AI Visibility for Mental Health Practices: Prompts, Facts and Fixes

People rarely ask an assistant for "a therapist." They ask for a specific fit. Here are realistic examples (hypothetical, written for this page):

• "Find a therapist near me who takes Aetna and does EMDR."
• "Which therapists offer telehealth and are licensed in both Colorado and Utah?"
• "Is there a counselor in Austin accepting new clients for teen anxiety?"
• "What's the difference between CBT and DBT, and who runs DBT groups near me?"
• "Affordable therapy with a sliding scale in Chicago."
• "Couples therapist in Portland with evening sessions."
• "Does [practice name] take Medicaid?"

Each prompt combines modality, payer, geography, license and availability. Every one of those is a fact you have to keep true in public.

Treat your prompt list as a proxy, not the real input. OpenAI says ChatGPT search typically rewrites a user's query into one or more targeted queries before it searches. Your job is to make sure the underlying facts are correct wherever an engine might look.

Measure each engine on its own. They draw on different indexes and behave differently.

• Google AI Overviews and AI Mode. Google states a page must be indexed and eligible to show with a snippet to appear as a supporting link, with no additional technical requirements. Normal Google eligibility is the gate.
• ChatGPT. Search answers depend on OpenAI's OAI-SearchBot. As of October 2026, the app searches at the model's discretion. When it doesn't, it answers from training data: OpenAI lists a February 16, 2026 knowledge cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path.
• Perplexity. Indexes through PerplexityBot.
• Claude. Search uses Claude-SearchBot, and web search in Claude.ai is a user toggle.
• Microsoft Copilot. Grounds on Bing, so Bing index status matters separately from Google.

Otterly's 2026 experiment showed why blended scores mislead: after Google deindexed one test site, Microsoft Copilot produced 2,718 of its 3,934 tracked citations while ChatGPT, Gemini and Google's AI features barely cited it. Health practices share this pattern with the wider AI visibility picture for medical practices.

Therapy practices change faster than most local businesses. Caseloads fill and open. Clinicians join and leave. Insurance panels get added or dropped. A therapist adds a telehealth license in a new state.

That creates two problems:

1. Frozen answers. A retrieval-off answer reflects whatever the model learned before its cutoff. Nothing you publish this month reaches it.
2. Old pages still get fetched. A retrieved answer is only as current as the pages the engine pulls, including a two-year-old directory profile or a blog post announcing a clinician who has since left.

The next move is a single facts page on your site: clinicians, license states, modalities, insurers accepted, fees or sliding-scale policy, telehealth versus in-person, and a dated line on new-client availability. Stating a specific fact on a fetchable page gives an engine a sentence it can quote, which is the only positive experimental signal for on-page content in our evidence base. If you can't keep "accepting new clients" current, write "availability updated [date], call to confirm" instead. A site rebuild can quietly break this, so read up on how a migration affects AI answers before you redesign.

These are the facts worth checking first for any therapy practice. The list reflects which details change often, not a measured error rate.

The free checker on this site asks up to three buyer questions and shows whether engines name you and what they say. It's a snapshot, so rerun it after you fix facts.

For "near me" and insurance prompts, engines often cite third-party pages rather than practice sites. The evidence shows a strong association between being named on those cited sites and appearing in the answer. It doesn't promise an effect, but it tells you where conflicting facts can come from.

Check these for consistency with your facts page:

• Therapist directories. Every profile, including old ones for clinicians who left.
• Insurer provider search tools. Payers keep their own lists, and errors there flow into answers about coverage.
• State licensing board lookups. These are the record for license states. Make sure names match how you present clinicians.
• Google Business Profile and Bing Places. Hours, address, telehealth flags.
• Referral lists from local hospitals, schools or employee assistance programs.

Pick one canonical name, address and description and use it everywhere. Our guide to managing your practice as an entity covers how to keep those details aligned.

Mental health pages carry more risk than most. Client information should never be on a public page. Put intake forms, client portals and anything with personal details behind a login, and mark utility pages noindex.

For public pages with passages you don't want quoted, Google and Bing document noindex, nosnippet, max-snippet:0 and data-nosnippet as the on-page controls. Bing says content marked data-nosnippet is still indexed but excluded from snippets and AI summaries. These are suppression tools. Applied by accident to a template, they hide your real facts, so check where they're set.

Don't try to win with volume. In Otterly's 2026 experiment, two new sites with roughly 1,000 AI-generated posts each were deindexed by Google (after about 14 and 90 days) with no manual action reported. Google's spam policy defines scaled content abuse as "when many pages are generated for the primary purpose of manipulating search rankings and not helping users." Hidden text or instructions aimed at AI crawlers are worse: they're manipulation and a reputational risk. If you're wondering how engines treat health topics, see whether AI search is stricter about health content.

Access is where the evidence is strongest. Run these checks:

1. robots.txt, per bot. Rules are independent. Blocking GPTBot only stops training use; OAI-SearchBot controls ChatGPT search. A blanket AI block can remove you from search answers. A short, correct example:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

If you want to opt out of training but stay in answers, read whether to block GPTBot.

2. Firewall and CDN. A firewall can return 403 to a crawler robots.txt allows. Since September 15, 2026, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and free-tier customers, while Search stays allowed. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.

3. JavaScript. Vercel found ChatGPT and Claude crawlers fetch JavaScript but don't execute it. If your insurers, clinicians or availability load inside a scheduling widget, those crawlers may see nothing. Put the facts in the server-rendered HTML.

Week 1: write down the truth

Build the facts page: clinicians, license states, modalities, insurers, fees, telehealth, dated availability. Make sure it renders without JavaScript.

Week 2: open the gates

Check robots.txt per bot, test firewall responses for each AI crawler, and confirm key pages aren't carrying stray noindex or nosnippet. Confirm index status in Google Search Console and Bing Webmaster Tools separately.

Week 3: clean up third parties

Update every directory, insurer and business profile to match. Remove departed clinicians. Fix licensing board name mismatches.

Week 4: measure and set a cadence

Ask your real prompts on each engine and record what each one says. See which competitors AI recommends for the same prompts. Set a monthly review tied to staffing and panel changes. Drop tasks with no evidence behind them using AI SEO best practices sorted by evidence.

None of this guarantees a mention. It removes the blocks and wrong facts you control.

Should a therapy practice block AI crawlers to protect client privacy? robots.txt isn't a privacy tool. Client information shouldn't be on public pages at all, so keep portals and intake forms behind a login. Blocking search crawlers such as OAI-SearchBot or PerplexityBot removes your public pages from those engines' search answers without protecting anything private. If you object to training use, you can block GPTBot separately, since OpenAI says each setting is independent.

Do I need schema markup or an llms.txt file for AI answers? No. Google says structured data isn't required for generative AI search and that Google Search ignores llms.txt. Ahrefs' matched test found AI Overview citations fell 4.6% after pages added schema. Existing markup does no harm, but it isn't a lever for AI answers. Your time is better spent on accurate facts in the HTML and on crawler access.

Should each clinician have their own page? It helps keep facts precise. A clinician page can state license states, modalities, populations served and insurers in plain sentences an engine can quote. The catch is maintenance: when someone leaves, remove or redirect their page and update every directory listing, or assistants may keep recommending a therapist who no longer works with you.

How often should I update dates on my pages to stay current in AI answers? Update a page when a fact changes, not on a schedule. No vendor states that page dates influence selection, and the main freshness study had no control group. Bumping a modified date without changing anything doesn't make your facts truer. A dated availability line is useful because it tells readers and engines when that specific fact was last checked.