What gets a therapist named by ChatGPT and other AI assistants
The study at a glance
Answers
8,986
600 questions put to ChatGPT, Google AI Mode, Gemini, Perplexity and Claude.
Cities
50
40 US and 10 Canadian metro areas, across 12 specialties and question types.
Practices compared
4,786
The Google Maps top 20 for each city and specialty.
What gets a therapist named by AI assistants?
Quick Answer
In our November 2026 study of 8,986 AI answers across 50 US and Canadian cities, practices on Google's first page for a specialty search had about 3.2 times the odds of being named. Naming the specialty on the practice's homepage went with about 2.5 times the odds. Psychology Today was the most cited source on four of five assistants.
The results in one table
| What we measured | Result |
|---|---|
| Answers that named at least one local clinician or practice | 95% to 100%, depending on the assistant |
| Answers that cited Psychology Today | ChatGPT 66%, Google AI Mode 34%, Gemini 82%, Perplexity 99%, Claude 73% |
| Clinicians named by ChatGPT that came with a Psychology Today profile link | 30% |
| Practice is on Google’s first page for the specialty search | 3.19 times the odds of being named (2.74 to 3.70) |
| Specialty is named on the homepage | 2.49 times the odds (2.20 to 2.82) |
| Homepage links to a page about the specialty | 2.01 times the odds (1.75 to 2.31) |
| Practice is in the Google Maps top 3 | 1.91 times the odds (1.70 to 2.15) |
| Fees shown, insurance mentioned or "accepting new clients" on the homepage | No detectable difference |
| Same question asked twice: providers named in both answers | ChatGPT 18%, Google AI Mode 15% |
| Named providers that were also in the Google Maps top 20 | ChatGPT 20%, Google AI Mode 28%, the other three 6% to 16% |
Therapist AI Visibility Study, Reframe Practice, November 6, 2026. Odds figures come from one model with all factors entered together, pooled across the five assistants. The ranges in brackets are 95% confidence intervals.
How we ran it
We could not find an independent study of how AI assistants pick therapists, so we ran our own.
| Date | November 6, 2026. One day. |
|---|---|
| Questions | 600. That is 50 metro areas (40 in the US and 10 in Canada, across large, mid-size and small markets) times 12 questions: anxiety, couples, EMDR for trauma, adult ADHD, teens, postpartum, OCD and ERP, grief, LGBTQ affirming, men, insurance or benefits, and online therapy. Each was written the way a person looking for a therapist would ask it. |
| Assistants | ChatGPT search and Google AI Mode, five runs of every question each. Gemini and Perplexity, two runs each. Claude, one run. That is 9,000 requests. 14 failed, which left 8,986 answers. |
| Who was named | Every provider or resource in every answer was sorted into fixed categories (individual clinician, practice, directory, national platform and so on). About 78,000 mentions in all. |
| Comparison group | The Google Maps top 20 for each city and specialty: 4,786 practices. For each one we recorded its Google rank for the same search, read its homepage the way an AI crawler does (plain HTML, no JavaScript; 87% of sites could be read) and searched its name for directory listings and other mentions. |
| Matching | AI mentions were matched to Maps practices by website domain or name. A hand check of 60 random matches found about 95% correct. |
| Analysis | How often practices were named, by factor, and a logistic regression with all the factors entered together and errors clustered by city. |
AI assistants name specific therapists
Between 95% and 100% of answers named at least one local clinician or practice. Claude declined to name anyone in 2.7% of answers. Every other assistant declined in under 1%.
ChatGPT named about 6 local providers per answer. Google AI Mode named about 3.6, Perplexity about 6, Claude about 6.7 and Gemini about 14.
Most of them were not the practices on the map. Only 20% of the providers ChatGPT named were in the Google Maps top 20 for the same search. For Google AI Mode it was 28%, and for the other three it was 6% to 16%.
Psychology Today is the source they lean on most
Psychology Today was cited in 66% of ChatGPT answers, 82% of Gemini answers, 99% of Perplexity answers and 73% of Claude answers. Google AI Mode cited it in 34%.
It also made up 30% of everything ChatGPT's own background searches returned. The next source was 4%. Of the individual clinicians ChatGPT named, 30% came with a link to their Psychology Today profile.
What went with being named
We compared the Maps practices the assistants named with the ones they skipped. These are the factors that went with being named, pooled across all five assistants.
| Factor | Odds of being named | 95% range |
|---|---|---|
| On Google’s first page for "[specialty] therapist [city]" | 3.19 times | 2.74 to 3.70 |
| Specialty named on the homepage | 2.49 times | 2.20 to 2.82 |
| Homepage links to a page about the specialty | 2.01 times | 1.75 to 2.31 |
| In the Google Maps top 3 | 1.91 times | 1.70 to 2.15 |
| Google reviews, for each roughly threefold increase | 1.18 times | 1.14 to 1.22 |
| Mentions on other websites, for each roughly threefold increase | 1.10 times | 1.01 to 1.20 |
The first four held on each of the five assistants on its own.
Most practices are not on Google's first page for their specialty and are not in the Maps top 3. In that group, practices whose homepage linked to a page about the specialty were named in 14.2% of cases. Practices without one were named in 3.8%. A case is one practice, one question and one assistant.
This is a correlation. A practice with specialty pages may also put more into its website in general, and that could account for part of the difference.
What showed no effect
Some things we expected to matter showed no detectable difference in whether a practice was named. Each range below includes 1, which means no difference.
| Factor | Odds of being named | 95% range |
|---|---|---|
| Fees shown on the homepage | 0.93 times | 0.75 to 1.15 |
| Insurance mentioned on the homepage | 1.06 times | 0.92 to 1.23 |
| "Accepting new clients" on the homepage | 0.91 times | 0.72 to 1.16 |
| Google star rating | 0.99 times | 0.87 to 1.12 |
| Homepage length | 0.98 times | 0.92 to 1.06 |
| Group practice compared with solo | 1.03 times | 0.87 to 1.21 |
Fees and insurance still matter to the person reading your site. Keep them there for that reader.
The answers change from one run to the next
We asked ChatGPT and Google AI Mode every question five times. Two ChatGPT answers to the same question shared only 18% of the providers they named. For Google AI Mode it was 15%.
ChatGPT named about 6 providers per answer and about 19 different providers across the five runs. Of the Maps practices ChatGPT named at least once, 15% were named in all five runs and 37% were named in only one.
The assistants also disagree with each other. Pairs of assistants shared between 19% and 45% of the Maps practices they named for the same question.
One search where your practice is missing tells you very little. You have to ask several times and count.
What this does not show
- It shows what went with being named. It does not show cause. A practice with specialty pages or a first-page Google ranking may also put more into its website and marketing in general.
- It is one day, November 6, 2026. AI assistants change their sources often.
- Website factors were read from the homepage only. A specialty page that the homepage does not link to was counted as missing.
- ChatGPT answers came from its consumer search and Google AI Mode answers from Google’s results page. Gemini, Perplexity and Claude were queried through their developer APIs with web search on, which can differ from their apps.
- The comparison group is the Google Maps top 20 for one search per question. Practices outside it are counted only in the answer-level numbers.
- Matching AI mentions to practices was about 95% correct on a hand check.
- We did not read Psychology Today profiles. We counted how often the assistants cited the site.
- All questions were in English.
- We did not measure whether being named led to inquiries or clients.
No one can promise a practice a place in an AI answer, and nothing on this page does.
What you can do with it
None of these guarantees that an AI assistant names your practice. They are the places the data points to.
- 1
Name each specialty you want more of on your homepage, and give it its own page.
Link that page from the homepage. Practices that did both were named more often on all five assistants.
- 2
Work on your Google ranking for "[specialty] therapist [city]".
This was the strongest factor in the study. The assistants lean on ordinary search results.
- 3
Keep each clinician’s Psychology Today profile complete and specific about specialties.
It was the most cited source on four of the five assistants.
- 4
If you see clients at an office, keep your Google Business Profile accurate.
Being in the Maps top 3 went with being named, most of all on Google AI Mode. Reviews did too, but we do not suggest asking clients for them. Many regulators and ethics codes restrict that.
- 5
Keep your fees and insurance details on your site.
They did not go with being named, and the people reading your site still need them.
- 6
Check more than once.
Ask the same question at least five times and count how many answers name you. One search where you are missing tells you very little.
If you want to see where your own practice stands, the free Practice Checkup covers your Google presence, Psychology Today profile and website.
FAQ
Do AI assistants recommend specific therapists?
+
Yes. In our November 2026 study, 95% to 100% of answers from ChatGPT, Google AI Mode, Gemini, Perplexity and Claude named at least one local clinician or practice when asked a therapy-seeker question for a specific city. Refusals were rare. Claude declined to name anyone in 2.7% of answers and the others in under 1%.
Does Psychology Today matter for AI search?
+
Psychology Today was the source the assistants cited most in our study. It was cited in 66% of ChatGPT answers, 82% of Gemini answers, 99% of Perplexity answers, 73% of Claude answers and 34% of Google AI Mode answers. Of the individual clinicians ChatGPT named, 30% came with a link to a Psychology Today profile.
Does showing fees or insurance help a practice get named by AI?
+
It did not in our data. Fees, insurance and "accepting new clients" wording on the homepage showed no detectable difference in whether a practice was named. They still matter to a person deciding whether to contact you, so keep them on your site.
How can I check whether ChatGPT names my practice?
+
Ask the question a client would ask, with your specialty and city, at least five times, and count how many answers name you. In our study two ChatGPT answers to the same question shared only 18% of the providers they named, so a single search tells you very little.
Does this study prove what makes AI name a practice?
+
No. It is a one-day snapshot from November 6, 2026, and it shows what went with being named. That is a correlation. A practice with specialty pages may also put more into its website in general. Showing cause would take controlled tests, where one thing is changed and the result is measured before and after.
Citing this study
Reframe Practice, Therapist AI Visibility Study, run November 6, 2026. reframepractice.com/guides/therapist-ai-visibility-study. By Jesse, Registered Psychotherapist.
Related: our earlier 10-city ChatGPT test, what AI can tell about your practice and how clients find therapists.
