Guide
Appearing in Claude
Claude has no ranked index to be listed in. It either knows something about you already, or it goes and looks. Those are two different problems with two different fixes.
Two questions wearing one coat
Asking how to appear in Claude collapses two quite different situations, and the advice is opposite in each.
What it knows without looking is a compressed impression formed from a great deal of text, most of it not yours. You cannot edit it, you cannot submit a correction, and it can be years out of date. What you can do is make the accurate version of your story easy to find everywhere else.
What it finds when it searches is ordinary web content fetched in the moment. This half is fully mechanical and entirely within your control: a page that can be reached and read, saying what you do.
Almost everything useful you can do sits in the second half, which is worth knowing before paying anyone to influence the first.
Three agents, and two of them decide what a user sees
Claude-User is the live fetch. When somebody asks Claude about your site, this is what goes and looks. Block it and Claude cannot open your pages when it is asked to, which is the outcome almost nobody intends when they add a rule about Claude.
Claude-SearchBot reads the web for searches made in Claude. Block it and your pages are left out of what Claude can find when it searches, which costs you in exactly the answers this page is about.
ClaudeBot gathers content for training. Blocking it is a decision about whether your words contribute to future models. It affects neither of the two above.
The mistake to avoid is a blanket rule that catches all three because they share a name. Rules naming Claude-Web or anthropic-ai can simply go: Anthropic no longer lists either. The crawler check reads your robots.txt and reports which of them you are currently allowing.
Why it describes you wrongly, and what fixes it
When a model gets your business wrong, the cause is nearly always that the clearest statement about you on the open web is not on your own site. A homepage that opens with a slogan and never names the trade leaves a gap, and something else fills it, often an old directory entry or a half-right summary on a third-party listing.
The repair is slow but reliable: say the plain thing in the visible words of your own pages, so that the correct version is the easiest one for anything to find.
Questions about appearing in Claude
How do I get my business mentioned by Claude?
There is nothing to submit your site to, so the question is really two questions. For what Claude knows without looking, that comes from training data and from what the wider web says about you, neither of which you control directly. For what it finds when it searches, let Claude-SearchBot and Claude-User reach your pages and make them say plainly what you do.
What is the difference between ClaudeBot, Claude-SearchBot and Claude-User?
ClaudeBot gathers content for training. Claude-SearchBot reads the web so that Claude can find you when it searches. Claude-User fetches a page live when somebody asks Claude about it. The two that change what a user sees right now are Claude-SearchBot and Claude-User. Older rules naming Claude-Web or anthropic-ai can go, because Anthropic no longer lists either.
Should I block ClaudeBot?
That is a judgement about training rather than about visibility, and reasonable people land both ways. Blocking it keeps your content out of training data. It does not stop Claude finding your page when it searches, which is Claude-SearchBot, or opening it when a user asks, which is Claude-User. Decide them separately rather than blocking everything with the word Claude in it.
Why does Claude get facts about my company wrong?
Because anything it knows without looking is a compressed impression of what the web said, which can be out of date or drawn from someone describing you badly. The repair is the same as everywhere: state the facts plainly in the visible text of your own pages, so that when it does look, the correct version is the easiest one to find.
Does Claude cite sources?
When it searches, yes, it shows what it drew on. When answering from what it already knows, there is nothing to cite and it will say so if asked. That distinction matters when you are testing: the same question can produce a cited answer one time and an uncited one the next, depending on whether it chose to look.
Check what Claude can reach and read.
All three of its agents, and every other named AI crawler, read from your live robots.txt. Free, and it takes seconds.