Name your clinicians
Author and medical-reviewer bylines with real credentials on every clinical page. This is the single strongest signal AI uses to decide whether a health page can be quoted.
Optimise Content →Solutions · By industry
AI answers health questions from clinical sources only.
Patients ask AI before they book, and AI answers from Mayo Clinic, the NHS and PubMed, not from marketing pages. Prefer shows which health prompts skip you, which clinical sources decide them, and what to publish so AI names you without hedging.
“best treatment for chronic insomnia”Condition prompt
Per engine
ChatGPTAbsent
PerplexityAbsent
ClaudeAbsent
ChatGPTAbsentPerplexityAbsentAI OverviewsAbsent
ClaudeAbsentNamed in the answer
Mayo Clinic guidance
NIH recommendationsWho decides itmayoclinic.orgClinical bodies answer the condition. Brands come later
Condition prompts are answered from clinical literature. No brand is named unless a trusted body has already named it.
The clinical math
Health answers name sources before brands.
How AI engines answer health questions, across 290 tracked healthcare domains.
82%of health answers carry a see-a-doctor caveat
AI hedges harder here than in any other category.
6.3×clinical bodies out-cite health brands
Mayo, NHS, NIH and PubMed carry the answer.
64%of health brands have no author or reviewer credentials
No named clinician means no citation.
4-10 wksfrom shipped fix to measurable movement
Slowest of any vertical, because clinical sources move slowly.
Prefer benchmark · 290 healthcare domains · 6 engines · Q2 2026.
The clinical bar · 01
Trust is earned before it is given.
Health is the strictest category AI handles. It answers from clinical bodies, names a brand only when a trusted source already did, and attaches a caveat almost every time. Marketing language actively lowers your odds.
“best online sleep clinic”
ChatGPT, answered this morning.
For chronic insomnia, cognitive behavioural therapy is the recommended first-line treatment. Programmes with published clinical evidence include Sleepio, which has randomised trial data, and Somryst, which is FDA-cleared. Availability depends on your insurer and state. Speak to your doctor before starting any treatment programme.
Sourcesmayoclinic.org
sleepio.comnih.gov
2services named1caveat attached0times Corva appears
Who decides · 02
Clinical bodies write the answer.
Health citations concentrate in medical institutions, government health services and peer-reviewed literature. Brand pages appear only when they carry named clinicians and published evidence.
Where citations come from
One tracked category, 230 prompts, last 30 days.
- Clinical institutions
- 34%
- Government health
- 21%
- Peer-reviewed
- 16%
- Health publishers
- 14%
- Directories & reviews
- 9%
- Your owned pages
- 6%
DomainTypeCitation share
1mayoclinic.orgTreatment named, no provider listClinical15.2%
2nih.govGovernment12.7%
3
nhs.ukGovernment9.8%
4
healthline.comNot in 2 relevant roundupsPublisher8.4%
5
pubmed.ncbi.nlm.nih.govYour trial not indexedPeer-reviewed7.1%
6
zocdoc.comProfile missing insurersDirectory5.6%
7
reddit.comr/insomnia threadsCommunity4.9%
8
corvahealth.comNo reviewed-by credentialsYour domain2.4%
Domains cited across 230 tracked health prompts, last 30 days. Clinical citations are earned with evidence, not content.
The prompt map · 03
Six families decide a care decision.
Two of them do not exist in other verticals: safety and access. Both are answered from registries and directories you may not have updated in years.
Corva Health, per family per engine
Status of the best prompt in each family.
CitedNamedAbsent
FamilyExample promptChatGPTPerplexityAIOClaudeWhat wins it
Conditionbest treatment for chronic insomniaPublished evidence a clinical body will cite
Providerbest online sleep clinicPublisher roundups + named clinicians
CostCBT-I cost without insuranceA published price range, not a quote form
Safetyis Corva Health legitimateCorrect registry data + accreditation page
ComparisonSleepio vs Corva HealthYour own evidence-led comparison pages
Accesssleep specialist taking Aetna in TexasStructured insurer, state and wait-time data
Live from the Corva demo workspace. Safety and access prompts run daily, because stale registry data compounds.
The playbook · 04
Six plays win health categories.
Each play maps to the Prefer tool that runs it, so this is a queue, not a PDF.
Publish your evidence
Outcomes, trial data and methodology in readable HTML, not a case-study PDF. AI names services whose results it can check.
Create Content →Fix your registry data
NPI, state licences, accreditation and directory profiles. Access and safety prompts are answered from these records, not from your site.
Action Center →Publish real prices
A range beats a quote form. Cost prompts need a number, and today a third-party estimate answers them for you.
Optimise Content →Earn the health roundups
Healthline, Verywell and condition-specific guides decide which services get named. Prefer finds the roundups you are missing and briefs the pitch.
Action Center →Watch the caveat
Track the warning AI attaches to your name and where it comes from. In health, one unresolved complaint outweighs ten good reviews.
Answer Engine Insights →Provider vs digital health · 05
Clinics and platforms lose different answers.
A local provider fights on access and insurance. A digital health platform fights on evidence and safety. Prefer ships a prompt set for each.
Providers and clinics
Local, licensed, insurance-bound. Where you are exposed:
Access“takes Aetna, near me, same week”
Specialty“sleep specialist, paediatric X”
Cost“with insurance, self-pay price”
Reputation“is X good, X reviews, wait times”
Exposure: a directory profile is out of date, so AI recommends a competitor.
Digital health
National, evidence-led, under scrutiny. Where you are exposed:
Condition“best treatment for X”
Evidence“does X work, X clinical trial”
Safety“is X legitimate, is X FDA cleared”
Comparison“X vs Y, X alternatives”
Exposure: no published evidence, so AI names a rival with a trial behind it.
Proof · 06
Evidence published. Named without hedging.
“We had the outcomes data the whole time. It just lived in a PDF nobody could read. We published it with our clinicians' names on it, fixed our registry entries, and AI started recommending us instead of skipping us.”
0 → 27%mention share on provider prompts6% → 88%pages with reviewer credentials10 weeksto first measurable move+41%assisted bookings from AI referrals
Questions
Asked plainly.
How do healthcare brands get recommended by AI assistants?
By being clinically verifiable. AI answers health questions from medical institutions, government health services and peer-reviewed literature, then names services whose clinicians, licences and published evidence it can confirm. Marketing claims without a named reviewer are ignored.
Why is AI visibility harder in healthcare?
Because health sits at the strictest end of YMYL, your money or your life. AI hedges on almost every health answer, cites clinical bodies six times more often than brands, and will not recommend a service it cannot verify is licensed and legitimate.
What is E-E-A-T for health content?
Experience, expertise, authoritativeness and trust. In practice it means named authors with real credentials, a medical reviewer on every clinical page, citations to primary literature, and a visible update date. AI uses the same signals when deciding what to quote.
Which prompts should a healthcare brand track?
Six families: condition, provider, cost, safety, comparison and access. Safety and access are unique to healthcare and are answered from registries and directories rather than from your website.
Which sources do AI engines cite for health answers?
Clinical institutions first, then government health services and peer-reviewed literature. In Prefer's Q2 2026 benchmark of 290 healthcare domains, clinical institutions took 34% of citations, government health 21%, peer-reviewed 16%, and brands' own pages 6%.
Is AI-generated health content safe to publish?
Only with clinical review. Prefer drafts from your own approved sources, attaches a citation to every claim, refuses to write a number it cannot source, and routes every draft through your medical reviewer before publish.
How do we handle HIPAA and patient privacy?
Prefer reads public answers and public sources only. It never touches patient data, and nothing in the workflow requires access to your EHR or any protected health information.
How long until health AI answers change?
Four to ten weeks, the slowest of any vertical, because clinical and government sources are re-crawled infrequently. Registry and credential fixes on your own pages move fastest.
Get your free AI visibility report
in about 10 minutes.
See how answer engines describe your brand today, and where the openings are to outpace the competition.