Quick Answer
Yes. When asked to recommend a therapist in a specific city, ChatGPT with web browsing lists 3 to 5 named providers with explanations. The recommended therapists consistently share certain traits: strong websites with specialty pages, Google reviews, and presence across multiple directories. Therapists with only a Psychology Today profile are rarely mentioned.
The experiment
The setup was simple. I asked ChatGPT (GPT-4, web browsing enabled) the same prompt in 10 different cities:
I chose 10 cities of varying sizes across the US and Canada to see if patterns hold regardless of market:
Methodology notes
Each search was conducted in a fresh ChatGPT conversation (no prior context)
Web browsing was enabled (the default for ChatGPT Plus and free users with search)
I recorded the first response without follow-up prompts
I then looked up each recommended provider to document their web presence
Searches were run over two days in February 2026
Important caveat: ChatGPT's recommendations change over time and vary based on conversation context. This is a snapshot, not a definitive ranking. Your results will differ if you run the same experiment tomorrow. The value is in the patterns, not the specific names.
What ChatGPT recommended in each city
Quick Answer
ChatGPT typically recommends 6 to 15 therapists or practices per city, organized into categories. It pulls from practice websites, Google Business Profiles, therapist directories like GoodTherapy and Zencare, and review sites. Practices with the word 'anxiety' in their name or domain appear in almost every city's results.
Across all 10 cities, ChatGPT recommended over 90 distinct providers. It gave far more recommendations than expected. Here is what it returned in each city:
| City | Providers | "Anxiety" in name | Had own website | Directories cited |
|---|---|---|---|---|
| Austin | 6 | 2 (ATCA, Central TX Anxiety) | 6/6 | Zencare, PT |
| Denver | 13 | 0 | 10/13 | Westside Behavioral |
| Toronto | 12 | 0 | 10/12 | Theralist, OntarioTherapists |
| Portland | 9 | 3 (Portland Anxiety Clinic, Portland OCD & Anxiety, Portland Anxiety Counseling) | 8/9 | GoodTherapy, Zencare, SonderMind |
| Nashville | 11 | 2 (Nashville OCD & Anxiety, Nashville Anxiety & Performance) | 9/11 | GoodTherapy, Sofia Health, PT |
| Chicago | 8 | 1 (Calm Anxiety CBT Clinic) | 8/8 | Charlie Health |
| Boise | 16 | 1 (K-Counseling & Anxiety Treatment) | 12/16 | GoodTherapy, Recovery Ways |
| Atlanta | 4 | 1 (Anxiety Specialists of Atlanta) | 4/4 | GoodTherapy, SonderMind, PT |
| San Diego | 8 | 0 | 7/8 | GoodTherapy, Grow Therapy |
| Raleigh | 7 | 1 (Light For Your Path, Anxiety & Mental Health) | 7/7 | PT, Nourish Carolina |
Practice names are included because they are publicly visible businesses. This is observational, not an endorsement of any specific provider.
What the recommendations looked like
In every city, ChatGPT organized its response into structured categories. Not a flat list. It separated "Highly Rated" from "Specialized Anxiety-Focused" from "Additional Options." For each recommended therapist, it included:
The practice or therapist name, bolded, with a short description
Why it was recommending them specifically for anxiety
Direct links to their website with ?utm_source=chatgpt.com tracking appended
References to the directories it pulled from (GoodTherapy, Zencare, PT, etc.)
A "Tips for Choosing a Therapist" section at the end of every response
An offer to narrow by insurance, telehealth, or therapy style
Two details stood out. First, ChatGPT appended ?utm_source=chatgpt.com to every link it provided. This means if you check your website analytics, you can see exactly how much traffic ChatGPT is sending you right now. Most therapists have never checked this.
Second, ChatGPT treated this as a conversation, not a search. Every response ended with: "Would you like me to narrow these down based on your insurance, budget, or whether you prefer in-person vs. virtual therapy?" It is actively trying to match clients to therapists, not just listing options.
The shift: A Google result says "here are your options." ChatGPT says "here is who I recommend, and here is why. Want me to help you narrow it down?" That is a fundamentally different discovery model. The user does not browse. They follow the recommendation.
The numbers behind this shift
This is not a niche trend. The data on AI search adoption is striking:
800M
weekly active ChatGPT users (late 2025)
Source: First Page Sage
527%
year-over-year growth in AI-referred website visits
Source: Authoritas / Search Engine Land
48.7%
of LLM users with mental health challenges use them for support
Source: Sentio University Survey
5x
AI search converts at 14.2% vs. Google's 2.8%
Source: Superprompt (2025 data)
That last number matters most. AI search traffic does not just visit. It converts at 5 times the rate of Google traffic. People who find you through ChatGPT are more likely to become clients because they already received a personalized recommendation with reasoning, not a page of 10 blue links.
Meanwhile, Psychology Today referrals have dropped 77 to 94% for many therapists. One practitioner's PT profile views went from 32,000 in 2020 to 2,600 in 2025. OpenAI is even building a formal therapist referral network within ChatGPT, with credentialing, matching algorithms, and handoff protocols. This is not slowing down.
The 7 things recommended therapists had in common
Quick Answer
Across 90+ recommended providers in 10 cities, the strongest predictors were: having the searched condition in the practice name or domain, a dedicated website with specialty pages, presence on directories ChatGPT reads (GoodTherapy, Zencare, Psychology Today), dedicated /anxiety or /anxiety-therapy URLs, and evidence-based approaches like CBT explicitly mentioned on their site.
After documenting every recommended provider across all 10 cities and reviewing their online presence, clear patterns emerged. These align with larger studies on what drives AI recommendations.
According to research by Onely, the weight of AI recommendation factors breaks down roughly as: being mentioned across reputable sites (41%), reviews and ratings (16%), specific client examples (14%), and brand mentions elsewhere on the web (11%). A separate Ahrefs study of 75,000 brands found that web mentions correlate 3x more strongly with AI visibility than backlinks do.
Here are the seven specific patterns from our experiment:
The practices with "anxiety" in their name dominated
This was the single most striking pattern. In 8 out of 10 cities, a practice with "anxiety" literally in its name appeared in the top recommendations: Anxiety Treatment Center of Austin, Portland Anxiety Clinic, Calm Anxiety CBT Therapy Clinic (Chicago), Nashville Anxiety and Performance Counseling, Anxiety Specialists of Atlanta.
The keyword domain effect was equally visible. The providers ChatGPT linked to directly had URLs like anxietyaustin.com, nashvilleanxiety.com, anxietytherapistchicago.com, and portlandanxietycounseling.com. Having the searched condition in your domain name appears to be a strong AI signal.
This does not mean you need to rename your practice. But it does mean that having a dedicated page with a clear URL like yourpractice.com/anxiety-therapy is likely more impactful than you think.
They all had a real website
Over 85% of recommended providers had their own website. Not a Wix one-pager. Not just a Psychology Today profile. A real site with multiple pages, content about their approach, and clear contact information.
The exceptions were a handful of providers that ChatGPT found through directory aggregation rather than direct websites. But the providers it featured most prominently and linked to directly all had their own sites.
They had dedicated specialty pages
The majority of recommended providers had a specific page on their website about anxiety. Not a bullet point in a list of specialties. A full page describing their approach to anxiety treatment, who they help, and what therapy looks like. Many had URLs like /anxiety-therapy, /therapy-for-anxiety, or /specialties/anxiety.
This makes sense. ChatGPT is looking for content that matches the query. A page titled "Anxiety Therapy in Austin" with 500 words of relevant content is much easier for AI to match than a generic "About Me" page that mentions anxiety in a list of 15 specialties.
ChatGPT read specific directories (not just Google)
This was a surprise. ChatGPT did not just search Google and return the top results. It pulled from specific therapist directories and cited them by name. The directories it referenced most frequently across all 10 cities:
GoodTherapy (appeared in 6 out of 10 cities)
Psychology Today (appeared in 5 out of 10)
Zencare (appeared in 3 out of 10)
SonderMind (appeared in 3 out of 10)
Grow Therapy (appeared in 2 out of 10)
Local/regional directories: Theralist (Toronto), OntarioTherapists, Nourish Carolina, Sofia Health, Recovery Ways
If you are listed on GoodTherapy and Psychology Today but not on Zencare or SonderMind, you are missing surfaces where ChatGPT is looking. Being on multiple directories does not just help with traditional SEO. It gives AI more sources to cross-reference.
They mentioned evidence-based approaches by name
ChatGPT consistently highlighted providers who named their modality. "Evidence-based approaches like CBT" appeared in the response for every single city. Practices that explicitly mentioned CBT, ACT, EMDR, DBT, or mindfulness-based approaches on their websites were described in more detail than those that did not.
This makes sense. When a client asks for help with anxiety, ChatGPT is looking for signals of credibility. Naming a specific, researched approach ("We use CBT and exposure therapy for anxiety") is a stronger signal than "I help people with anxiety."
Group practices appeared more than solo therapists
In 9 out of 10 cities, the majority of recommendations were group practices or counseling centers, not solo practitioners. The solo therapists who did appear (like Julie Hsu in Austin or Sara Morgan in Austin) all had something the groups had: a well-structured website with dedicated specialty content.
Group practices have a natural advantage. They have more content, more reviews, and more directory listings simply because they have more clinicians. But solo practitioners can level the field by building a site with the same depth of content on their specialties.
Smaller cities got more recommendations, not fewer
Boise (population ~230,000) got 16 recommendations. Atlanta (population ~500,000) got 4. This was counterintuitive. The explanation: in smaller markets with fewer therapists, ChatGPT had to search harder and ended up listing more options. In larger markets with more competition, it was more selective.
If you practice in a smaller city, you have a genuine advantage. There is less competition for AI visibility, and ChatGPT is more likely to find and recommend you even with a modest web presence.
The summary: The recommended therapists were not doing anything exotic. They had a website with dedicated specialty pages, they were on the right directories, and they named their modality. Practices with the condition in their name had an outsized advantage. The bar is not high. It is just that most therapists have not reached it.
Who ChatGPT ignores
Equally telling is who did not get recommended. In a city like Chicago with over 10,000 licensed therapists, ChatGPT mentioned 8. In Atlanta, just 4. Here are the profiles it consistently skipped:
Therapists with only a Psychology Today profile
If your entire online presence is a PT profile, ChatGPT has very little to work with. PT profiles are structured but generic. They all look the same to an AI. There is not enough unique, parseable content for a confident recommendation.
One-page websites with no substantive content
A Squarespace template with your photo, a paragraph about you, and a contact form is not enough content for AI to understand your practice. There is nothing to extract, nothing to match against a query.
Therapists with no Google reviews
Zero Google reviews means zero social proof for the AI. Even 3 to 5 honest reviews give ChatGPT something to reference when explaining why it recommends you.
Therapists who list 15+ specialties with no depth on any
A profile that says "I specialize in anxiety, depression, trauma, PTSD, OCD, ADHD, couples, families, teens, grief, anger, self-esteem, life transitions, and stress" does not signal expertise in anything. ChatGPT reads that as "generalist" and looks for someone more clearly specialized.
Therapists on platforms like Rula or Alma without individual presence
Platform therapists who appear only as a profile within a larger company directory are rarely recommended individually. The platform itself might get mentioned, but the individual therapist does not.
The uncomfortable truth: In a city with 10,000 licensed therapists, ChatGPT recommends 8. That is 0.08%. The other 9,992 do not exist as far as AI search is concerned. This is not fair, and it does not correlate with clinical competence. But it is the reality of how an increasing number of clients are finding therapists.
Free: ChatGPT Visibility Checklist
A 12-point checklist to evaluate whether your practice is visible to AI search tools. Covers website structure, review presence, directory listings, and content optimization. Takes 20 minutes.
Free download. No spam. Unsubscribe anytime.
What you can do about it (this weekend)
Based on the patterns from this experiment, here are the highest-impact actions you can take. They are listed in order of effort, starting with what you can do today.
Run the experiment on yourself (15 minutes)
Open ChatGPT and ask: "Recommend a therapist for [your main specialty] in [your city]." See if you appear. Then ask about your specific niche. Then search for a competitor's specialty. Take notes on who gets recommended and what they have that you do not.
This alone will tell you exactly where you stand. Most therapists have never done this.
Create specialty pages on your website (2 to 4 hours)
For each condition you treat, create a dedicated page. Not a bullet point. A full page with a clear heading ("Anxiety Therapy in [Your City]"), 300 to 500 words about your approach, who you typically help, what the first session looks like, and a clear way to contact you.
Write it in your voice. Describe what you actually do with anxious clients. The specificity is what makes AI (and humans) trust the content.
One more thing: keep your pages updated. Content updated within the last 30 days gets 3.2 times more AI citations than stale content. Even small updates (adding a recent FAQ, updating your availability) signal freshness.
Add FAQ sections to your site (1 to 2 hours)
FAQ sections are the easiest content for AI to parse. Write 5 to 8 questions your clients actually ask, and answer them in 2 to 3 sentences each. "How long does therapy take?" "Do you accept insurance?" "What is your approach to anxiety?" "How do I know if therapy is right for me?"
Put FAQ sections on your homepage, your specialty pages, and your about page. This structured Q&A format is exactly what AI tools look for when generating recommendations. According to Mentionlytics, 61% of pages cited by ChatGPT include rich schema markup (compared to only 25% of Google's top results). Sites with clear heading hierarchies and bullet points are 40% more likely to be cited.
Ask for Google reviews (ongoing)
If you have fewer than 5 Google reviews, this should be a priority. At an appropriate point in treatment (usually termination or a natural check-in), a simple "If you found our work together helpful, a Google review helps other people find this kind of help" is sufficient.
Aim for 10 reviews as a first milestone. That puts you ahead of the majority of therapists in most cities.
Audit your directory listings for consistency (30 minutes)
Google your practice name and check every listing that appears. Is the phone number correct? The address? Your specialties? Make sure your name, address, and phone number are identical everywhere. Update any outdated listings.
If you want to take this further, we have a more detailed guide on how clients are finding therapists in 2026 that covers the full 5-step visibility audit.
Limitations of this experiment
This experiment has real limitations, and it would be irresponsible not to name them:
ChatGPT results are not stable.
Running the same prompt twice can produce different recommendations. The patterns are consistent, but the specific providers change. Do not treat any single result as definitive.
Web presence does not equal clinical competence.
The best therapist in your city might have no website and zero reviews. Being recommended by ChatGPT does not mean someone is a good therapist. It means they are visible to AI. These are different things.
This is a single query type.
I only tested "anxiety therapist in [city]." Results for trauma, couples therapy, child therapy, or other specialties may show different patterns. Anxiety is one of the most commonly searched conditions, but it does not represent the full picture.
Google rankings and ChatGPT recommendations are not the same thing.
A Chatoptic study of 1,000 queries found only 61 to 62% overlap between what Google ranks and what ChatGPT recommends. Being #1 on Google does not predict being #1 in ChatGPT (rank order correlation was near zero at 0.034). About 77% of AI optimization comes from strong traditional SEO, but the remaining 23% requires different strategies.
This could change tomorrow.
OpenAI updates ChatGPT's search capabilities regularly and is actively building a formal therapist referral network with credentialing and matching algorithms. The factors that influence recommendations today may shift as the technology evolves. What will not change: the trend toward AI-mediated discovery.
The ethical question: Should AI be recommending therapists based on web presence? That is a legitimate concern. Web presence is not a proxy for clinical skill. But the same critique applies to Psychology Today rankings, Google search results, and even word-of-mouth referrals. No discovery mechanism evaluates clinical competence directly. The question is not whether AI search is perfect. The question is whether you want to be visible in the channel where a growing number of clients are looking.
Frequently asked questions
Does ChatGPT recommend specific therapists by name?
+
Yes. When web browsing is enabled, ChatGPT recommends specific therapists and practices by name, typically listing 3 to 5 providers with explanations for each recommendation. Without browsing, it provides general guidance on how to find a therapist.
What makes ChatGPT recommend one therapist over another?
+
Based on this experiment and published research (Onely, 2025), the key factors are: being mentioned across multiple reputable sites (41% weight), having Google reviews (16%), brand mentions on the web (11%), and having a well-structured website with clear specialty pages and FAQ sections.
How many people are using ChatGPT to find therapists?
+
ChatGPT has over 800 million weekly active users. While there is no specific data on how many use it to find therapists, therapists are increasingly reporting new clients who found them through ChatGPT. One psychologist reported that ChatGPT referrals now exceed Psychology Today referrals.
Can I pay to get recommended by ChatGPT?
+
No. As of early 2026, there is no paid placement within ChatGPT recommendations. Recommendations are based on organic web presence factors. This may change as OpenAI explores monetization, but for now it is entirely organic.
Does ChatGPT read my Psychology Today profile?
+
ChatGPT can access PT pages when browsing is enabled, but it weighs your own website, Google reviews, and multi-site presence more heavily than any single directory profile. Therapists with only a PT profile are rarely recommended.
How do I get my practice recommended by ChatGPT?
+
Build a website with dedicated specialty pages and FAQ sections. Get Google reviews (aim for 10+). Be listed consistently across multiple directories. Write helpful content in your area of expertise. Keep your NAP consistent everywhere.
Is it ethical for AI to recommend therapists?
+
This is an active debate. The concern is that AI recommendations are based on web presence rather than clinical competence. However, the same critique applies to every other discovery mechanism. The key ethical issue is transparency about how recommendations are generated.
Will OpenAI build a formal therapist directory?
+
According to Behavioral Health Business, OpenAI is developing formal therapist referral capabilities. The timeline is not public, but it signals that AI-based therapist discovery will become more structured.
Related guides
Guide
How Clients Find Therapists in 2026
The full data on where therapy clients search
Guide
What Clients See When They Google You
The 3-second judgment clients make from page 1
Guide
Psychology Today Stopped Working
The data behind the PT decline
Guide
The Real Cost of Psychology Today in 2026
The ROI math most therapists have not done