberately - rather than blocking anything that claims the name - means matching this token's own published ranges at the edge.
Check an IP against this botCommonly confused with Amzn-User
Amzn-SearchBot indexes on a schedule and does respect robots.txt; this token fetches live and may not. Easy to conflate given the naming.
The same role and the same qualified robots.txt language at OpenAI. Both need edge enforcement if a block must hold.
Amazonbot is the only Amazon crawler documented as potentially training AI models. This one explicitly is not.
Amzn-User FAQs
Does Amzn-User respect robots.txt?
Not fully. Amazon states it may not follow all robots.txt directives because the traffic is user-initiated. Treat a Disallow as advisory and enforce at the server, CDN, or WAF if you need it to hold.
Is my content used for training if I allow this?
No. Amazon documents that Amzn-User is not used for generative AI model training. Only Amazonbot carries that role.
How do I block Amzn-User properly?
Filter the user-agent at the edge and scope the rule to the live IP address list Amazon publishes for this token specifically - not the Amazonbot or Amzn-SearchBot lists, which cover different crawlers.
Why does Amazon run three separate crawlers?
To separate purposes cleanly: Amazonbot for general crawling that may train models, Amzn-SearchBot for search indexing without training, and Amzn-User for live user-initiated fetches. Each has its own token and its own IP list.
Can Amzn-User read your page right now?
Test any URL and see exactly what AI crawlers receive.