Claude Rank Tracking: How to Track Your Brand in Claude
How to track whether Claude mentions and cites your brand: when it searches, the three Anthropic crawlers, a manual protocol, and the tools that cover it.
By Citlyze Research. Published . 5 min read.
On this page
Claude rank tracking means running your buyer prompts through Claude on a schedule and recording whether it names your brand, where in the answer, and which sources it cites. Claude decides on its own when to search the web, so the same question can come back as a search answer with citations one day and an answer from training data the next. A useful tracking program measures both, separately, over repeated runs.
Claude sends less traffic than ChatGPT, but the people who use it skew technical and tend to ask detailed buying questions. If you sell software, developer tools or professional services, it is often the second assistant worth tracking. This guide covers what makes Claude different to measure, the crawler rules that decide whether it can read your site, and how to track it by hand or with a tool.
When Claude searches, and when it doesn't
Anthropic's web search guide says Claude invokes its search tool when a question benefits from current information, and in the current experience there is no toggle: Claude searches when it helps. Users can still force a search by asking for one, or stop it by telling Claude not to search. When it does search, Anthropic says every response includes citations, with links to the sources.
That gives you two kinds of answer to track:
- Search answers come with a source list. You can see which pages Claude read, whether one of them is yours, and which third-party pages it trusted for your category.
- Answers from training data have no sources. They show what the model already believes about your category, which changes only when the model is retrained.
The split matters for diagnosis. A brand missing from search answers usually has a retrieval problem: the pages Claude finds for that question don't mention it. A brand missing from no-search answers has a reputation problem that content shipped this month won't fix quickly. Record which kind each answer was, or the two will blur into one number that moves for reasons you can't explain.
Which index Claude searches is not something Anthropic documents for text results. Its search guide says only that image results come from Bing. So treat ranking well in any one search engine as helpful, not as a guarantee of retrieval.
The three Anthropic crawlers
Claude reaches your site through three user agents, and Anthropic's crawler page gives each a different job:
| User agent | What it does | What blocking it stops |
|---|---|---|
| ClaudeBot | Collects web content that may be used to train Anthropic's models | Your future pages entering training data |
| Claude-User | Fetches pages when a Claude user asks a question | Claude reading your page to answer that user |
| Claude-SearchBot | Indexes pages to improve Claude's search results | Your pages being indexed for Claude's search |
The common mistake is a blanket rule. Blocking every Anthropic agent to keep content out of training also blocks Claude-User and Claude-SearchBot, and with them every chance of being cited in a search answer. If your concern is training, block ClaudeBot by name and leave the other two allowed:
User-agent: ClaudeBot
Disallow: /
Anthropic says its bots honor robots.txt and that the rule must sit on every subdomain you want covered. It also warns that blocking its IP addresses may not work reliably, so use robots.txt rather than a firewall rule. Check the CDN too: a bot-fighting setting there can block all three agents while robots.txt allows them. Our AI crawlers guide covers the CDN side, and the free AI crawler access checker shows what your robots.txt currently tells each agent.
What to measure
Score every Claude answer on the same fields you use for other engines, so the numbers compare:
- Mentioned or not. Whether your brand appears anywhere in the answer.
- Position. Named first, listed among others, or added as an afterthought.
- Cited or not. For search answers, whether one of your URLs is in the source list.
- Sources. Every domain Claude cited, so you can see which third-party pages shape the answer.
- Sentiment and accuracy. Whether the description is correct and favorable, including pricing and positioning.
- Answer type. Search or no search, as above.
Mention rate across repeated runs is the headline number. One answer tells you almost nothing, because identical prompts return different answers on every engine, and Claude is no exception.
Tracking Claude by hand
A manual baseline takes an afternoon and tells you whether Claude is worth paying to track:
- Pick 15 to 20 buying questions in full sentences, the way a buyer would type them. Finding buyer prompts covers how to build the list.
- Run each one in a fresh chat, three times, from an account with no history about your company. Claude's memory and project context steer answers toward what it already knows about you.
- Log the fields above in a sheet, one row per run.
- Repeat the next week with the same wording. Change nothing in the prompts, or you can't tell a real move from a reworded question.
On a free Claude account, Anthropic notes that usage limits reset every five hours, so spread a large set across the day. After two weeks you have a mention rate per prompt with enough runs to see whether you are absent, present but unstable, or reliably named.
Tools that track Claude
Claude coverage varies more than any other engine's, and it is often an extra. Read the plan page, not the homepage, before you buy.
Citlyze tracks Claude as a paid add-on on every plan. Each pack tracks up to 25 of your prompts in Claude every day, with its answers and the sources it cites, and you can add up to eight packs. Claude then shows up as its own engine in filters and charts, next to the seven answer surfaces every plan includes. We build Citlyze, so test that claim on a trial like any other. Pricing has the details.
Rankscale names Claude among the engines it tracks, from $17/month billed yearly, which makes it the cheapest way to see Claude next to many other engines.
Promptwatch lists Claude among the models its plans track, from $95/month.
Otterly.AI offers Claude on every plan as one of its seven engines, at three credits per answer instead of one.
For the wider field, sorted by engine coverage, see the best AI rank trackers.
Free Claude tracking
There is no first-party report for Claude: Anthropic offers nothing like Bing Webmaster Tools' Copilot citations or ChatGPT's Business Hub. So free tracking is the manual protocol, plus one signal from your own logs. Claude-User visits mean someone asked Claude a question and it fetched your page to answer. Count them per page in your server or CDN logs and you have a rough measure of which pages Claude reads on users' behalf, even though it doesn't tell you what the answer said.
The manual route stays practical at around 20 prompts. Beyond that, or once you want competitors scored on the same runs, the hours add up faster than a tool's fee.
Start with the questions you lose elsewhere
If you already track ChatGPT or Perplexity, start your Claude set with the prompts where those engines leave you out. If Claude names you where the others don't, the gap is in the sources the others trust, which is useful to know. If Claude leaves you out too, the problem is wider than one engine, and how ChatGPT picks its sources is a good place to start on the retrieval side.
Keep reading

Engines
ChatGPT Business Hub: What Its Mentions and Clicks Really Count
ChatGPT Business Hub shows a claimed business its mentions, clicks and topics in ChatGPT. How to claim it, what each number counts, and what it leaves out.
, 6 min read

Measurement
Search Console's AI Mode Report: What It Shows and Hides
Search Console's generative AI report counts AI Overviews and AI Mode impressions. What it leaves out, when to opt out, and a monthly workflow.
, 7 min read

Engines
Gemini Rank Tracking: How to Track Your Brand in Gemini
How to track brand mentions in Gemini answers, why there is no GeminiBot to unblock, and how Gemini differs from AI Overviews and AI Mode.
, 5 min read