Discover how AI assistants see your brand
See when ChatGPT, Gemini and the other assistants name, recommend or cite your brand: the answers, question by question.
ReadClaude visibility
When Claude names you, and with which passages.
Claude visibility: TrustInsider asks Anthropic’s models the questions in your panel through the official API, with and without web search, and for each answer keeps mentions, recommendations, cited passages, country, city and model.
Searches per answer
How many times can Claude search before answering?
With search on, Claude decides by itself whether and how many times to search, within a cap set for each answer. Anthropic counts the number of searches and it arrives with the answer: it is an observed fact, not a parameter we choose.
The search is run directly by the model, without going through a code environment: so the answer contains only text, reasoning, searches and results, and every element is read with the same rule.
The settings that go with a measurement question to Claude with search, separated from the facts that are read only in the answer.
| Item | Who decides it | Where to read it |
|---|---|---|
| Search cap | The collection profile, before each send. | In the request that goes to Anthropic. |
| Searches run | Claude, question by question. | In the count Anthropic returns. |
| Country and city | The panel’s market. | In the location sent with the search. |
| Cited pages | Claude, while writing. | In the citations tied to the text. |
Cited passages
From which passages of pages does Claude take its sentences?
When it searches, Claude ties a sentence to the search result it draws it from. TrustInsider keeps the address of each cited result next to the sentence, so you see which information about your brand comes from your site and which from a third-party page.
Only citations that point to a search result count. A name written without a citation is a mention, and a page never cited does not become a source because it appears among the results.
A sample Claude answer with search on, about fictional wineries: the recommendation, the winery’s visits page cited, the competing winery named and a tourist guide cited for opening hours.
QuestionWhich winery should I visit at the weekend for a tasting with lunch?
Claude with search · sample answer about fictional brands
For a tasting with lunch I recommend Tenuta Calvarossa, which offers visits by booking on Saturdays, as stated on the estate’s visits page. Cascina Morleto also opens on Sundays, according to a local tourist guide.
Recommendation
Claude recommends the winery for that request: it counts as a recommendation.
Citation
The sentence about visits is tied to the estate’s page: the brand’s site is a cited source.
Competitor
The competing winery is named as an alternative: its presence is recorded.
Source
A third-party guide cited for opening hours: it enters the external sources with its domain.
Reasoning and text
What remains of Claude’s reasoning in a collection?
Claude’s models can reason before answering. Only the text of the answer enters the measurement: the reasoning is not read and does not count as a mention, while its tokens stay in the receipt, because they are part of the cost.
The parts of a Claude answer with search, split between what enters the visibility reading and what enters only the cost receipt.
| In the visibility reading | In the cost receipt | |
|---|---|---|
| The text | Mentions, recommendations and competitor names in the text. | The output tokens of the whole answer. |
| The reasoning | Never read and never counted as presence. | Its tokens, included in the output. |
| The searches | The cited pages next to the sentences. | The number of searches counted by Anthropic. |
| The service | No effect on the reading of the brand. | The service tier declared for the answer. |
Frequently asked questions
Claude and your visibility: questions and answers.
How a refusal and an interrupted answer are told apart, what changes without search, and why the API’s answers can differ from those of the Claude app.
When is a Claude answer a refusal?
When Anthropic closes it declaring it a refusal. An answer that does not reach its natural end is incomplete. In both cases the outcome stays in the log and the answer does not enter the count of appearances.
What does Claude measure without search?
How the model knows the brand, without consulting pages. If an answer without search still reports searches, it is marked incomplete: the two modes are not mixed.
Does the Claude app answer the same way?
Not necessarily: the app has its own settings and the history of the person using it. TrustInsider reads Anthropic’s official API and says so next to every measure.
Can I see Claude next to the other assistants?
Yes: on the same questions and in the same market, each assistant has its own cell. The cells sit side by side in AI visibility, without being added together.
Related paths
From Claude to the other assistants, question by question.
Claude takes sentences from specific passages of pages. Below you find AI visibility for the other assistants question by question, how the platform works, the published methodology and support on credits and data.
From question to action
Learn how to use TrustInsider: choose the brand and the questions, read the collected answers, update your content and compare the results.
ReadThe published method, with weights and limits
How every TrustInsider measure is built: sampling of AI answers, sources and rights, index calculation and declared limits.
ReadFrequently asked questions on plans, credits and data
Help with access, credits, collections, integrations and data: guides, FAQs, the site assistant and support requests from your workspace.
Read