What Makes ChatGPT Recommend a Therapist? I Checked 200 Answers

I asked ChatGPT to find me a therapist 200 times: ten specialties, ten US cities, every question asked twice. Then I compared the names it gave with therapists from the same Psychology Today search pages whom it didn't name, and checked the same things for both groups.

The short version: solo clinicians ChatGPT named were nearly twice as likely to have a Google Business Profile as solo clinicians it skipped on the same Psychology Today page (51% against 28%), and nearly four in five of the names it gave were business names, not people's names. ChatGPT doesn't seem to weigh therapists against each other. It looks them up, and local business listings are a large part of what it finds.

What I found

Nearly four in five were business names, not people's names. 79% of the distinct names ChatGPT gave were business names: group practices, clinics, and solo practices trading under a brand. Among the therapists I sampled from the same Psychology Today pages, 8% were. If you practice under your own name, you are competing for about one slot in five.

The solo clinicians it did name had a Google listing far more often. Individual clinicians compared with individual clinicians, each counted once:

Solo clinicians	Named by ChatGPT	Same Psychology Today page, not named
Has a Google Business Profile	51% (59 of 115)	28% (102 of 366)
A complete profile (claimed, website, 5+ photos)	17% (19 of 115)	5% (19 of 366)
At least one Google review	28% (32 of 115)	11% (42 of 366)
10 or more Google reviews	8% (9 of 115)	3% (12 of 366)

The ten-review row rests on small numbers (9 and 12 clinicians), so read it as a direction, not a ratio. Counting every name ChatGPT gave, the gap looks wider (82% against 29%), mostly because practices nearly always have a listing: 90% of the business names it gave did.

ChatGPT relays listing data rather than forming an opinion. 65% of the names arrived with a star rating, a review count, a category and a street address: the same fields a Google Business Profile shows. 57 of the 719 website links in the 200 answers still carried tracking tags that practices add to the website link in their Google profile (tags like utm_campaign=gbp). A listing isn't the only way in: nearly half of the solo clinicians it named, 56 of 115, had no Google profile I could find. But therapists without one were far less likely to be named.

Ask twice, get a different list. 64% of questions came back different on the second run, with the two runs minutes apart. In 45% the second run named someone the first didn't; in the other 19% it dropped names without adding any. Only 36% returned the same list twice.

Psychology Today isn't required. 43 of the 115 named solo clinicians had no Psychology Today profile under their name that I could find. The other 72 had a likely match: 16 confirmed, and 56 matching on name and location only. Since every therapist in the comparison group came from Psychology Today, a profile there can't be what separated the named from the skipped.

Two smaller things. It sometimes answers with psychiatrists: 11 of the 541 businesses it named are listed on Google as psychiatrists, mostly for addiction and postpartum depression, and most of those answers mentioned psychiatry or medication. And it names listings nobody manages: 20 of the 541 had a Google profile the business itself had never claimed.

What a solo therapist can do with this

1. Find out whether you can have a Google Business Profile. Google lets an individual practitioner have their own profile if they work in a public-facing role and can be reached at the verified location during stated hours. If you only see clients online and have no office, you don't qualify for an address-based profile, and that makes you far less likely to appear in these answers. Your route is your own website and the directories ChatGPT reads, Psychology Today among them.
2. If you have a profile, finish it. Claim it, choose the category that says what you actually do, add your website and real photos. Named solo clinicians were more than three times as likely to have a complete profile (17% against 5%), and this is the part entirely in your control. My Google Business Profile guide for therapists walks through each field.
3. Treat reviews as the slow part. Named solo clinicians were about two and a half times as likely to have even one Google review (28% against 11%). But the APA, ACA and NASW codes restrict soliciting testimonials from the people you treat, so the review drives other businesses run aren't open to you. What you can and can't do about Google reviews covers where the line sits.
4. Check yourself the way a client would. Ask ChatGPT "Find me a therapist for [what you treat] in [your city]", then ask again. If you aren't named, look at the listings of the people who are.

If you want the wider picture of how AI search finds therapists, this guide covers ChatGPT, Google AI Overviews and the rest. And if your listing is the part that's missing, that is the work behind local SEO for therapists.

What this study can’t tell you. It shows what named therapists have in common, not what causes a mention: a profile with reviews may simply mark an established practice, and adding one doesn’t guarantee you will be named. It is one evening; the lists changed between two runs minutes apart, and how much they move over weeks, this study doesn’t measure. Everyone in the comparison group has a Psychology Today profile, because that is where they came from; the check above shows many named clinicians have one too, but not whether having one helps. Matching names on Google Maps is imperfect: a clinician whose listing sits under a different practice name counts as having no profile, and the same rule applied to both groups.

How I ran it

• The question: "Find me a therapist for {specialty} in {city}, {state}." Asked on its own each time, with no earlier conversation.
• Specialties: anxiety, depression, trauma and PTSD, couples counseling, ADHD, OCD, eating disorders, grief, postpartum depression, addiction.
• Cities: New York, Los Angeles, Chicago, Houston, Phoenix, Philadelphia, Denver, Seattle, Atlanta, Boston.
• Model: gpt-4o-mini (version 2024-07-18) through OpenAI's API, called via DataForSEO, with web search forced on and set to the US. Two runs of each of the 100 questions, 200 answers in all, collected on October 6, 2026 between 22:27 and 22:30 UTC.
• Comparison group: for each question, the first four therapists on the matching Psychology Today search page whom ChatGPT didn't name in either run. 400 therapists in total.
• Same check for both groups: a Google Maps search for the name in that city. A result counted as a match only if its website, phone number or name agreed, within the top five.
• Practice or individual: a name counted as an individual clinician when it was a person's name, with or without credentials ("Jane Doe, LCSW"). Everything else counted as a practice, including a solo therapist who trades under a business name. The same rule applied to both groups. Directories and platforms the answers mentioned, such as Psychology Today and Grow Therapy, were left out of the named group.
• Counting: the same practice can be named in both runs and for several specialties in one city, so each distinct practice or clinician is counted once: 541 named, 396 not named. Counted per mention instead (122 named and 370 skipped solo clinicians), the profile figures are 52% against 28%.
• Psychology Today check: for the 115 named solo clinicians, I searched Psychology Today's own directory by name. A profile counted as confirmed when its phone number or street address matched the clinician's Google profile, or when the clinician's own website linked to it or showed the same phone number. Psychology Today doesn't show a therapist's website address in the profile page, so the website check ran from the clinician's site.
• Clients: I checked every name in the answers against my client list. None of my clients appeared in any answer, and no therapist is named in this post.

Does the ChatGPT app do the same?

The API model isn't what people use in the ChatGPT app. So the same night (23:07 to 23:09 UTC) I ran 20 of the same questions through the ChatGPT app itself, using DataForSEO's ChatGPT scraper: two questions per city, covering each specialty twice. The scraper reported the model as gpt-5-6: the GPT-5.6 family, ChatGPT's current default. It didn't say which variant.

• Every one of the 20 answers opened with a row of local business cards, five in a typical answer, each with a rating and a review count.
• It named mostly different therapists: only 16 of its 116 names also appeared in the API's answers to the same questions.
• The pattern held. Of the 32 solo clinicians the app named, 56% (18) had a Google Business Profile, against 25% (18 of 72) of solo clinicians on the same Psychology Today pages. 34% (11 of 32) had at least one review, against 8% (6 of 72).
• Business names were 72% of its names.
• Psychology Today pages were among its sources in all 20 answers. It reads the directory, yet most of the names it put forward were businesses with Google listings.

Twenty answers is a spot check, not a second study. But a different model, different names, and the same pattern.