When Someone Asks AI About Fertility, These Seven Sources Write Your Answer

Quick answer: When a patient asks ChatGPT, Gemini, Claude or Perplexity a fertility question, the answer is assembled from a small, repeatable set of sources — and almost none of them belong to your clinic. Published citation research says health answers lean on institutional sources (Mayo Clinic, Cleveland Clinic, Wikipedia, NHS, PubMed — over 75% of citations in one 2026 study of ChatGPT health answers), registry data (SART and CDC), and a layer of directories and advocacy sites. Your own website is a minor character in that story. You can't write any of those sources. You can only make sure they agree about you — because when they don't, engines don't pick a side. They leave you out.

• AI health answers are built from a narrow source pool. In a 5,472-citation study across ChatGPT, Gemini, Claude and Perplexity, 30.7% of citations came from health media, 23.1% from commercial and affiliate sites, and 22.9% from academic sources — with 62.4% of all citations coming from domains rated DR 81–100
• For fertility specifically, the load-bearing sources are: SART and CDC ART (the numbers), ASRM / ReproductiveFacts.org (the definitions), PubMed Central (the literature), institutional health media like Mayo and Cleveland Clinic (the explainers), FertilityIQ and review directories (the entity-and-reputation layer), and RESOLVE (coverage, support and options)
• The federal Fertility Success Rates Act is why clinic-level fertility data exists in the first place — every US ART program must report success rates annually. AI can read that data whether or not your website matches it
• Each engine leans differently: ChatGPT favors health media, Gemini favors government and NGO sources, Claude balances research, Perplexity leans commercial and user-generated. Same clinic, four different source shelves
• You cannot insert yourself into SART or CDC. You can only make your numbers, credentials and descriptions agree — because when sources disagree, engines omit rather than adjudicate
• No strategy here requires naming or beating a competitor. None of it is a hack. It's consistency with sources you don't control

How we know what engines read

Two published analyses anchor this map, and we're naming them because the claims should be checkable:

1. "AI's Sources of Truth: How Chatbots Cite Health Information" (July 2025) — 5,472 citations collected from ChatGPT (GPT-4o browsing), Gemini 2.5 Flash, Claude Sonnet 4 and Perplexity (Sonar), across health prompts. Findings used here: source-type shares, domain-authority distribution, freshness, and per-engine leanings.
2. "Authority Signals in AI Cited Health Sources" (arXiv:2601.17109, January 2026) — 100 questions drawn from HealthSearchQA (a 3,173-question consumer health set curated by Google Research), run through ChatGPT 5.2 Pro; 615 cited sources coded. Finding used here: over 75% of citations came from established institutional sources.

Fertility-specific source profiles below (SART, CDC, ASRM, FertilityIQ, RESOLVE) come from those organizations' own live pages, checked October 2026.

Layer 1: The registries — SART and CDC

This is the layer that owns your numbers, and it predates AI entirely.

The Fertility Success Rates and Clinic Compliance Act requires every US ART program to report its success rates annually. That reporting produces two public faces:

• CDC ART (cdc.gov/art) — the federal report. Covers all reporting US clinics, so it's the more complete dataset. It presents cumulative success rates, including embryo transfers occurring within a year after an egg retrieval, broken out by patient age group — with its own plain-language caution that "average chances may not reflect actual success rates during ART treatment for an individual or couple."
• SART (sart.org and its public reports) — the professional society's registry, covering member clinics, with clinic-level multi-year reports, a patient success-rate predictor, and patient-facing explainers. SART's own guidance page spells out the three variables that move any cycle outcome — endometrium, embryo and transfer difficulty — and then warns, in its own words, that "using SART statistics to compare clinics may not influence a patient's personal chance for success."

Two things engines take from this layer: the rates themselves (live birth per retrieval vs. live birth per transfer are different denominators, and both appear in the public data) and the framing caveats, which get repeated in answers whenever a patient asks "is this clinic good." A clinic whose website leads with a differently-defined success number than its registry entry hands an engine a conflict it can't resolve. Engines don't split the difference on conflicts like that. They just leave you out — and you never see the omission happen.

What you control here: nothing in the registries. Everything about whether your site agrees with them.

Layer 2: The societies — ASRM and ReproductiveFacts.org

ASRM (American Society for Reproductive Medicine) is the professional society; its patient-education site, ReproductiveFacts.org, is where definitions live — fact sheets and booklets on topics like age and fertility, stress and infertility, and what ART actually involves, plus patient videos. SART is ASRM's affiliated society, and their SART Fertility Experts podcast series is hosted on the same patient site, so this layer interlocks.

When a patient asks "what is unexplained infertility" or "how many embryos should be transferred," engines reach for definitional, society-backed pages. This layer answers the informational questions — the ones your clinic's service pages were never shaped to win.

What you control here: nothing. But quoting and aligning your own definitions with this layer's vocabulary is free.

Layer 3: The literature — PubMed Central

In the 5,472-citation study, the single most-cited domain was PubMed Central — 385 citations, still less than 0.1% of the total, which tells you how fragmented the citation pool is. The same study found chatbots rely more on summaries and interpreters of science (59%) than on peer-reviewed research directly (41%). In fertility, that means the research layer usually enters answers through a middleman: a health-media explainer or a society page citing the study, not the paper itself.

What you control here: if you publish or are cited in peer-reviewed fertility research, that's an authority signal no marketing spend replicates. If you don't, this layer is background — you compete for the summary layer instead.

Layer 4: Institutional health media — Mayo, Cleveland Clinic, Wikipedia

The arXiv authority-signals study found over 75% of ChatGPT's health citations came from established institutional sources — Mayo Clinic, Cleveland Clinic, Wikipedia, the NHS and PubMed leading the list. In the appendix, Mayo Clinic alone drew 61 citations at a domain authority of 92. The Outcomes Rocket study independently found institutional health media "play a prominent role."

This is the explainer layer: "what is IVF," "IVF success rates by age," "how much does IVF cost." These queries have enormous volume, and the institutions own them because their pages are open-access, recently updated, medically reviewed, and structured as direct answers — the exact attributes the citation studies associate with being picked.

What you control here: you don't get cited from this layer as a clinic. You do compete with it for adjacent questions no institution has answered specifically — your city, your patient population, your financing structures.

Layer 5: Directories and review platforms — FertilityIQ, Healthgrades, WebMD

This is the entity layer, and it's the only place your clinic appears as a fact instead of a pitch.

FertilityIQ is the fertility-specific one: doctor and clinic profiles with 1–10 ratings, "strongly recommend / neutral / don't recommend" splits, review counts, and category-level scoring covering billing, nursing, scheduling and an attention-to-detail score based on patient-reported errors. It also runs education courses featuring faculty from named academic medical centers — its IVF course alone shows 709k views, with its PGT-A course at 73k. When an engine is asked which fertility doctors are worth considering, structured profiles like these are exactly the kind of attributable, specific-claim content it can lift.

The general health directories show up in the same role: in the arXiv ChatGPT study, WebMD drew 13 citations at DA 94 and Healthgrades 2 at DA 73. Modest counts — but they're entity records about specific practices, which is a different function than the explainer layers above.

What you control here: claim and keep accurate every profile that exists about you — name, physicians, locations, insurance, services. It's the one layer where you show up as a first-class entity, and a conflicting or half-filled profile is the quietest reason an engine decides you're not worth naming.

Layer 6: RESOLVE — coverage, support and options

RESOLVE: The National Infertility Association (resolve.org) owns a distinct question set: insurance coverage ("Get Coverage at Work"), financial resources, support groups, and family-building options — its advocacy page notes it has defeated 50 anti-IVF bills since the overturn of Roe v. Wade.

Ask an AI "does insurance cover IVF" or "how do I afford fertility treatment" and RESOLVE-style sources dominate, because they're the only institutional players in that question set. Notably, this is also a layer where patient forums and Reddit threads surface — Perplexity in particular leans commercial and user-generated relative to the other engines.

What you control here: very little directly. But if your clinic's financing, insurance and benefits information contradicts what RESOLVE-type sources describe, you lose the coverage conversation.

The map at a glance

Each engine reads a different shelf

From the citation study's per-engine analysis: ChatGPT draws heavily from health media. Gemini focuses on government and NGO sources — which, for fertility, means CDC and SART get disproportionate weight there. Claude approaches research parity between summaries and studies. Perplexity favors commercial and user-generated content, averaging the most citations per answer (14.97, vs. Claude 13.99, ChatGPT 13.59, Gemini 12.29).

The practical read: a clinic with clean registry data and institutional placements reads strongest in ChatGPT and Gemini; one with strong directory and review coverage reads strongest in Perplexity. Most agencies sell "AI visibility" as one channel. It's four — with four different reading lists.

What to do this week

• Reconcile your site against your registry entry. Pull your clinic's CDC ART and SART pages and compare the definitions your website uses — live birth per transfer vs. per retrieval, age bands, what's counted. If they disagree, fix the site. Every time. This is the single highest-value step in fertility AEO, and it costs nothing but an afternoon
• Claim every FertilityIQ and Healthgrades profile that exists about you. Incomplete or conflicting entity records are the quietest reason clinics don't get named. Our AI visibility walkthrough has the manual test
• Write your informational pages in definitional shape — the direct answer in the first two sentences, aligned with ASRM vocabulary. That's the form the explainer layers use, and it's what engines lift. What AEO means for healthcare practices covers the mechanics
• Cover the coverage question honestly. Insurance and financing queries route to RESOLVE-type sources; a clear, current page on what your clinic accepts and how financing works keeps you in that conversation instead of contradicting it
• Test across all four engines, not just one. How AI search differs from Google explains why the shelf differs by engine — and which health AI assistants can actually surface a clinic covers who can name one and who can't

One honest caveat

This map is built from two published citation studies (2025 and January 2026) and the live structure of each named source, checked October 2026. Citation pools shift as engines update — the studies themselves note findings are snapshots. Fertility-specific peer-reviewed research on AI answers is still thin, and one 2024 evaluation concluded AI was not yet reliably able to give evidence-based fertility recommendations, particularly around embryo transfer. Nothing here is a claim about any individual clinic's quality or visibility, and none of this is medical advice — it's a map of where machine answers come from, which is a marketing question, not a clinical one.

Frequently Asked Questions

Can a fertility clinic get cited directly by AI engines?

Sometimes, mostly in the entity layer (directories, review platforms, branded queries) — rarely in the explainer or registry layers, which are dominated by societies, registries and institutional media. The realistic goal for most clinics isn't becoming a cited source; it's being named accurately inside answers assembled from the sources above.

Does having good SART data make AI recommend my clinic?

It makes you quotable where it matters. Registry data feeds the comparisons engines run, and clean, consistent data removes the main reason an engine omits a clinic. But the registry is one input among seven layers — coverage, reviews, definitions and directory records decide whether you get named at all.

Why do some clinics get named by ChatGPT but not Gemini?

Different shelves. ChatGPT leans health media; Gemini leans government and NGO sources, which weights registry data more heavily for fertility. Directory and review coverage moves Perplexity more than it moves Gemini. Our breakdown of the health AI assistants covers which engines can surface clinics at all.

Is FertilityIQ actually read by AI engines?

FertilityIQ's value is structural: physician and clinic profiles with specific, attributable claims — ratings, recommend rates, review counts — in a crawlable form, which is the content shape citation studies associate with being lifted. It's one entity record among several; the discipline is claiming it and keeping it accurate alongside Healthgrades, your Google Business Profile and your own site.

What about Reddit and patient forums?

Real presence in the citation pool, especially for Perplexity, which leans commercial and user-generated. Forums surface for experiential questions — cost stories, treatment experiences, coverage battles. Clinics can't manage forum content; they can make sure official sources don't contradict it.

Does our own website matter at all?

Yes — for branded queries, hyper-specific questions no institution has covered, and as the consistency check engines run against everything above. But the citation studies are clear that institutional sources dominate the general health question set, and a clinic site is not going to out-authority a registry or a society. Win the questions nobody else has answered, and keep your facts aligned with the sources that do.

Curious where your clinic stands? Claim Free Snapshot — I'll check how AI currently describes your practice and which of these seven layers it's reading for you, before anyone talks retainers, or see how the audit works first.

The clinics AI describes accurately are the ones whose sources already agree. Make yours agree.