Pediatric Clinic Acquisition Study methodology.

• Format: phone interviews with clinic owners and senior leaders
• Fieldwork: September and October 2026
• Clinics included: 624, made up of 463 single-location clinics, 97 clinics with two to four locations and 64 groups with five or more
• Specialties called: applied behavior analysis (ABA), speech-language pathology, occupational therapy and pediatric dentistry
• Reference period: the 12 months before each call
• Inclusion rule: respondents rated their confidence in their answers at 75% or higher

Each call followed the same two-part format. For every acquisition channel, we asked the respondent a version of this question:

“If you had to take your best guess, how many of your new clients in the last 12 months have come from [channel]?”

Once the channel estimates were recorded, we asked how confident the respondent was in their answers. These were conversations with busy owners rather than a scripted questionnaire, so the exact wording shifted slightly from call to call, but the substance of both questions stayed the same on every call.

The four channels were referrals and word of mouth, meaning families who arrived through a recommendation; paid ads on Google and Meta; Google Search, meaning families who found the clinic organically and reached out through its website; and AI referrals, meaning families who told the clinic they discovered it through an AI tool such as ChatGPT.

An AI referral in this study is a family who asked an AI assistant such as ChatGPT, Gemini or Perplexity to recommend clinics, was pointed to the clinic in that answer, and then contacted the clinic themselves. The category does not include leads captured by a chatbot on the clinic’s own website.

We kept a clinic in the study only when the respondent rated their confidence at 75% or higher. Every clinic below that line was dismissed entirely and none of its answers appear anywhere in the results, because an estimate the owner doubts would only add noise to the picture. The 624 clinics in the published figures are the ones that passed this filter.

We built the call list by scraping public listings of pediatric clinics and then randomized the order of that list, so that no region, specialty or clinic size was called first by design. Callers worked through the randomized list in order, and we stopped calling once we judged that we had collected enough responses to report the results with confidence.

We did not record each clinic’s specialty during the calls. The study therefore cannot break results down by ABA, speech, occupational therapy or dentistry, and we cannot say how many clinics of each type are in the final sample.

We grouped every clinic by its number of locations. For each clinic, we turned its answers into the share of its new clients that came from each of the four channels, and then averaged those shares across all the clinics in the group. Every clinic counts equally in its group’s average, whether it took on ten new families last year or several hundred.

Averages are rounded to whole percentages and the four channels in each group add up to 100%. The same figures appear in the results table, the charts and the downloadable CSV.

As soon as a clinic’s answers were recorded, we removed the clinic from the call list so it could not be counted twice, stripped its name from its answers and added the anonymous figures to the combined dataset. We handled the data this way because acquisition numbers are competitive information, and no owner wants rivals to see exactly how their families find them. Nothing in the published data can be traced back to a single practice.

Every figure in the study is self-reported. Owners and leaders gave their best estimates from memory and experience, not exports from a CRM, call tracking or analytics, so individual answers may be off and the results should be read as a picture of where clinic leaders see their families coming from.

The confidence filter raises the quality of the answers we kept, but it may tilt the sample toward clinics that watch their lead sources closely, and those clinics could differ from the ones that do not. Because the call list came from public online listings, clinics without an online presence are not represented, and because we stopped calling at a point we judged sufficient rather than at a fixed target, the group sizes are uneven. The five-or-more group has 64 clinics, so its figures carry more uncertainty than the single-location group’s.

Because the results average clinics rather than pooling every client, a small clinic carries the same weight as a busy one within its group. The figures describe the typical clinic in each size group, which can differ from the share of all new families across those clinics combined.

AI referrals deserve a specific note. Families often ask an AI tool for options and then search for the clinic by name or call it directly, which means website analytics usually credit those visits to Google or to direct traffic. An owner who hears “we found you on ChatGPT” at intake will count that family as an AI referral, so self-reported AI shares tend to run higher than analytics-based measures of AI traffic.

Finally, the study describes where new clients came from at clinics of different sizes. It does not measure advertising return, cost per acquisition or whether any channel caused a clinic to grow.

The data and charts are free to reuse under a Creative Commons Attribution 4.0 license. Please credit the Driftlss Pediatric Clinic Acquisition Study 2026 and link to the study. The figures are available as a CSV download.

• November 2, 2026: study and methodology first published.