Skip to content
Built for AI agentsConnect yours
Claude logo

Claude

Claude brand tracking: get recommended by Claude

Claude answers a smaller audience than ChatGPT but a disproportionately senior one, and it recommends sparingly. See which Claude answers name a competitor instead of you, the pages it quoted to decide, and whether a change moved it.

Start 7-day free trialAll platforms

Card required, nothing charged for 7 days

  • API with web search
  • Tuesdays and Fridays
  • The same in every market

API with web search

Claude is asked twice a week, through its own API

We ask Claude through its API with web search switched on. Each Claude answer costs several times as much as one from any other engine, so Claude runs on Tuesdays and Fridays while the other five run six days a week, and every Claude figure says how many answers it rests on.

Tuesdays and Fridays.
Two runs a week, never on consecutive days, so one bad day never costs the whole week.
The same answer everywhere.
Claude's API takes no location, so the market your workspace measures does not change what Claude says.
Search, recorded.
Web search is switched on for every question, and each answer records whether Claude used it.

The collection week

Claude's two days in a six-day week

Every Claude answer is a paid call to Anthropic's API plus a fee for each web search it runs, several times what an answer costs from any other engine we track. So Claude is asked on Tuesday and Friday: never two days running, so its two readings come from different days and one bad day never costs the week. The other five are asked Monday to Saturday.

The days of the week each engine is asked your prompts: Claude on Tuesday and Friday, the other five Monday to Saturday, none on Sunday.
EngineMTWTFSS
ChatGPTaskedaskedaskedaskedaskedaskednot asked
Claudenot askedaskednot askednot askedaskednot askednot asked
Geminiaskedaskedaskedaskedaskedaskednot asked
Perplexityaskedaskedaskedaskedaskedaskednot asked
Google AI Overviewsaskedaskedaskedaskedaskedaskednot asked
Google AI Modeaskedaskedaskedaskedaskedaskednot asked

Sunday is the day no engine is asked: it is the quietest day for buying questions, so it is the cheapest reading to lose.

What two days a week changes

  • 32 answers a prompt, 2 of them Claude'sEach prompt you track produces 32 answers a week: 30 from the five engines asked six days, and 2 from Claude. Plan allowances are counted on the same schedule.
  • A count before a percentageUntil Claude's figure rests on five answers in the window, its card reads as a count, such as 3 of 4, rather than as 75%.
  • Alerts wait for a bigger moveA per-engine alert compares one period with the one before and fires only when the change is larger than chance could produce on that many answers. Fewer answers means Claude has to move further before it alerts.
  • Still in every pooled figureClaude's answers count toward your headline, which pools all six engines, and toward the sources behind your answers.
How plans count answers

Answers with no brand

When Claude names nobody, the answer still counts

Claude is more willing than the other engines to answer without recommending anyone. On those prompts no competitor holds the recommendation yet, so the answer is stored in full and counted, and the answer view says no tracked brand was named instead of showing an empty list.

Counted, not dropped.
Only our own collection failures leave the denominator, so an outage never reads as your absence and a brandless answer never hides.
Short source lists.
Claude tends to cite only a few pages, so each one it does cite is worth knowing.
Every word kept.
The whole answer, so you can read how Claude framed the choice even when it named no one.

Citations in two passes

The pages Claude quoted come first in its source list

Beside each answer, Claude's API returns two things: the passage it drew on from each page it quoted, and every result its web searches turned up. We keep both. Quoted pages lead the list in the order Claude cited them, and the pages its searches returned follow, so a page Claude read but did not quote still counts as cited. When Claude recommends a competitor, its quoted pages are the clearest guide to where your brand needs to be.

Quoted first.
The pages behind Claude's own sentences head the list, because those are the ones that shaped what it said.
Read, and still counted.
A page Claude's search returned counts as a citation even when no sentence rests on it, which is what lets a Claude answer read as cited, not named.
One row per page.
A page that was both quoted and returned by the search is listed once, in its quoted place.

Fabrikam alternatives for a remote team

Claude, 02 Oct 2026 06:17

Example data
1

The answer, and the pages it quoted

For a remote team moving off Fabrikam, Contoso is the closest like-for-like swap: the same task and client views, and projects move across with their templates. Northwind is the lighter choice if you mostly need shared task lists and async updates.

  • “Import your Fabrikam projects with their templates, task lists and client views intact.”

    contoso.com/switch/fabrikam

  • “Shared task lists and async check-ins, built for teams spread across time zones.”

    northwindtraders.com/remote-teams

2

What its search returned

  • Switching from Fabrikamcontoso.com/switch/fabrikamalready quoted
  • Fabrikam plans and pricingfabrikam.com/pricingadded
  • Northwind for remote teamsnorthwindtraders.com/remote-teamsalready quoted
3

Citations, as the answer view lists them

  1. #1contoso.comyoursquoted
  2. #2northwindtraders.comquoted
  3. #3fabrikam.comsearched

Per-engine visibility

Claude brand tracking on a smaller weekly sample

Two runs a week means Claude's count grows more slowly than the other engines', so whether a change worked takes longer to show here. Below five answers its card shows the fraction rather than a percentage, because a percentage there would be a guess, and Claude is never averaged with ChatGPT, Gemini, Perplexity or Google's two surfaces.

What it takes to be named by Claude

If something here is unclear, ask us before you sign up rather than after.

Start 7-day free trial

Card required, nothing charged for 7 days

How do I get Claude to recommend my brand?
Look at the pages Claude quoted when it named someone else. Claude cites few sources per answer and quotes the ones behind its own sentences, so those pages are the clearest guide to where your brand needs to be and what it needs to say. The pages its searches returned count too, and the Sources screen shows the sites read beside a competitor and never beside you. After a change, the Tuesday and Friday runs show whether Claude's answers moved, though with two runs a week that takes longer than on the other engines.
Can RankHound track brand mentions in Claude?
Yes. We ask Claude your category prompts every Tuesday and Friday, store each answer verbatim, and record which brands it named and which sources it cited. Collected through the API with web search enabled.
Why is Claude checked less often than the other engines?
Each Claude answer costs several times what the others do, because it goes through Claude's own API with web search rather than a product surface. So Claude runs twice a week and the other five run six days a week. Every Claude figure carries the number of answers behind it, so the smaller sample is never hidden.
Why do Claude's answers vary between runs?
Answers are generated rather than looked up, so wording, brand order and citations vary between runs, and Claude decides for itself whether to search before it answers. With two runs a week its sample builds slowly, so until it rests on five answers Claude's card shows a fraction rather than a percentage.
What makes Claude different from the other assistants?
Claude cites fewer sources per answer and is noticeably more willing to say it does not know. That makes an absence less informative than it looks: on many category prompts it names nobody at all. Those answers still count, because an answer that names no brand is a real observation about your category, and the answer view says so rather than leaving a blank.
Does the market I choose change Claude's answers?
No. Claude's API takes no location, so it answers the same wherever the question comes from. The market your workspace measures applies to ChatGPT, Gemini and Google's two surfaces, not to Claude or Perplexity.
Is Claude blended with the other engines anywhere?
Never into a score. Claude answers through an API twice a week, and averaging it with engines read from a product surface six days a week would produce a number that describes neither. Where the Overview does pool the engines, as your headline does, it says so on the screen.
Does Claude search the web for every answer?
Web search is switched on for every question we ask Claude, but Claude decides whether to use it, and may run up to five searches for one answer. Each answer records whether it searched, in the answer view's Web search row. An answer given without a search still counts, and has no pages behind it to cite.
Which pages count as a Claude citation?
The pages Claude quoted in its answer and the pages its web search returned. Quoted pages come first, in the order Claude cited them, then the rest of the search results, each page listed once. So your page counts as cited when Claude's search returned it, even if no sentence in the answer rests on it.

Related reading

7-day trial

Get recommended by Claude, starting with who it names now

Name the competitors Claude might recommend instead of you, and the prompts to ask it. Every Claude answer is kept in full, with the pages it quoted and the count behind its rate, so you can see where to work and whether it worked.

Start 7-day free trialHow to measure AI visibility

Card required, nothing charged for 7 days.

On every plan

  • Every answer stored in full, with its citations.
  • Every rate shown with the number of checks behind it.
  • Cited-but-not-named reported as its own state.