# The Fertility AI Answer Gap

Field Study · By Sandeep Kumar · Published 2026-11-05

**Quick answer:** We ran 60 real patient questions about fertility and IVF through ChatGPT. **85% of the answers (51 of 60) cited zero clinics.** ChatGPT answered the question — cost, success rates, genetics, insurance — but never pointed the patient to a clinic. When it did cite a clinic, the clinic's own site was the source in only 8.4% of all citations. The provider market is mostly invisible in AI answers, and that invisibility is concentrated in a handful of question clusters a clinic page can own.

## TL;DR

- **60 questions, one run each,** collected from real observed patient searches (Reddit, Quora, clinic Q&A, People Also Ask, autocomplete), US-focused, run through ChatGPT via the Cloro API on 2026-11-05
- **Ghost rate: 85%** — 51 of 60 answers contained no clinic citation at all. 9 answers (15%) cited at least one clinic
- **The gap is concentrated.** Genetics, diagnosis, donor egg, and risks clusters are 100% ghost. Cost is 71% ghost, success rate 86%, clinic choice 78%
- **Source economics:** of 190 total citations, 51.1% came from clinical/educational sources (CDC, SART, ASRM, academic health), 34.2% news/other, 6.3% aggregators, and **8.4% from clinic websites**
- **The read:** when a patient asks ChatGPT about IVF, the answer is built almost entirely from non-clinic sources. A clinic page that answers the exact question, in patient language, with citable data, is the missing ingredient in that answer
- **This is an AI-visibility measurement, not a clinical or quality judgment about any clinic**

## How we ran it

We built a battery of 60 patient questions on fertility and IVF — 30 verbatim strings observed in real patient searches, 30 paraphrased from observed fragments — across 11 intents: cost (7 questions), procedure (10), success rate (7), insurance (5), genetics (4), diagnosis (3), donor egg (2), risks (3), clinic choice (9), age-specific (8), LGBTQ+/single (2). Each was run once through ChatGPT (Cloro API, `taskType CHATGPT`, country US) on 2026-11-05, at a cost of 300 credits. The API did not disclose a model version, so we report the engine as "ChatGPT via Cloro API, Nov 2026."

**Definitions.** A *citation* is a source URL or citation pill returned with the answer. A *clinic citation* is either a clinic-site domain cited, or a clinic entity named in the answer. A *ghost answer* has zero clinic citations. *Opportunity Index* = demand weight (from provenance volume: high 3, medium 2, low 1) × ghost rate — it ranks where patient demand meets clinic-less answers.

**Consistency check.** Manual spot-check of all 9 clinic-citing answers for conflicting figures about the same clinic within one answer: none found (e.g., q08 lists different prices for different clinics — Radiance at $6,500 advertised vs. Pittsburgh Fertility "call for pricing" — not a contradiction).

**One honest limitation.** One run per question, one engine, one day. ChatGPT answers vary run to run, so shares carry noise; other engines may behave differently. Findings are correlational, not predictions.

## The ghost rate

| Metric | Value |
| --- | ---: |
| Questions run | 60 |
| Answers with zero clinic citation (ghost) | 51 (85%) |
| Answers with ≥1 clinic citation | 9 (15%) |
| Total citations across all answers | 190 |
| Clinic-site citations | 16, across 12 distinct clinics |
| Clinic entities named in answer text | 6, across 2 questions |

When a patient asks about cost, success rates, or genetics, ChatGPT produces fluent, useful general information — and the local provider market is invisible. For a clinic, each ghost answer is an unclaimed AI-visibility slot.

## Four clusters are 100% ghost

Opportunity Index by cluster (demand weight × ghost rate):

| Cluster | Questions | Ghost rate | Opportunity Index |
| --- | ---: | ---: | ---: |
| Genetics (PGT-A, testing) | 4 | 100% | 2.25 |
| Diagnosis (AMH, FSH, testing) | 3 | 100% | 2.00 |
| Donor egg | 2 | 100% | 2.00 |
| Success rate | 7 | 86% | 1.84 |
| Cost | 7 | 71% | 1.73 |
| Procedure | 10 | 90% | 1.71 |
| Risks | 3 | 100% | 1.67 |
| Insurance | 5 | 80% | 1.60 |
| Clinic choice | 9 | 78% | 1.56 |
| Age-specific | 8 | 88% | 0.88 |
| LGBTQ+/single | 2 | 50% | 0.50 |

Genetics is the sharpest example: PGT-A questions — cost, controversy, whether it's worth it — got **zero** clinic citations. A clinic page that answers "Should I do PGT-A?" with its own testing protocol, pricing, and age-banded data is currently the only way a clinic appears in that answer.

## The 12 clinic sites that got cited

Every clinic-site citation across the 60 answers — one row per site, with the questions it appeared in:

| Clinic site | Domain | Answers | Citations |
| --- | --- | ---: | ---: |
| Fertility Centers of New York | fertilityny.org | 2 (q30, q44) | 3 |
| Radiance Fertility | radiancefertility.com | 1 (q08) | 2 |
| Shady Grove Fertility | shadygrovefertility.com | 1 (q70) | 2 |
| CCRM | ccrmivf.com | 1 (q04) | 1 |
| Pittsburgh Fertility | pittsburghfertility.com | 1 (q08) | 1 |
| IVF Florida | ivfflorida.com | 1 (q12) | 1 |
| NYC IVF | nycivf.com | 1 (q12) | 1 |
| Rejoice Fertility | rejoicefertility.com | 1 (q19) | 1 |
| Cambridge IVF (NHS UK) | cambridge-ivf.nhs.uk | 1 (q30) | 1 |
| Women & Infants | fertility.womenandinfants.org | 1 (q36) | 1 |
| Florida Fertility Institute | floridafertility.com | 1 (q70) | 1 |
| Chelsea Fertility NYC | chelseafertilitynyc.com | 1 (q71) | 1 |
| **Total** | **12 sites** | **9 distinct** | **16** |

Clinic entities named inside answer text, without a citation pill: ETSU Health Fertility & Urogynecology, Rejoice Fertility, Southeastern Center for Fertility and Reproductive Surgery, The Reproductive Medicine Group, Shady Grove Fertility — Brandon, Shady Grove Fertility — Tampa Westshore.

*These are AI citation outputs at a point in time, not quality judgments.*

## Clinical bodies supply half of all citations

Share of all 190 cited sources:

| Source | Citations |
| --- | ---: |
| Clinical/educational (CDC, SART, ASRM, academic health) | 97 · 51.1% |
| News and other | 65 · 34.2% |
| Clinic websites | 16 · 8.4% |
| Aggregators | 12 · 6.3% |

_Bars share a 0–97 citation scale._

_Source: Pager Studio run, 60 answers, 2026-11-05._

85% of the cited material is clinical, educational, or general web. About 1 in 12 citations points to an actual clinic — even though the question is about choosing one.

## ChatGPT pulls from ASRM and ACOG, not clinics

Domains appearing most across the 190 citations: ASRM (32), ACOG (16), ReproductiveFacts (14), NHS (8), CDC (6), SART (6), HFEA (5), Mayo Clinic (5), Cleveland Clinic (5), PubMed (4), MedlinePlus (4), and then a long tail including ivfcost.co, KFF, fertilityny.org, sartcorsonline.com, thefertilityindex.com, Carrot, GoodRx, Tricare.

The top of the table is dominated by clinical bodies — the *educational* ecosystem. No clinic site appears among the eleven most-cited domains (4 or more citations each); the best any clinic does is fertilityny.org, tied with six other domains at 3 citations. None of the five most-cited domains overall — ASRM, ACOG, ReproductiveFacts, NHS, CDC — is a clinic site.

## What ChatGPT says instead of a clinic

Observed patterns from the stored answers:

- **Cost (q01):** quotes national ranges ($15k–$30k per cycle), gives a generic breakdown, names no clinic
- **Cost with a published price (q08):** one of only two cost answers that cited a clinic, and the one that quoted a price — "I found one clinic north of Pittsburgh advertising IVF starting at just $6,500" (Radiance Fertility) + Pittsburgh Fertility. Transparent published pricing is what gets cited
- **Insurance (q44):** describes ideal plan coverage features; no clinic is named because insurance answers are plan-level, not provider-level
- **Age (q71):** gives a general age-limit table ("Until 35… 35–39… 40–42…") with no clinic policies
- **LGBTQ+ (q70):** one of only two answers where ChatGPT named multiple clinics (with q19, a "near me" clinic-choice question) — Shady Grove Brandon/Tampa and Reproductive Medicine Group, real local entities the answer tied to a location (Wimauma, FL). Local relevance triggers clinic citations
- **Emotional/support (q34):** long empathetic framing, zero citations

## Ghost answers cite half as much

Sixteen answers came back with no external source at all — 27% of the run, generated with zero citations.

| Citations in answer | Answers |
| ---: | ---: |
| 0 | 16 |
| 1 | 0 |
| 2 | 12 |
| 3 | 9 |
| 4 | 7 |
| 5 | 5 |
| 6 | 2 |
| 7 | 5 |
| 8 | 1 |
| 9 | 1 |
| 10 | 1 |
| 11 | 0 |
| 12 | 1 |

_Citations per answer, all 60 answers._

_Source: Pager Studio run, 2026-11-05._

Non-ghost answers averaged 5.3 citations against 2.8 for ghost answers — when ChatGPT cites a clinic it also cites more broadly. Clinic presence tracks a richer, better-sourced answer.

## Four surprises in the data

1. **A UK NHS clinic appears in a US run** (Cambridge IVF, q30). Search-country signals are imperfect; international clinics can leak into US answers
2. **Location helps, but it is not a precondition:** all three "near me" questions in the battery cited a clinic — yet 6 of the 9 clinic-citing answers came from questions carrying no location at all (age limit, insurance, embryo transfer, PGT cost). Even a national question cites a clinic when a clinic page is the citable source
3. **Ghost answers are thinner:** fewer citations, less grounding. The gap is not just "no clinic" — it's lower overall answer quality, which is an even stronger argument for clinic pages adding grounded, citable data
4. **Aggregators are marginal (6.3%):** ChatGPT is not leaning on comparison sites; it's leaning on clinical bodies. So the fight for a citation is against ASRM/ACOG content, not against FertilitySync

## Downloads

Snapshot files from the 2026-11-05 run, read-only copies of the stored answers:

- [Per-question analysis (CSV)](https://www.pagerstudio.space/study-fertility-ai-answer-gap/analysis.csv) (6.2 KB) — All 60 per-question rows: ghost flag, intent, citation counts, source domains, clinic-site domains, entities named
- [Question provenance (JSON)](https://www.pagerstudio.space/study-fertility-ai-answer-gap/battery_60.json) (11.6 KB) — Every question mapped to its original observed string, source, volume, and intent
- [Raw answers (ZIP)](https://www.pagerstudio.space/study-fertility-ai-answer-gap/raw-answers-2026-11-05.zip) (140 KB) — All 60 raw ChatGPT answers from the 2026-11-05 run

Data comes from public search and forum surfaces; no users are identified.

## What this means for a clinic

In every ghost answer, ChatGPT answers the *question* but not the *referral*. The missing ingredient is a clinic page that:

- answers the exact patient question in patient language, not brochure-speak
- publishes the specific data AI can cite — costs, success rates, testing protocols, age-banded outcomes
- is structured so AI can attribute it: clear entities, clinic name, contact, schema

One clear signal from the data: in question 08, the clinic with a transparent published price — Radiance at $6,500 — got cited by name. Clinics that publish the number get named; clinics that hide it don't exist in the answer.

The gap is real, concentrated in a few clusters, and addressable with the right page. That is what we do: [IVF & fertility AEO](https://www.pagerstudio.space/ivf-fertility-aeo).

## Frequently Asked Questions

### Why did you only test ChatGPT?

Because it is the engine patients name most, and because one engine with a transparent, reproducible run is a cleaner study than five engines with mixed methodology. Other engines may behave differently; that is a stated limitation, not an oversight.

### Doesn't 85% ghost mean ChatGPT is just bad at this?

No. ChatGPT answers the questions well — the content is useful and accurate. What it fails to do is *refer* to a clinic. That distinction is the whole point: the patient gets help, and the clinic gets nothing. The referral gap is an opportunity, not an indictment of the engine.

### Are the clinics cited the "best" clinics?

No. The study measures visibility, not quality. A named clinic appears because its site or entity data was citable for that specific question. A clinic that is absent can be excellent — it is simply not in the AI answer. This is a marketing measurement, not a medical one.

### How can a clinic appear in these answers?

By being the page that answers the ghosted question with citable specifics: transparent prices, protocol detail, age-banded success rates, testing guidance. The full mechanism is in [what AI reads for fertility](https://www.pagerstudio.space/blog/sources-ai-reads-for-fertility-what-clinics-need-to-know) and [which clinics ChatGPT recommends](https://www.pagerstudio.space/blog/which-fertility-clinics-does-chatgpt-recommend).

### Was your own data run through this study?

No. The study is a clean, independent corpus of patient questions; no Pager Studio client data or site was included. It is published openly for any clinic to audit.

## Methodology and limitations

Every figure in this study traces to the stored answers: `battery_60.json` (question provenance), `analysis.csv` (per-question rows), and `raw-answers-2026-11-05.zip` (60 raw answers). All are available in the Downloads section above.

**Sample.** 60 patient questions on fertility/IVF, all tracing to real observed patient questions: 30 verbatim-observed strings (Reddit / Quora / clinic Q&A / People-Also-Ask / autocomplete), 30 paraphrased from observed fragments. US-focused; India cluster excluded. Intents covered (11 clusters): cost (7), procedure (10), success_rate (7), insurance (5), genetics (4), diagnosis (3), donor_egg (2), risks (3), clinic_choice (9), age_specific (8), lgbtq_single (2).

**Engine and run.** ChatGPT via the Cloro API (`taskType CHATGPT`), country=US, one run per question (60 runs), 300 credits. Data collected 2026-11-05. The API's `model` field was null at collection time, so the engine is reported as "ChatGPT via Cloro API, Nov 2026"; model version not disclosed by the API. No re-runs; no variance absorption (single run per question — disclosed limitation).

**Definitions.** Citation = a source URL/label or citation pill returned with the answer (domain-level pills). Clinic citation = a cited source whose domain is a clinic site, OR a clinic entity named in the answer (from the API's entity extraction). Ghost answer = an answer with zero clinic citations. Source taxonomy = clinic-site / cdc-sart-clinical (CDC, SART, ASRM patient education, academic health systems, medical publishers) / aggregator (comparison & index sites) / community (Reddit, Quora) / news-other. Opportunity Index (per intent cluster) = demand weight (high=3, medium=2, low=1) × ghost rate.

**Limitations.** Single platform, single point in time (ChatGPT, Nov 2026, US). Findings are correlational, not predictions; other engines may behave differently. One run per question; ChatGPT answers vary, so shares carry run-to-run noise. We evaluate AI visibility — whether and which clinics are named/cited. This is not a clinical or quality judgment about any clinic, and it is not medical advice.

**Disclosure.** Pager Studio is a commercial service; the fertility clinics we work with pay for AEO deployments. We ran this corpus for our own product development and publish it openly so any clinic can audit its own AI presence. ChatGPT is a trademark of OpenAI; all other marks belong to their respective owners and are used for identification only.

---

Want to see where your clinic lands in these answers? [Claim Free Snapshot](https://www.pagerstudio.space/book) — I'll check how AI currently describes your practice before anyone talks retainers, or [see how the audit works](https://www.pagerstudio.space/pricing) first.

The gap is real, concentrated, and addressable. **Answer the question, get the referral.**
---

Site: https://www.pagerstudio.space
Full site content for AI systems: https://www.pagerstudio.space/llms-full.txt
Site index: https://www.pagerstudio.space/llms.txt
Sitemap: https://www.pagerstudio.space/sitemap.xml
Contact: hello@pagerstudio.space — https://www.pagerstudio.space/contact