Claude is one of the AI models Gumshoe tracks in your Visibility Audit (formerly known as a report). Claude with live web search is available on Pro, as a separate model option you select yourself. This article explains how Claude generates its answers inside Gumshoe, how to turn live search on, and what to expect if you don't.
How does Claude answer questions in a Gumshoe audit?
That depends on which Claude option you have selected.
The standard Claude options answer from foundational (training) knowledge only. They do not look anything up while your audit runs, and they return no source URLs.
The (Web Search) options search the live web before answering. When one of these is selected, every prompt in your audit goes to Claude with Anthropic's own web search tool attached, plus an instruction to search for current information rather than answer from memory. The search happens on Anthropic's side. Gumshoe does not send your prompts to a separate search engine, and no crawler is involved. Claude decides what to look up, reads the results, and answers.
This is different from using Claude yourself in a chat window, where it may search for one question and not the next. Inside a Gumshoe audit the mode is fixed for whichever option you selected, which is what keeps your results comparable from run to run.
How do I turn on Claude's live web search?
On a Pro project, open your project settings, find Model families, and select the Claude (Web Search) options. If you are creating a new project, you can also select them in the final step of setup.
These options are not selected for you. An existing Pro audit keeps the models it already has until you add them yourself, so nothing changes about your audit on its own.
Pro has no limit on how many model families one audit can run, so adding the Web Search options adds to your selection rather than replacing anything.
What is the difference between Claude and Claude (Web Search)?
The standard options reflect what Claude knows from training, the same as before. The (Web Search) options let Claude pull in current information before it answers, so your Visibility Audit captures what Claude says when it is actively looking things up.
The other difference is citations. A (Web Search) answer comes back with the source URLs Claude used, and those are stored with your audit the same way Gumshoe captures citations from Perplexity, Gemini, and Google AI Overviews. The standard Claude options return no sources at all, so they contribute nothing to your Citations or Sources views.
Because both options run the same underlying model, you can select both and compare them side by side: the same model, the same prompt, one answer from training and one from a live lookup.
Is this available on Starter?
No. Claude with live web search is included in Pro only. Starter projects can still run the standard Claude models.
You will see both (Web Search) options in your model list on Starter, greyed out, with the note "This model is not available for this subscription." That is expected. The list shows the full Pro model pool so you can see what a Pro subscription adds.
Does my Pulse Audit use Claude's live web search?
No. Your Pulse Audit runs a fixed set of models that cannot be changed, and Claude with live web search is not among them. Selecting a model in your project settings governs your Visibility Audit.
This is true on both plans, whether your Pulse Audit runs daily on Pro or weekly on Starter.
Does this change how many conversations my audit uses?
Yes. Every model you select runs every prompt in your audit, so adding a (Web Search) option adds one conversation per prompt per audit run. Selecting both Claude (Web Search) options adds two conversations per prompt per run.
Your Audit settings page recalculates the monthly total as you tick models, so you can see the effect before you save.
Will my Visibility Score change?
It may, and not only because Claude is seeing fresher information. Your Visibility Score reflects the mix of models in your audit, so adding a model changes the mix. An audit that adds Claude with live web search is not directly comparable to its own earlier runs.
Treat the first run after the change as a new baseline rather than a month over month comparison.
Can my content still show up in Claude without live search?
Yes. Claude's foundational knowledge refreshes on a training cycle, so strong content and third-party mentions can land you in Claude over time. We have seen client content picked up by Claude within a few months of publishing, which suggests those cycles are moving faster than they used to.
One thing to focus on: keep publishing high-quality content and earning mentions. If you are not running Claude's (Web Search) options, watch Perplexity as your early indicator. Live search models like Perplexity tend to reflect new content first, while foundational-only models follow on their own cycle. If you do have Claude's (Web Search) options selected, you will see those reflect new content faster too.
What should I do next?
If you are on Pro and not yet running the Claude (Web Search) options, add them and compare the two Claude answers side by side on your next run. Seeing the same model answer once from training and once from a live lookup is the clearest picture you can get of whether fresh content is reaching Claude at all.
For the full steps, including what happens when you remove a model, see How do I change the models in an existing visibility audit?