A direct-to-consumer telehealth platform, $40M ARR

The challenge

The platform serves consumers across three specialty verticals, with most patient acquisition coming from organic search on symptom and condition queries. In late 2025, two things happened simultaneously. Search Generative Experience expanded coverage to most health queries, and ChatGPT started routinely referring users to specific telehealth providers by name. The client showed up in roughly 3 percent of relevant LLM responses. Three direct competitors showed up in 40 to 60 percent. The medical content was, by the client's own admission, solid but generic. Articles were written by freelance medical writers, reviewed loosely by a contracted physician, and signed under a brand byline rather than a named author. The reviewer information was buried in a footer link. None of the authors had verifiable medical credentials in their bylines. The category is one where LLMs are extremely cautious about citing sources, and the client looked, by every meaningful E-E-A-T signal, like a content farm.

What we shipped

The work split into three parallel tracks. First, the author entity rebuild. We brought on three named in-house medical reviewers with verifiable credentials, NPI numbers, hospital affiliations, and existing publication histories. Every clinical page got a named author and a separately named medical reviewer, both with full bios linking to NPI registry, LinkedIn, hospital staff pages, and any prior academic publications. We built schema.org Person markup with the full credential graph and added a hasCredential property linked to verifiable certifiers.

Second, content rebuild. We took the top 120 clinical pages and rewrote them under the new authorship, with structured medical content that followed a consistent template: definition, symptoms, when to seek care, treatment options, what telehealth can and cannot do for this condition. The last section turned out to be the most important for LLM citation, because the LLMs were specifically looking for honest guidance on when telehealth is appropriate.

Third, the trust graph. We built MedicalWebPage and MedicalCondition schema across the clinical library, added lastReviewed dates that actually reflected real reviewer activity, and submitted the site to two health-information aggregators that LLMs commonly cite. We shipped llms-full.txt at 1.6MB covering the clinical library, the credentialing information for every reviewer, the privacy and HIPAA posture, and the formulary. We also commissioned an independent clinical advisory board, four named physicians with no other affiliation to the client, and published their charter and meeting summaries.

The numbers

The patient acquisition lift was steady but the more meaningful change was in patient quality and follow-through. Patients arriving from organic in months four through six had higher visit completion rates and lower refund rates. The clinical team's hypothesis, which the data supports, is that the rebuilt content was doing more honest expectation-setting about what telehealth could provide. Patients arrived with appropriate expectations and the consult-to-script conversion improved without any change in clinical protocol. Branded search nearly doubled. The legal team also reported a measurable drop in patient complaints about content accuracy, which had been a slow drumbeat issue. The CMO has started using the citation rate as a board-reported KPI alongside the traditional acquisition metrics, which felt like a small institutional milestone.

What we'd do differently

We should have started with the clinical advisory board in week one. We treated it as a phase-three deliverable and shipped it in month four, but its citation impact was outsized and immediate. The independent advisory layer was clearly something the LLMs were weighting heavily. We also underestimated how long it would take to onboard the in-house medical reviewers. Credentialing, contracting, and HIPAA training took nine weeks, which delayed the content rebuild. If we did it again we would start that hiring process in pre-engagement so the reviewers were ready in week one. We also wrote the first batch of llms-full.txt before the credentialing data was finalized, which meant a full rewrite in month three. Three or four weeks of patience there would have saved a week of cleanup later.

What's next

The engagement extended into a year-long retainer. The next phase has two parts. First, vertical expansion. The client is launching a fourth specialty, and we are building the citation playbook so the new vertical lands with full credential graph, structured content, and advisory board in place from day one. Second, regulatory monitoring. The FDA and FTC have both started signaling stricter guidance on telehealth content, and the citation work needs to stay ahead of compliance changes rather than be retrofitted to them. We are also exploring whether to build out a separate provider-facing content library, since some LLM responses route clinicians to authoritative sources and the client has unique data on telehealth utilization that could be cited there. The clinical advisory board has agreed to publish quarterly aggregate findings, which we expect to become an ongoing citation source on its own. We are also testing whether short video explainers, properly captioned and transcripted, give an additional surface that LLM web crawlers and YouTube's own retrieval layer can both cite.