Why we fetched your site
Somebody typed the domain into Rankli. We write SEO articles in a business’s own voice, and the only way to learn that voice is to read the pages they already publish, so we read a handful of pages once and write from what we found.
RankliBot does not crawl the web. It has no index, no discovery queue and no schedule. It fetches a site because a person typed that domain into a form minutes earlier, and it stops when it has enough to work with.
There are two such forms, and one of them does not require an account. A customer entering their domain during setup is the first. The second is our public analyser, where anyone can ask what we would write for a site, so a visit can be someone evaluating us against your site rather than your own team. That page is metered for exactly this reason: at most three visits to any one site in a day, whoever asks, and the result is held for 24 hours so a second person asking about the same domain costs your server nothing.
The third is the agent readiness check, which asks for exactly five URLs: the homepage, /robots.txt, /llms.txt, /llms-full.txt and /sitemap.xml, once each, inside a twelve-second budget, at most three times a day for any one site. The report is stored, so a shared link to it never visits your site again.
Exactly what it requests
- The homepage.
- Up to five internal links from that homepage, chosen for how likely they are to describe the business, six pages in total, at most.
- /sitemap.xml, and one child sitemap if the first is an index. We keep up to 200 URLs from it as internal-linking targets, and fetch none of them.
The whole visit is bounded by a twenty-second budget, shared across every request. It runs once per request to read a site, never on a schedule. We request HTML and read the text; we do not execute JavaScript, submit forms, follow links off your domain, or fetch images, video or files. The public analyser skips /sitemap.xml altogether: it links no articles, so it has no use for the list.
Blocking it
Because a fetch only happens when someone asks us to read that specific site, the simplest fix is to tell them to stop. If you would rather block it at the edge, match the user-agent string RankliBot. Requests come from Vercel’s network, so there is no fixed IP range to allow-list.
Blocking it is fine. When a site is unreadable to us we say so on the spot and ask its owner to describe the business in their own words, and every article after that is written from what they told us rather than from a guess.
The other string you might see
rankli-publisher/1.0 (+https://www.rankli.io) is a different job: it posts finished articles into a CMS that a customer has connected with their own credentials. It only ever talks to a site whose owner gave us a key, and it only writes where they told it to.
Getting hold of us
Write to milletgilmore@gmail.com if RankliBot did something you did not expect. See also our privacy policy.