KazevoBot
If you found KazevoBot in your access log, this page explains what it was doing there. The short version: it is not a crawler. It has no frontier, no schedule pointed at the open web, and it never follows a link on your site to discover a second page. Every request it makes is one URL, fetched once, because a person clicked something in Kazevo.
What it is not
- It does not spider. One request means one page. KazevoBot will not walk your navigation, your sitemap, or your internal links.
- It does not recrawl on a schedule. There is no background job that revisits a URL it has seen. If it comes back, someone ran another check.
- It does not collect training data. Kazevo trains no models of its own, and fetched pages are not sold, syndicated, or accumulated into a corpus. What we will say plainly, because a bot page that omits it is not worth reading: where a fetch feeds a content brief, the structure extracted from it is sent to the third-party models Kazevo runs on (currently OpenAI and Google) as part of producing that brief.
- It does not republish your content. Where a page is read for competitive analysis, what is extracted is its heading structure — the shape of the answer, not the prose.
The requests it makes
KazevoBot sends a different purpose string for each kind of fetch, so you can tell them apart in a log rather than guessing. These are the complete set.
| User-Agent | What it fetches | What triggers it |
|---|---|---|
KazevoBot/1.0 (+https://kazevo.com/bot; robots check) | Your /robots.txt, and nothing else. | Us, immediately before fetching a page of yours for competitor analysis or a brand presence check — to find out whether we are allowed to. Cached briefly, so a burst of checks reads it once. |
KazevoBot/1.0 (+https://kazevo.com/bot; retrieval eligibility check) | Your /robots.txt, your homepage, and /llms.txt if you publish one. | A Kazevo customer running an eligibility check on a site they have added to their workspace. |
KazevoBot/1.0 (+https://kazevo.com/bot; content audit) | One page, at a URL typed into the audit tool. | A customer auditing a specific page for answer structure and schema. |
KazevoBot/1.0 (+https://kazevo.com/bot; competitive content analysis) | The heading structure of a page already ranking for a query, to brief a writer on what the answer needs to cover. | A customer generating a content brief. The URLs come from a search-result set, never from following links on your site. |
KazevoBot/1.0 (+https://kazevo.com/bot; entity audit) | A customer’s own site, checking whether their brand resolves to one identity. | A customer running the entity audit on a site they own. |
KazevoBot/1.0 (+https://kazevo.com/bot; page repair analysis) | The live HTML of one of your own pages, so a repair is scored against what is actually published rather than a stored draft. | A customer repairing a page on a site they have connected. |
KazevoBot/1.0 (+https://kazevo.com/bot; brand presence check) | A single third-party page — a roundup, listicle or directory — to check whether a brand name already appears on it. | A customer reviewing source opportunities, to avoid pitching a page that already mentions them. |
One request does not carry the KazevoBot token. The free visibility checker, which anyone can run against their own domain without an account, identifies itself separately:
KazevoVisibilityCheck/1.0 (+https://www.kazevo.com/tools/ai-visibility-check)Rate and volume
Because every fetch is tied to a person clicking something, volume is bounded by product usage rather than by a crawl budget. A check reads a handful of URLs at most, requests time out after 10–15 seconds, and redirects are followed. You should not see sustained traffic from KazevoBot; if you do, we would rather hear about it than not.
Blocking it
You are entitled to, and nothing in Kazevo penalises a site for it. The supported way is robots.txt, and it is honoured:
User-agent: KazevoBot
Disallow: /A blanket User-agent: * rule applies to us too — you do not have to name us to be respected. Three details worth stating, because they are the ones people get burned by elsewhere:
- No robots.txt means yes. A 404 — or a 403 on the file itself — is read as "this site published no rules", which is what every real crawler does.
- A broken robots.txt means no. If your robots.txt returns a 5xx or we cannot reach it, we do not fetch the page. Kazevo tells its own customers that a persistent 5xx on robots.txt reads as a site-wide disallow; behaving otherwise would be advice we do not take ourselves.
- It applies to the fetches that land on you unasked — competitor analysis and the brand presence check. Checks a Kazevo customer runs against a site they have registered as their own are not gated: the eligibility check exists to fetch and evaluate your robots.txt, and gating it on the file it is reporting on would break the diagnostic precisely when the answer matters.
If you would rather not rely on us behaving, block by user agent at your server, CDN or WAF. For example, in nginx:
if ($http_user_agent ~* "KazevoBot") { return 403; }When a site blocks us, the customer who ran the check is told the page was blocked. Kazevo does not fill the gap with an inferred or invented result — an unread page reads as unread.
Getting in touch
Questions about a request in your logs, or a block you want us to honour at our end, go to [email protected]. Include the user-agent string and a timestamp and we can tell you which check it came from.