Who We Are What We Do Census Playbook FocusFlow Book a Roadmap Call
AI Search & GEO

How do I measure whether AI search mentions my center?

The short answer

You measure AI search visibility with a fixed prompt panel, brand-mention tracking, referral traffic signatures, and call source attribution. Build a list of 20 to 40 prompts a real family would ask, run them monthly in ChatGPT, Gemini, and Google's AI Overviews, and log whether your center is absent, cited, named, or recommended. Pair that with AI referral traffic in analytics and a disciplined how-did-you-hear-about-us question at intake. Good looks like steady month-over-month movement up that ladder, ending in admissions whose source traces to an AI answer.

Build a prompt panel and run it on a schedule

A prompt panel is a fixed list of 20 to 40 questions, asked the same way, in the same engines, every month. Build it in four buckets:

  • Local discovery: best drug rehab near your city, medical detox in your county.
  • Qualified discovery: rehab in your state that takes a given payer, PHP programs near your city for young adults.
  • Brand verification: is your center legitimate, reviews of your center.
  • Comparison: your center versus the competitor you lose admits to.

Run each prompt in ChatGPT, Gemini, and a Google search that triggers an AI Overview, in logged-out or fresh sessions. Log engine, prompt, date, and outcome in a spreadsheet. It takes an afternoon a month and replaces guessing with data.

Two rules keep the panel honest. Use the plain, slightly messy phrasing families actually type, not prompts only your marketing team would write. And resist editing the panel every month: you can add prompts, but changing existing ones resets your trend line.

Score four levels, not yes or no

Mentions are not binary. Score each prompt on a four-step ladder:

  1. Absent. You do not appear at all.
  2. Cited. Your site appears as a linked source, but the answer does not name you.
  3. Named. The answer mentions your center among the options.
  4. Recommended. The answer presents you as a primary suggestion, with accurate details.

Score for accuracy too: a mention that lists the wrong phone number or a program you no longer offer is its own action item. Month over month, you want prompts climbing the ladder. Moving from absent to cited across ten local prompts is real progress even though no family has heard your name yet. The work that drives the climb is laid out in GEO for addiction treatment centers and structuring content for AI citations.

Watch the referral signatures in analytics

AI engines leave fingerprints. In GA4, build a report filtered to referral sources like chatgpt.com, gemini.google.com, perplexity.ai, and copilot.microsoft.com. The volume will look small, and that is expected: most AI influence converts as a branded search or a direct phone call after someone reads an answer, not as a referral click. Treat rising AI referrals, rising branded search, and rising direct traffic as one combined signal that machines are talking about you. Watching only the referral line undercounts the effect badly.

The same logic applies inside Search Console: a rise in impressions for your brand-name queries often follows AI mentions, because people verify what a chatbot told them before they call anyone.

Close the loop with call attribution

The measurement that matters most is the one attached to admissions. Train admissions staff to ask how the caller found you, and log answers like asked ChatGPT or saw the AI answer on Google as distinct sources in the CRM. Pair that with call tracking so every call carries a source automatically, and roll it all up into cost per admission by channel. Our ImpactEngine platform handles the source logging and follow-up automatically; see what we do.

What good looks like month over month: more prompts scored cited or better, brand queries answered accurately, AI referrals and branded search trending up together, and the first admissions with an AI source logged. If the panel shows you absent everywhere, start with why your center isn't showing up in ChatGPT, or have us run the baseline for you.

Questions operators ask

AI answers change every time I ask. Is this data even reliable?
Any single answer is unstable; the aggregate is not. Twenty to forty prompts scored monthly gives you a trend line that smooths out the randomness. You are measuring direction, not a fixed ranking.
How often should I run the panel?
Monthly is the right cadence for most centers. AI visibility moves on a scale of weeks to months, so weekly checks add noise without insight, and quarterly checks let problems run too long before anyone notices.
Is there software that tracks this automatically?
Yes, AI brand-mention tracking tools exist and are maturing quickly, and we use them in client engagements. A disciplined spreadsheet is enough to start, and it forces you to actually read the answers families are seeing.
Put this to work

Want this handled for you?

Thirty minutes with people who already run this playbook for treatment centers across the country.

Get Your Customized Roadmap