A family in Scottsdale asks ChatGPT for a pediatric dentist. A retiree in Tampa asks Claude which estate planning attorneys handle Florida probate. A young couple in Denver asks Gemini for a buyer's agent who knows the Highlands neighborhood. A new patient in Boston asks Perplexity for a dermatologist who treats melasma. Four different assistants, four different answers — and four very different reasons one practice gets named while another, sometimes the better-known one, does not.
The four dominant AI assistants don't share a single algorithm. They share a family of techniques (retrieval-augmented generation, training-corpus recall, knowledge graph lookups) but each one weights sources differently. If you want to be recommended consistently, you have to understand what each platform is actually doing under the hood — and then you have to find the common denominator that gets you cited on all of them.
The Four Platforms Are Not Interchangeable
Most professionals assume "AI search" is one thing. It isn't. Here's how the four leading assistants actually decide who to recommend.
ChatGPT: Browsing-Enabled Recall + Bing's Index
When a user asks ChatGPT about a local professional, two things can happen. If browsing is disabled, ChatGPT pulls from what it learned in training — which means stable, well-documented businesses with long histories of mentions get the nod. If browsing is enabled (the default for most users now), ChatGPT performs live retrieval against Bing's index and synthesizes an answer from what it finds.
This has three concrete implications:
- Bing indexing matters. If your site isn't crawled by Bing, you're invisible to live-browsing ChatGPT. Submit your site through Bing Webmaster Tools, not just Google Search Console.
- Recency gets weighted. A dental practice in Austin that published a new "Same-Day Crowns" service page last month is more likely to surface for that query than a competitor whose site hasn't changed in two years.
- Structured data passes through. Schema markup (LocalBusiness, Physician, Attorney, RealEstateAgent) gives ChatGPT clean, parseable facts — hours, specialties, service areas — that frequently appear verbatim in responses.
Claude: The Editorial-Weighted Conservative
Claude is the most conservative of the four. Anthropic has trained Claude to defer to authoritative sources and to hedge when sources conflict. In practice, Claude rarely recommends a business based on a single website or a thin directory profile.
What Claude does surface: editorial mentions. If you've been quoted in a local newspaper, cited in a professional association publication, profiled in a regional business journal, or referenced in an academic context, Claude is far more likely to name you. A personal injury attorney in Philadelphia who has been quoted twice in The Inquirer and is featured in Super Lawyers will outrank a competitor with a flashier website but no editorial footprint.
For healthcare, Claude leans heavily on Healthgrades, Doximity, and ZocDoc. For legal, it's Avvo, Martindale-Hubbell, and Super Lawyers. For real estate, it's Zillow and Realtor.com. These aren't just SEO directories to Claude — they're treated more like credentialing sources.
Gemini: Google Knowledge Graph + Business Profile
Gemini is the only assistant that's deeply integrated with the Google knowledge graph and Google Business Profile. This makes it the most directly influenceable of the four — and the most punishing if your profile is incomplete.
A complete, verified Google Business Profile with:
- Accurate primary and secondary categories
- Full service menu with descriptions
- Recent photos (last 90 days)
- A steady cadence of reviews (and owner responses)
- Updated hours, including holiday hours
- Posts published in the last 30 days
- Q&A entries answered by the owner
…will dominate Gemini's local recommendations for that category in that geography. We've seen orthodontic practices in Charlotte move from invisible to consistently recommended in Gemini within six weeks of bringing their Business Profile to full completeness — without changing anything else.
Perplexity: The Transparent One
Perplexity is the easiest platform to audit because it shows its work. Every recommendation comes with cited sources, hyperlinked. If you want to know why a Brooklyn family law attorney is being recommended, you can scroll down and see exactly which Avvo profile, which firm bio page, and which local news article Perplexity pulled from.
Perplexity's retrieval layer leans heavily on:
- Directory aggregators (Yelp, Healthgrades, Avvo, Zillow)
- Local and regional news outlets
- Wikipedia and Wikidata
- High-authority vertical publications
- The business's own website, if well-structured
Because Perplexity cites in real time, it's the best platform for tracking your AI visibility. If you can see yourself there, you've got a foothold. If you can't, you have a clear list of source gaps to close.
The Common Denominator: Why Some Practices Appear on All Four
When we audit practices that get cited across all four platforms, one pattern keeps repeating. It isn't a flashy website. It isn't paid placement. It isn't even superior services.
It's this: high-authority directory presence with consistent, complete profiles across the major sources each AI's retrieval layer indexes.
A dermatology practice in Minneapolis that gets named by ChatGPT, Claude, Gemini, and Perplexity will almost always have:
- A fully completed Google Business Profile (Gemini's primary source)
- Verified, complete profiles on Healthgrades, Doximity, and ZocDoc (Claude's preferred sources, Perplexity's retrieval base)
- A modern website with LocalBusiness and Physician schema (ChatGPT's structured-data path)
- At least a handful of editorial mentions — local news, association publications, alumni magazines (Claude's authority weighting)
- Consistent NAP (name, address, phone) across every directory it appears on
The boring truth: cross-platform AI visibility is won by entity consistency, not platform-specific tricks. This is what we mean when we talk about your AI Visibility Score — it's a measurement of how recognizable and consistently described your practice is across the source ecosystem that all four assistants draw from.
Understanding RAG: Why "Being in the Sources" Is the Whole Game
All four platforms (Gemini and Perplexity always, ChatGPT in browsing mode, Claude when given tool access) use some variant of retrieval-augmented generation. The mechanics matter:
- The user asks a question.
- The system converts the question into a search query.
- It retrieves a set of candidate documents from its indexed source pool.
- It feeds those documents to the language model as context.
- The model synthesizes an answer using only what's in that context.
If you're not in the indexed source pool — or you're in it but with thin, inconsistent, or outdated information — you can't be retrieved. And if you can't be retrieved, you can't be cited, regardless of how good your services actually are.
This is the single biggest mental shift professionals need to make about AI search. It is fundamentally different from traditional Google search, where ranking on page one was the goal. In AI search, the goal is to exist as a clean, complete entity in the sources each assistant indexes. Ranking is replaced by recall.
A Platform-Specific Playbook (That's Really One Playbook)
Here's the priority order we recommend for professionals starting from zero:
- Week 1–2: Audit and complete your Google Business Profile to 100%. This unlocks Gemini fastest and feeds back into Perplexity and ChatGPT browsing results indirectly.
- Week 2–4: Claim and fully complete every major vertical directory profile (Healthgrades and ZocDoc for medical, Avvo and Super Lawyers for legal, Zillow and Realtor.com for real estate). Match every detail to your Business Profile.
- Week 4–6: Add LocalBusiness and profession-specific schema to your website. Submit your site to Bing Webmaster Tools. Publish two to three substantive service or condition pages.
- Week 6–12: Pursue editorial mentions — a quote in a local outlet, a piece in a professional association newsletter, a guest column in a regional business publication. These are slow but compound especially well for Claude.
This is essentially the same playbook we walk through in detail in How to Rank in ChatGPT, and it works because every platform ultimately drinks from overlapping wells.
Where to Start
The fastest way to know which platforms are already citing you, which aren't, and exactly which sources are missing is to run a citation scan. You'll see, side by side, how ChatGPT, Claude, Gemini, and Perplexity currently describe (or fail to describe) your practice — and which gaps to close first.
Ready to find out where you stand across all four assistants? Run your free AI visibility scan and get a complete picture of your citation ecosystem in under two minutes.