ChatGLM-Spider, KimiBot and Timpibot appear in Aiola's crawler directory under Zhipu AI, Moonshot AI and Timpi. Their records describe different collection purposes and provide different evidence for identifying requests. A recorded user agent identifies a claimed crawler name; it does not authenticate the sender.
Who publishes a page about each one?
Zhipu's about page dates ChatGLM's launch to 2023. The ChatGLM-Spider record associates the crawler with Zhipu AI and lists bigmodel.cn. Its documentation field is empty, so that company history provides context without establishing a crawler policy.
The KimiBot record names Moonshot AI and links to a dedicated vendor policy. Unlike the other two records, it also links to a published IP feed. Those fields provide additional material for checking a request's claimed identity.
Timpi's about page reports more than 2.2 billion indexed web pages and over 1,700 decentralized nodes. Its search engine uses its own web index. The Timpibot record names Timpi and lists timpi.com as the vendor domain. The company page describes its infrastructure, while the directory supplies the crawler-specific fields.
What does each crawler collect?
Aiola's ChatGLM-Spider note describes public web collection for Zhipu AI's ChatGLM model ecosystem. The record categorizes it as training and gives its kind as crawl. It does not specify file types, individual datasets, or which model receives any particular page.
KimiBot's directory note describes content potentially used for foundation-model training. That wording leaves the eventual use of any individual page unspecified.
Timpibot's note describes crawling public pages for Timpi's decentralized search index. Its stored category is nevertheless training, and its kind is crawl. This reflects Aiola's classification rule: crawl maps to Training, while search and index map to Indexing.
The bucket therefore needs to be read alongside the crawler's purpose. Timpibot's directory note supports a search-index purpose, without stating that Timpi trains a model on your page. All three records use the training category, including KimiBot.
Which string appears in the log?
The table reproduces Aiola's stored samples and tokens. ChatGLM-Spider and Timpibot have token-only samples, with no documentation URL attached to either record. Treat these as directory identifiers rather than vendor-confirmed complete request headers.
| Crawler | User-agent sample | robots.txt token | IP feed in record |
|---|---|---|---|
| ChatGLM-Spider | ChatGLM-Spider | ChatGLM-Spider | None listed |
| KimiBot | Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; KimiBot/1.0; +https://www.kimi.com/policies/kimi-crawlers | KimiBot | kimibot.json |
| Timpibot | Timpibot | Timpibot | None listed |
Kimi's published header matches the stored sample. Its linked JSON feed contains IPv4 prefixes. The absence of equivalent links on the other records only describes those records; it establishes no worldwide publication claim.
How does Kimi say to use robots.txt?
Kimi's crawler policy recommends robots.txt at the site root, separately for each subdomain. Each crawler has independent settings.
For a full block, its example uses:
User-agent: KimiBot
Disallow: /Its path example instead disallows /private/ and /paid/. Another example uses Crawl-delay: 1, setting a one-second minimum request interval.
Kimi says a training disallow signals that content should not be used for model training. It publishes changing IP ranges for verification and warns that IP blocks can prevent reading updated robots.txt instructions.
What can you say about the other two?
Neither ChatGLM-Spider nor Timpibot has a documentation URL or published IP feed in Aiola's directory data. Their records therefore cannot establish whether either crawler honors a particular robots.txt directive. A company about page does not supply that missing policy.
You can write groups using the stored tokens, but these examples carry no vendor-backed compliance promise:
User-agent: ChatGLM-Spider
Disallow: /
User-agent: Timpibot
Disallow: /Keep those groups distinct from KimiBot's documented example. When describing a rule change, distinguish the rule you published from behavior actually observed in requests. Neither record supplies an IP-based check for the two token-only samples.
How do you collect these hits in Aiola?
Crawlers do not run JavaScript, so the browser tag cannot see them. One server or edge snippet from the Analytics setup guide posts each matching hit. Only known crawler user agents are stored.
What should you check after a rule change?
The AI view shows crawl totals, bots, pages, and activity over time. It includes Answers, Indexing, Training and Other buckets, AI-answer referrals, crawler 404s, and discovery-file health.
All three records use the Training category. Aiola's Analytics documentation describes these views.
A recorded crawl is not proof that the vendor cited the page. An empty period is not proof that the vendor cannot reach the site.
Try it in Aiola
Download Aiola free for macOS 12 or later and Windows 10 or later. Hosted Analytics needs Pro or the 7-day trial, and the pricing page lists both. Open the AI view to inspect requests for these three crawlers.
