Claude optimization for medspas

1 · What Claude does when a patient asks about a medspa

We measure Claude against its default model with web search enabled. A patient asks Claude a local medspa question, and it can run a search, weigh what it retrieves, and compose an answer that names a short list of specific clinics: the same shortlist compression the other four surfaces show. A real patient today gets a direct, named recommendation, not a page of links.

Across the ten South Bay practices in our first study, Claude named the practice in about 23% of answers, our second-highest surface after ChatGPT. Claude also tends toward caution: when the sources it retrieves disagree with each other, it is often more conservative about naming a brand at all rather than naming one anyway.

2 · How Claude picks who to recommend

Anthropic documents three named crawlers. Claude-SearchBot indexes web content to improve search result quality, the one that governs whether Claude’s search feature can cite your site at all. Claude-User performs a live fetch at the moment a specific patient’s question needs your page. Anthropic also runs ClaudeBot, generally described as collecting content for model training.

If Claude cannot fetch your pages, whether through a robots.txt disallow on the citation-gating bots or a firewall turning the crawler away, it cannot cite what it cannot read. Beyond crawl access, the same pattern holds as on every surface we audit: which sources are found (your own site, local directories, review aggregators), whether those sources agree with each other and are cited elsewhere, and whether your pages state the treatment explicitly enough to match a specific query.

3 · Why most medspas do not appear

In our first study of ten South Bay practices, Claude named the typical clinic in only about 23% of answers, meaning it was silent more than three times out of four. The same failure modes recur across every surface we measure: no treatment-specific page a search can resolve to the query, absence from the third-party directories that cover the metro, name-address-phone details that disagree across the clinic’s own site and its listings, and no structured data confirming what the page’s prose claims.

It is a small, early sample (ten practices, directional, not the last word), and Claude’s retrieval layer changes continuously, so a single snapshot is not a verdict.

4 · What we measure on Claude

Every audit runs the same 36 patient query types (expanded into 162 query-runner combinations across all five surfaces) against Claude’s default model with web search enabled. For each query we record three things:

• Mention. Was the medspa named in Claude’s answer?
• Citation. Was the medspa’s URL or property cited as a source?
• Rank position. Ordinal position among named competitors in the answer.

These roll up into a Claude-specific visibility score (0–100) that sits alongside the scores for ChatGPT, Gemini, Google AI Overviews, and Perplexity in the combined audit. Methodology detail, including dedup rules, is on the methodology page.

5 · The fix patterns that move Claude rank

We do not chase surface-specific tricks. The audit ships the same fixes ordered by projected lift, not by activity volume:

• Confirm Claude-SearchBot and Claude-User are not disallowed in robots.txt and are not being turned away by CDN bot management. Whether to also allow ClaudeBot, the training crawler, is the owner’s call about model training, and our audit does not count it either way. That is not the same trade-off as Google-Extended on the Gemini page: blocking that one does cost you Gemini.
• Treatment pages with explicit, retrievable copy: one page per major procedure, with the procedure name, location qualifier, and price band stated plainly.
• LocalBusiness, Service, and FAQPage structured data on the site, kept consistent with the prose.
• Name-address-phone consistency across Google Business Profile, Yelp, and any third-party directory that covers the metro.
• At least one third-party citation per major treatment line, since Claude weighs whether the sources it finds already agree with each other.

For the deeper how-to view written for owners doing this without an agency, see how to rank your medspa on ChatGPT (the same fixes carry over to Claude).

6 · Claude vs the other four surfaces

Claude is one of five AI surfaces we audit. The other four have different ranking logic:

• ChatGPT (OpenAI) is currently the largest AI assistant by user volume and issues its own retrieval pass against the live web for local queries.
• Gemini (Google) sits on top of Google’s knowledge graph and local business data, a different source mix from the others.
• Google AI Overviews shares some retrieval surface with Gemini but runs its own citation pipeline off the main Googlebot index; treating the two as one would smooth over real visibility gaps.
• Perplexity (Sonar) is a web-search answer engine that leans heavily on Reddit and other UGC for local queries, a different citation profile again.

A medspa can rank well on Claude and poorly on Gemini, or the reverse. The audit measures ChatGPT, Claude, Gemini, Google AI Overviews, and Perplexity together, because a single surface gives a badly incomplete picture.

Frequently asked

• Does Claude actually recommend specific medspas to patients?
  Yes, in the same way the other four AI surfaces do. Claude’s default model answers with web search enabled, and across the ten South Bay practices in our first study it named the practice in about 23% of answers, our second-highest surface after ChatGPT. Claude also tends toward caution: when the sources it retrieves disagree with each other, it is often more conservative about naming a brand at all.
• What actually controls whether Claude can cite my site?
  Anthropic names three crawlers: ClaudeBot, Claude-SearchBot and Claude-User. Claude-SearchBot indexes web content to improve search result quality, which is what lets Claude’s search feature find your site. Claude-User is the live fetch Claude runs at the moment a specific patient’s question needs your page. ClaudeBot is the training crawler: Anthropic’s documentation describes disallowing it as excluding your site from future model training, and does not tie it to citations. So our audit flags Claude-SearchBot or Claude-User being disallowed in robots.txt or challenged by a firewall, since Claude cannot cite a page it cannot read, and treats allowing ClaudeBot as your own decision about training.
• Is the fix different for Claude than for the other AI surfaces?
  No. The audit ships the same fixes ordered by projected lift across all five surfaces: retrievable treatment pages, consistent structured data, name-address-phone consistency, reviews that name treatments, and third-party citations. What differs by surface is which crawlers need to reach your site, not the underlying content and citation work.
• Why doesn’t my Google Business Profile show up in Claude?
  A Google Business Profile is a Google record. It says nothing about you anywhere else, so if it is the only place your details appear, there is nothing outside Google corroborating them to Claude or any other surface. We have not measured how any assistant weighs one source against another, and anyone quoting you a formula for it is guessing. What we can say is that a listing with no third-party citations and no on-site structured data leaves Claude, and anything reading the page, with a single unverifiable source.