AI for therapists usually means one of two things: tools that write clinical notes, or tools that help people find and choose you. This page is about the second. Some people now ask AI assistants to shortlist therapists, AI can draft your marketing, and chat widgets can answer questions at 2 a.m. Some of that is useful. Some of it, especially anything near a crisis or client data, should stay human.
In short: AI marketing for a therapy practice means three jobs: being named when people ask AI assistants for a therapist, drafting content and ads faster, and answering practical questions on your site. Keep client information out of tools without a BAA, never let AI handle a crisis (point to 988), and review everything published under your license.
What AI marketing means for a therapist practice
The search demand shows the split. "ai for therapists" and "chatgpt for therapists" each draw about 390 US searches a month, with clicks costing about $20.60 and $18.04. "ai tools for therapists" draws about 50 ($19.38), "ai for counselors" about 40 ($17.75) and "ai tools for counselors" about 10 ($12.21) (Google Ads Keyword Planner data via DataForSEO, US, 12-month average to Sep 2026). Many of these searchers want note-taking tools.
Keep the two categories apart:
| Item | Clinical documentation AI | Marketing AI |
|---|---|---|
| Examples | Session note drafting, transcription | Content drafts, ad copy, AI visibility checks, website chat for practical questions |
| Touches client data? | Yes, always | Shouldn't, except a chat or form that collects inquiries |
| BAA needed? | Yes | Only for tools that receive inquiries or client information |
| Whose decision? | Yours, as the clinician, with informed consent | Yours, as the business owner |
This page covers the right-hand column. For the wider picture across specialties, see AI marketing for medical practices.
How patients use ChatGPT, Gemini and AI Overviews to choose a therapist
Choosing a therapist involves questions people find hard to ask a person: What's the difference between a psychologist and an LCSW? Does EMDR work for childhood trauma? Who near me takes my insurance and sees teenagers? AI assistants answer all three in one conversation, and Google's AI Overviews put a summary above the usual results.
That changes where your first impression happens. A prospect may hear your name from an assistant, then check your website and directory profile before writing to you. If the assistant names three other practices instead, you never learn you were passed over. Our guide to AI visibility for practices explains how these answers are assembled.
How people ask AI assistants for a therapist
People type to assistants the way they'd talk to a friend, with more detail than a search box gets.
Illustrative example (prompts a prospective client might type):
- "Find me a therapist in Austin who takes Aetna and does EMDR for trauma."
- "Who's a good couples counselor near Lakewood, Colorado with evening telehealth?"
- "My 15-year-old might have OCD. What kind of therapist should we look for in Raleigh?"
- "Is there a therapist in Toronto who specializes in postpartum anxiety and has a sliding scale?"
Assistants answer from what they can find and read: practice websites, directory profiles, Google listings and what others write about you. Each prompt combines a specialty, a location, a logistics detail (insurance, evening hours, sliding scale) and often an age group. Practices that state all four plainly have more to be matched on. No one can promise an assistant will name a particular practice.
To see what assistants say about you now, run the free AI visibility check in Medfy's audit. It asks Claude and Gemini, with live web search, whether they name your practice, and checks whether the crawlers behind ChatGPT, Perplexity, Claude and Gemini can read your site.
AI notes, AI chat and the ethics codes: APA, ACA, NASW and AAMFT
None of the major codes bans AI, but each sets limits that apply directly.
- Client data. The APA Ethics Committee's FAQ on AI and social media (opens in a new tab) (November 2023) says clinicians "should not input identifiable client data into ChatGPT or other machine learning tools." NASW 1.07(m) (opens in a new tab) asks social workers to protect the confidentiality of electronic communications, and ACA H.2.d (opens in a new tab) asks counselors to use current encryption standards for websites and technology-based communications.
- Clinical judgment. The AAMFT Code (opens in a new tab), revised effective January 1, 2026, says therapists "do not use technology in place of their own independent treatment decisions" (6.2) and must use documentation technology that meets professional and legal standards (6.5).
- Testimonials and claims. AI makes fake praise easy. APA Standard 5.05 (opens in a new tab) bars soliciting testimonials from current clients, ACA C.3.b extends that to former clients, and APA 5.01 bars false or deceptive statements about your services, whoever drafted them.
- Crisis. An AI chat on a therapy site must never try to handle a crisis conversation. Configure it to show the 988 Suicide & Crisis Lifeline (opens in a new tab) (call or text 988, or chat at 988lifeline.org, 24/7) and 911 the moment a message suggests risk, and to alert your team.
What to automate and what to keep human:
| Automate, with your review | Keep human |
|---|---|
| First drafts of educational posts and specialty pages | Any reply to an inquiry that mentions symptoms |
| Variations of your directory profile wording | Anything involving risk, crisis or safety |
| Ad copy variants checked against Meta and Google health policies | Deciding fit, and referring out |
| Answers to practical FAQs: fees, insurance, telehealth states | Responses to reviews |
| Weekly checks of what AI assistants say about you | Final approval of anything published under your license |
Directory profiles, your website and AI answers
Therapy has a constraint most specialties don't: the codes discourage soliciting client reviews, so you can't lean on review volume. Your website and directory profiles carry more of the weight.
- Your Psychology Today profile is one of the pages an assistant may find and draw on. Make the specialties, ages, insurance and telehealth states specific, and match them to your site.
- Your therapist website should have a page per specialty that answers the question in plain language, with your license type, location and logistics.
- Consistency across Google, directories and your site lets an assistant connect them to the same practice.
- Crawler access. If your site blocks the crawlers behind ChatGPT, Perplexity, Claude or Gemini, those assistants have less to go on.
Which AI tools are safe with patient information
Consumer AI chat accounts aren't covered by a business associate agreement (BAA), so nothing identifying a client belongs in them. OpenAI, Anthropic, Google and Microsoft offer BAAs only for specific business products. What their own pages said on October 22, 2026:
| Vendor | What its official page says |
|---|---|
| Anthropic | Its BAA article (opens in a new tab) covers commercial products: the first-party API once a BAA is signed and PHI use is enabled with Anthropic's sales team, and Enterprise plans where the owner turns on HIPAA settings and accepts the BAA; some features are excluded |
| Workspace customers who sign Google's BAA can use covered services (opens in a new tab), including the Gemini App (excluding Gemini in Chrome) and Gemini in Google Workspace; Google says customers without a signed BAA must not use PHI (opens in a new tab) in Workspace | |
| Microsoft | Its BAA (opens in a new tab) is offered through its data protection terms for in-scope enterprise services, and its in-scope list for commercial Office 365 includes Microsoft Copilot and Copilot Chat |
| OpenAI | We couldn't load OpenAI's help pages when checking; confirm current BAA eligibility directly with OpenAI before using any OpenAI product with client data |
Plans and terms change, so confirm before you rely on any of them. Our guides on whether ChatGPT is HIPAA compliant and on HIPAA and AI tools go deeper, and HIPAA for therapists covers the rest of your setup.
How Medfy uses AI for therapists
Medfy is AI marketing software for independent practices, built HIPAA-compliant by design. The $349/month base plan includes:
- Your website, rebuilt and hosted on your own domain, with a page per service, structured data and HIPAA-compliant forms, plus 500 AI credits a month for edits asked for in plain English.
- Automated SEO: researched articles every week, drafted from MedlinePlus and PubMed, which you approve or set to auto-publish. Most therapists will want to approve each one.
- AI visibility tracking: every week Medfy asks ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews the questions patients ask, and shows whether they name you and who they name instead.
- First-party analytics, including visits from AI assistants, and an AI CMO dashboard with a weekly brief.
- A patient CRM and inbox, access-controlled, with patient information in separate, encrypted HIPAA databases in the US or Canada.
Add-ons: Google Ads management ($199/month), Meta ads + AI creatives ($199/month) and a HIPAA AI chat widget ($49/month) that answers from your practice's information, captures new-patient requests and hands off to your team, covered by your BAA. Ad spend goes directly to Google and Meta, never marked up; billing is monthly and you can cancel anytime. Medfy doesn't answer phones, send email or text campaigns, request reviews or post to social media.
Before subscribing to anything, Medfy's free audit shows where you stand today. For the rest of the funnel, see our therapists marketing guide.
Frequently asked questions
Can ChatGPT recommend my therapy practice?
It can, but no one can promise it will. Assistants such as ChatGPT, Gemini and Perplexity build answers from what they can find and read: your website, directory profiles such as Psychology Today, your Google listing and what others say about you. Clear specialty pages, consistent details everywhere and a site AI crawlers can read improve your odds. Check what assistants say about you now before changing anything.
Is it ethical for therapists to use AI in marketing?
Yes, within the same rules as any marketing. Claims must be accurate, testimonials can't be solicited from current clients (APA Standard 5.05, ACA C.3.b), and you stay responsible for anything published in your name. Never put identifiable client information into a consumer AI tool, and never let AI write reviews or testimonials. Review every AI draft before it goes out under your license.
Is ChatGPT HIPAA compliant for therapists?
Not in its consumer form. HIPAA compliance depends on a signed business associate agreement and how you use the tool, and consumer AI chat accounts aren't covered by one. AI vendors offer BAAs only for specific business products, with conditions. Until you have a BAA for the exact product you use, keep client names, notes and anything identifying out of it.
Should a therapy website have an AI chatbot?
Only for practical questions, and only with a BAA. A chatbot can answer fees, insurance, telehealth states and how consults work at 2 a.m., and capture a consult request. It must never try to handle a crisis: anything that sounds like risk should show 988 and 911 straight away and alert your team. If a vendor won't sign a BAA or explain its crisis setup, skip it.