When patients start asking ChatGPT instead of Googling, a marketing program needs new instruments. Here’s how we built them, and what the data told us to do next.
A growing share of the people looking for treatment never click on a search result. They ask ChatGPT. They read the AI Overview at the top of Google and stop scrolling. They tell Gemini their situation and take the recommendation they’re given.
Most agencies are still calling this an emerging trend. For our client, a nationally recognized behavioral health network with eight web properties spanning inpatient, outpatient, virtual, and specialty programs, we made it a measured, managed marketing channel, the same way organic search became one twenty years ago.
AI visibility is not a trend we are watching. It is a channel we instrument, measure, and optimize like any other. This case study shows exactly how we do it across a multi-site network: the tracking architecture we built, the data it produced, and the strategy that data dictated.
At a Glance
- Eight web properties instrumented for AI visibility: referral tracking from ChatGPT, Gemini, Claude, Perplexity, and Copilot, plus AI Overview presence tracking in Google
- The network’s flagship site appears in Google AI Overviews for 60.9% of its tracked keywords
- 825 tracked AI-referred sessions in a single month across the network, 88.6% of them from ChatGPT, producing tracked calls and form submissions on nearly every property
- AI visitors don’t just browse: one specialty facility converted AI-referred sessions to contact events at a 12.5% rate
- A five-pillar optimization strategy now in motion: crawler access, homepage authority, conversational content, off-page citations, and tracking modernization
First, Instruments: You Can’t Manage What You Can’t See
Most conversations about “AI SEO” skip the unglamorous first step: measurement. We started there, defining a framework that separates potential visibility from actual visitor behavior, because they are different things, measured by different tools:
- AI referral sessions and key events (GA4): real people who clicked from an AI tool to a website, and what they did there. Actual foot traffic, actual conversions.
- AI Overview presence (Ahrefs keyword data): whether the sites appear within Google’s AI-generated answer summaries for tracked keywords. Potential visibility, not guaranteed clicks.
- Manual prompt testing: running standardized sets of industry-relevant prompts across AI platforms under controlled conditions (private sessions, disabled personalization, repeated runs) to see whether and how the brand appears in AI answers, including the sentiment of the description.
We built the tracking to correct for the three things that make AI traffic look smaller than it is: consent banners that block analytics, “dark AI” referrals that are mislabeled as direct traffic, and AI platform interface changes that shift referral numbers without any real change in visibility. Correcting for all three is what makes the numbers in this case study trustworthy.
The Baseline: What a Month of AI Traffic Actually Looks Like
With instrumentation in place across all eight properties, a single month of data gave us the network’s full AI-visibility portrait:
- 825 AI-referred sessions in a single month across the network, producing tracked key events on nearly every property, from calls to form submissions
- ChatGPT dominates: 88.6% of all AI referrals. Gemini, Perplexity, Claude, and Copilot are, for now, rounding errors by comparison
- Half of AI visitors (49%) land on a homepage. When AI tools recommend the network, they mostly send people to the front door to verify legitimacy and begin their journey
- The other half land on remarkably specific pages. AI engines matched granular user intents to granular pages: specialty clinical programs, workshop pages, insurance-and-admissions content, virtual treatment pages, and city-level program pages
- The flagship site now appears in Google AI Overviews for 60.9% of its tracked keywords, a visibility surface that did not exist as a category two years ago
Two findings stood out as more than trivia. First, AI-referred visitors convert. One specialty facility converted AI-referred sessions into contact events at 12.5%, a rate most channels would envy, consistent with the theory that someone who has already described their situation to an AI arrives unusually qualified. Second, the pages AI tools cite for specific questions are exactly the pages built with strong architecture and genuine substance: the same local program pages and service pages that win traditional rankings. AI search isn’t a separate content universe. It’s a stricter judge of the same one.
The Strategy: Five Pillars, Built From the Data
The baseline shaped a five-pillar program now underway across the network:
- Guarantee crawler access. With ChatGPT driving nearly 89% of AI referrals, OpenAI’s crawler (and its peers at Perplexity, Google, and Anthropic) must reach every property without obstruction. This includes auditing hosting and security layers that silently block AI crawlers.
- Strengthen homepage anchor equity. Since half of AI visitors land on homepages, each property’s homepage must clearly name and link every specialty, program, and location, because vague homepage copy means AI engines undersell the network’s actual depth.
- Pivot content to conversational Q&A. People ask AI tools full questions (“does insurance cover virtual treatment in California?”), not keyword fragments. Content is being restructured to answer those questions directly, in extractable, citable form, supported by healthcare-appropriate structured data. A network-wide schema plan is in progress.
- Build off-page authority AI engines trust. LLMs calibrate trust from what the wider web says: medical directories, authoritative citations, and brand mentions across the behavioral health space. Digital PR and directory accuracy are now AI-visibility work, not just brand work.
- Modernize measurement. Server-side tracking is under evaluation to close the consent-related tracking gap, so the network’s true AI traffic and conversions aren’t systematically undercounted.
None of these pillars is exotic. That’s the point. The organizations that will win AI search are the ones doing disciplined fundamentals (architecture, substance, schema, authority) while measuring the new channel honestly enough to steer.
What We Tell Clients About This Channel
Three things, candidly.
It is early, and it is already producing patients. Hundreds of monthly sessions and dozens of monthly conversions across a network are real people entering treatment journeys through a door that barely existed two years ago. The trend line only points one way.
It rewards the same work, judged more harshly. AI engines cite few sources per answer. “Winner takes all” isn’t a metaphor: being the third-best resource on a topic means being invisible in an AI answer. Every investment in genuinely authoritative, well-structured content now pays in two currencies, rankings and citations.
It has to be measured correctly. Consent banners, mislabeled traffic, and shifting AI interfaces will distort AI visibility reporting unless the tracking accounts for them. Ours does, which is why our clients can act on the numbers we give them.
Want to Know How Visible You Are in AI Answers?
Most organizations have no idea whether ChatGPT recommends them, misdescribes them, or has never heard of them. Finding out is now a measurable, improvable discipline.
Astute builds AI visibility programs on real instrumentation: referral and conversion tracking, AI Overview monitoring, controlled prompt testing, and the content, schema, and authority work that turns visibility into inquiries.
Talk with Astute about an AI visibility audit.
FAQs About AI Visibility
What is AI visibility?
AI visibility is how often, and how favorably, your brand and content appear in answers generated by AI tools like ChatGPT, Gemini, Claude, and Perplexity, and in AI features like Google’s AI Overviews. Where traditional SEO measures your position on a results page, AI visibility measures your share of voice inside the answer itself.
How do you measure AI visibility?
Three layers: analytics (tracking sessions and conversions referred from AI platforms in GA4), SERP data (monitoring which of your keywords surface AI Overviews and whether you’re cited, via tools like Ahrefs), and controlled manual testing (running standardized prompts in clean, unpersonalized sessions, repeated for consistency). Each layer answers a different question; no single tool tells the whole story.
Which AI platform actually sends traffic?
In our measurement across eight healthcare properties, ChatGPT drove 88.6% of AI-referred sessions, an order of magnitude more than Gemini, Perplexity, Claude, and Copilot combined. That distribution will shift over time, but today, ChatGPT is where AI-referral optimization earns its keep.
Do visitors from AI tools actually convert?
They do, at strong rates. Someone who has described their needs to an AI assistant and clicked its recommendation arrives pre-qualified. In our data, one facility converted AI-referred sessions into contact events at 12.5%, and network-wide, AI referrals produced tracked calls and form submissions every month.
How do you optimize content for AI search?
The fundamentals are familiar: genuinely expert content, clear site architecture, structured data, and off-page authority. The emphases shift. Content should directly answer full conversational questions in extractable form, homepages must comprehensively represent what you offer (AI tools route heavily to them), and AI crawlers need unobstructed access. Strong traditional SEO remains the best predictor of AI citations.
Is it too early to invest in AI visibility?
No, and the window is the point. The channel is still small enough that baselines are cheap to establish and category leadership is unclaimed, and it is growing fast enough that waiting means competing against whoever built the authority first. Instrument now, optimize the fundamentals, and be skeptical of anyone promising guaranteed placement in AI answers.
Curious what ChatGPT says about your business? Talk with Astute.