What is PanguBot?
PanguBot is attributed to Huawei and described in this directory as collecting web content for the Pangu model family. The attribution is marked unverifiable. That warning matters because a request containing PanguBot cannot be treated as confirmed Huawei traffic from the available evidence.
Huawei's Pangu documentation describes NLP, multimodal, computer-vision, prediction, and scientific-computing models managed through ModelArts Studio. It covers data processing, model training, deployment, and application development, but it does not document a web crawler named PanguBot.
The facts record supplies one stable token, PanguBot, and no complete user-agent example. It also has no Cloudflare catalog entry, published IP verification, or HTTP-signature directory. The token is enough to write a policy rule and search logs, not enough to verify the operator.
If the recorded purpose and attribution are accurate, pages fetched by PanguBot may be used as training material for Pangu models. There is no documented connection to a live AI search engine, retrieval service, or citation feature. Allowing the token should not be presented as a way to gain placement in Huawei AI answers.
Robots compliance is recorded as partial. No operator documentation in the source material explains which directives or request types are covered, so a robots.txt block is useful as a stated preference but cannot be assumed to stop every request. There is no supported basis for attributing the partial status to a particular fetching mode.
Sites can apply a conservative policy without resolving the attribution. Keep nonpublic content behind authentication, use an edge rule when a robots preference is ignored, and avoid granting special access to a client simply because it claims the PanguBot name.
