
ABA Marketing | Josh Blicker

Key Points:
ChatGPT referral traffic in GA4 shows recorded visits attributed to ChatGPT. For an ABA clinic owner, the practical question follows: did those visitors contact the practice, and could the clinic serve those families?
A referral report starts that investigation. Your intake records complete it. At The CMG, we connect marketing reporting with intake outcomes so owners can judge channels against qualified opportunities. The steps below explain how to inspect AI sources, organize the report, and carry available source information into internal reporting.
Clinics can find identifiable ChatGPT visits in Reports > Acquisition > Traffic acquisition. Google introduced its dedicated AI Assistant channel in May 2026. Recognized assistant referrals receive the ai-assistant medium and (ai-assistant) campaign name.
The starting view should use session dimensions, which describe visits. First-user dimensions answer a different question: how someone originally discovered the website.
A practical first check follows this sequence:
Google’s Traffic acquisition report instructions explain these controls. A missing report may reflect the property’s report menu configuration.

GA4’s AI Assistant channel excludes Google’s AI Overviews and AI Mode. Google includes those visits within Organic Search under its default channel definitions.
AI search traffic in GA4 therefore spans more than one reporting category. A single AI Assistant total cannot represent every visit influenced by an AI answer.
Other limits affect interpretation:
Direct traffic should remain separately reported unless evidence supports another source. Our healthcare AI search visibility guide covers testing mentions and citations beyond website clicks.
A custom group can organize AI referral traffic in GA4 when sources appear across several channels or you need a consistent historical view. The existing AI Assistant channel in GA4 may already cover your immediate reporting needs.
An Editor or Administrator can create a group under Admin > Data display > Channel groups:
Exact source rules need updating when new values appear. Broad matches such as “ai” can include unrelated websites.
Google confirms that custom channel groups apply retroactively to reports. They reorganize existing data; missing referrers remain missing. Setting a custom group as the property’s primary channel group changes that separate dimension going forward.
Healthcare AI traffic tracking must keep protected health information out of GA4. Google does not offer a Business Associate Agreement for Analytics and prohibits exposing protected health information to the service.
The reporting plan should separate website analytics from care-related records:
A generic event name does not remove the sensitivity of the underlying action. HHS explains that tracking obligations depend on the information disclosed and its context. Public healthcare page visits do not automatically establish that a disclosure involves protected health information.
The GA4 steps in this article apply only where analytics can operate without collecting prohibited information. Our ABA marketing data tracking article explains the broader separation between internal reporting and platform measurement.

ChatGPT traffic analytics becomes more informative when each action has an accurate label. A phone-button click shows an attempt to call. It does not prove that a conversation happened.
The same distinction applies across website actions:
| Action | What the record establishes |
|---|---|
| Phone-button click | Someone activated the phone link |
| Booking-link click | Someone opened the scheduling destination |
| Form start | Someone began the form |
| Successful submission | The form accepted a submission |
| Completed intake conversation | The intake team spoke with the family |
For interactions permitted in GA4, a configured event can become a key event. Care-related actions that reveal sensitive information should stay in protected reporting.
A test should confirm that a successful action records once, while an unsuccessful attempt does not produce a success event. Repeated submissions and calls also need deduplication inside intake reporting so one family does not become several inquiries.
AI search lead attribution connects an inquiry’s available source record with its later intake status inside controlled systems. GA4 alone cannot establish that an AI visitor became an enrolled client.
A customer relationship management system, or CRM, can retain the source attached to an inquiry alongside follow-up records. Our CRM and practice management comparison explains which operational tool handles each stage.
The website-to-CRM handoff needs to preserve the available referral or campaign source when an inquiry arrives. Call-tracking records need their own source connection. A matching timestamp alone should not identify a family or justify linking two records.
An ABA AI referral tracking record can include:
A parent’s answer to “How did you hear about us?” adds context, but self-reported discovery should remain separate from the recorded website source. Missing connections should stay labeled unknown.
Our multi-state ABA reporting work connected marketing and internal CRM data into a 12-week view by state, channel, intake cost, and children-in-service cost. That project illustrates the reporting approach; it does not establish AI-specific results.
Through Intake as a Service, we contact inbound inquiries, apply practice-specific qualification criteria, and record outcomes in the CRM. Those records help distinguish an inquiry from a family ready for the next intake step.
A monthly report should keep ChatGPT and Perplexity referral traffic visible as separate sources before combining assistant totals. Each row should show available sessions, unique inquiries, qualified inquiries, and later outcomes from protected reporting.
Illustrative example, not CMG results: A clinic records 100 ChatGPT sessions and links eight unique inquiries to that source. Four meet its qualification criteria, and two schedule assessments. The inquiry qualification rate is 4 ÷ 8, or 50%.
The report still needs context. Two assessments do not establish two enrollments, and recent inquiries may still be waiting for follow-up.
A consistent comparison with Organic Search should use:
Small totals deserve caution. One additional inquiry can change a percentage sharply. Unknown-source inquiries need their own row so channel totals do not imply complete attribution.

GA4 does not provide the person’s ChatGPT conversation or prompt. Source information and the page they entered can describe the website visit, but they cannot establish the wording of a private conversation.
An existing property can report AI sources alongside other channels. A separate property is unnecessary solely because the traffic comes from an assistant, and splitting reports can make channel comparisons harder.
Google says data processing can take 24 to 48 hours. A test missing from a standard report immediately after the visit does not establish a tracking failure. Recent data can change during processing.
AI referral reporting helps owners decide which sources deserve attention when session counts connect with qualified inquiries and later outcomes. Reliable definitions and honest reporting gaps support that decision without overstating what GA4 can see.
We help practices connect marketing performance with intake reporting and follow-up. A conversation with The CMG can focus on your current source reports, CRM handoff, and qualification records so your team can identify where attribution stops.