Writer: AI health search cites journals, not pharma websites

What Writer tested

Writer says it used Writer Agent to run 40 health queries across oncology, obesity, immunology and neurology. Half of the queries were patient oriented and half physician oriented. The queries went through four consumer answer engines: Gemini/Google, ChatGPT, Perplexity and Claude. Writer then sorted all 777 cited sources by type.

The company frames the work against a broader shift in behavior. According to the article, roughly two thirds of physicians and three quarters of patients now use AI-powered search tools when researching a diagnosis, and AI has overtaken social media as a first stop for health information. Google, it notes, now delivers an AI-generated summary as its first response rather than ten blue links.

Journals lead for doctors, GoodRx for patients

The results differ sharply by audience. When answer engines respond to HCP questions, 19.5% of citations come from academic journals, which Writer says is 4x as often as in patient queries. Across both audiences, hospitals and research institutions appeared around 15% of the time.

• The top domain for doctors was PubMed/PMC, cited 50 times.
• The top domain for patients was GoodRx, cited 15 times.
• Only one domain, fda.gov, appeared in both top-8 lists.

Commercial and manufacturer content was barely visible. Writer cites DataForSEO research that found health content is "especially citation-friendly" and that LLMs lean toward authoritative sources such as the CDC, Mayo Clinic and peer-reviewed publications. According to that research, commercial domains get discovered during retrieval but rarely reach the final cited answer for HCP-oriented content.

As an example, Writer says a ChatGPT question about treatment pathways for metastatic melanoma draws on published clinical literature, NCCN guidelines and institutional pages, with no pharmaceutical product page among the cited sources.

Writer's recommendations: publish more, publish faster

Writer argues that evidence has become a company's most visible asset. It says this does not change who owns the work: Medical Affairs and HEOR keep scientific decisions, while commercial teams own the visibility strategy and budget advocacy. The company stresses that the regulatory firewall holds and that publications should not be commercialized.

The main problem, according to Writer, is volume. If a therapeutic area has only a dozen papers among hundreds of sources an AI engine synthesizes, the evidence barely registers. Writer calls for more medical writers, HEOR studies, congress abstracts and real-world evidence analyses, noting that conference abstracts and posters are also indexed in PubMed.

On journal choice, Writer says NEJM, Lancet and JAMA reject over 95% of submissions and can take months to decide, and teams should keep aiming for them. But it recommends a portfolio that includes mid-ranked journals, which it says often decide in weeks. It cites a bibliometric study of general medical journals that found a median submission-to-acceptance time of 123 days.

Writer also says the volume needed is beyond what manual writing teams can produce, so AI-assisted drafting is necessary, with human experts in the loop and MLR governance intact. It points to a separate 2026 study in which expert reviewers could not reliably distinguish AI-written from human-written abstracts, which it takes as a reason not to shortcut peer review.

What it means for life science marketers

Writer's advice is to audit current spend on channels AI engines effectively ignore, then start with one therapeutic area and build a publication pipeline. It says display ads, sponsored content and SEO-optimized product pages do not influence what ChatGPT tells a patient or what OpenEvidence tells a physician, and that redirecting that spend is possible without new budget.

The company also proposes a new KPI: presence in AI-generated answers, in place of impressions, click-through rates and page rankings. The study is a vendor analysis of 40 queries, and the findings reflect Writer's own interpretation. Still, the pattern it reports points to closer cooperation between Marketing and Medical Affairs teams.