The 100 Most-Cited Websites in ChatGPT for High-Stakes Health and Medical Consumer Decisions

LLM Authority Index Research | ChatGPT | Health and Medical | July-September 2026

ChatGPT does not use the same citation source market for high-stakes health and medical consumer decisions that it uses across the broader web.

Across 1,116 deduplicated ChatGPT responses covering 14 health and medical decision verticals, Healthline ranked #1, appearing as a cited domain in 21.1% of eligible responses. Forbes ranked #2 at 14.8%, Reddit #3 at 14.0%, HearingTracker #4 at 7.4%, and Consumer Reports #5 at 7.0%.

The more important finding is structural. Only 28 domains in this health-specific Top 100 also appear in the previously published broad ChatGPT Top 100. The source market changes substantially when the prompt population is restricted to decisions such as hearing aids, online therapy, online pharmacies, medical alert systems, fertility clinics, online doctors, dental implants and addiction treatment.

This is a citation visibility study, not a clinical-quality ranking. A domain's presence means it appeared in captured ChatGPT citations for the studied prompt population. It does not mean ChatGPT endorsed the source, that the source is medically superior, or that citation caused a recommendation.

Answer Capsule

Which websites does ChatGPT cite most often for high-stakes health and medical consumer decisions?

Healthline ranks #1 at 21.1% Response Citation Coverage, followed by Forbes at 14.8%, Reddit at 14.0%, HearingTracker at 7.4%, and Consumer Reports at 7.0%. The category Top 100 account for 63.5% of all domain-response appearances in the studied ChatGPT health corpus. Only 45 domains remained in the monthly Top 100 in July, August and September, and the July-to-September Top 100 Jaccard similarity was only 32.5%, indicating substantial source-list movement below the most persistent leaders.

ChatGPT's health citation market is also highly platform-specific. The final Health child-study Top 100 lists show ChatGPT sharing 44 domains with Perplexity, 41 with Google AI Mode, 39 with Google AI Overviews, 39 with Gemini and 32 with Microsoft Copilot. By contrast, the two Google Health Top 100 lists share 76 domains with each other.

Key Findings

• Healthline is the leading ChatGPT health citation domain, appearing in 236 of 1,116 eligible responses, or 21.1%.
• Forbes ranks #2 at 14.8% and Reddit ranks #3 at 14.0%.
• Specialist source HearingTracker ranks #4 overall despite appearing in only one source vertical, driven by 55.7% response coverage in Hearing Aids.
• The health-specific Top 100 capture 63.5% of domain-response appearances, compared with about 58.8% in a method-consistent broad ChatGPT baseline.
• The head of the category is flatter than broad ChatGPT: the health Top 10 capture 24.8% of domain-response appearances versus about 32.1% in broad ChatGPT.
• Only 28 domains overlap between the health-specific and broad ChatGPT Top 100 lists.
• Only 45 domains persisted in the ChatGPT health Top 100 in all three months.
• July and September shared just 49 Top 100 domains, a Jaccard similarity of 32.5%.
• ChatGPT health shares 44 Top 100 domains with Perplexity, 41 with Google AI Mode, 39 with Google AI Overviews, 39 with Gemini and 32 with Microsoft Copilot.
• Healthline leads Online Doctors at 44.0%, Online Therapy at 64.1%, STD Tests at 52.5%, and Weight Loss and Metabolic Health at 28.3%.
• HearingTracker leads Hearing Aids at 55.7%.
• Forbes leads Medical Alert Systems at 46.1%.
• Amazon leads Online Pharmacies at 18.7%.
• Yelp leads the overlapping Fertility Clinics and IVF Clinics source labels at 26.9% in this captured ChatGPT slice.
• The public ranking excludes 318 technical infrastructure entries: 198 images.openai.com entries, 88 mapbox.com entries and 32 openstreetmap.org entries.

Questions This Study Answers

• Which websites dominate ChatGPT citations for high-stakes health and medical decisions, and how does that source hierarchy change by vertical?
• How persistent is the ChatGPT health source market over time, and how portable is that authority to other AI platforms?
• What should health publishers, healthcare brands and CMOs measure before allocating resources to ChatGPT-focused content, PR, partnerships or earned media?

Why This Research Belongs in Daily AI Search Marketing Decisions

This study is research-based, but its purpose is commercial decision support.

This Article sits at the intersection of two useful views of the same source market:

• the ChatGPT platform view, which asks which sources recur across high-stakes consumer decisions on ChatGPT; and
• the Health and Medical decision-family view, which asks which sources recur across six AI platform families for one high-stakes consumer category.

For health publishers, specialist sites, providers and marketplaces, that intersection creates a more precise claim than saying a domain is simply "visible in ChatGPT." A source can document whether it is visible specifically inside ChatGPT for defined health and medical consumer decisions, whether that visibility persists over time, and whether its authority is broad across many health verticals or concentrated in one specialist decision journey.

For healthcare brands, providers, CMOs, PR teams and agencies, the operating question is:

Which sources are already present around the exact health and medical decisions our customers ask ChatGPT to help them make?

Four findings matter for planning:

• the Top 100 account for 63.5% of response-level domain appearances in this ChatGPT health and medical slice, compared with about 58.8% across ChatGPT's broader high-stakes source market;
• only 28 domains overlap between this Health and Medical Top 100 and the broad ChatGPT Top 100;
• only 45 domains remained in the monthly ChatGPT Health Top 100 in July, August and September, and the median of the three monthly Top 100 Jaccard comparisons was 40.9%; and
• the final Health child-study Top 100 lists show ChatGPT sharing 44 domains with Perplexity, 41 with Google AI Mode, 39 with Google AI Overviews, 39 with Gemini and 32 with Microsoft Copilot.

The parent Health and Medical Citation Authority Study provides the category-level Persistence-Portability Gap interpretation. Under its reconciled final-platform methodology, it reports 61.3% temporal persistence, 25.8% cross-platform portability, and an approximate 35.5 percentage-point Persistence-Portability Gap for Health and Medical decisions.

That category gap is exploratory. Article 26 should not create a separate ChatGPT gap by subtracting its 40.9% monthly persistence from one platform-pair overlap. The statistics answer different questions.

This Article also shows why platform-specific monitoring matters even inside one decision family. ChatGPT Health is materially less persistent month to month than the two Google Health source markets, while its pairwise overlap with every other measured engine remains below 30% Jaccard similarity.

Current outside research supports the need to keep prompt population and source definitions attached to every health-source claim.

BrightEdge's August 31, 2026 healthcare comparison found that ChatGPT's five most-cited healthcare sources in its tracked set were government agencies or hospital systems, while Google AI Overviews elevated YouTube above every clinical source. BrightEdge healthcare authority study

Ahrefs' September 2026 broad U.S. ChatGPT study ranks Reddit first at 16.8% mention share, Wikipedia second at 7.0%, Consumer Reports third at 3.7%, and Forbes fourth at 3.1%. This Health-specific panel instead ranks Healthline #1, Forbes #2, Reddit #3, HearingTracker #4 and Consumer Reports #5. Ahrefs ChatGPT citation study

A 2026 medRxiv preprint analyzed 100 HealthSearchQA questions and 615 ChatGPT 5.2 Pro cited sources, reporting 75.7% of cited sources from established institutional source types. Its query population and source-class framework differ substantially from this commercial decision panel. medRxiv Authority Signals study

Tinuiti's Q3 2026 citation research spans multiple commercial categories, including OTC health, and continues to show strong category and platform effects in citation mix. Tinuiti Q3 2026 AI Citation Trends Report

Those studies should not be merged numerically with this one. The useful agreement is structural: ChatGPT citation behavior changes with query intent, source class, category, metric and denominator.

This article measures citation visibility in high-stakes consumer decisions. It does not measure clinical quality, medical accuracy, endorsement or whether a cited source caused ChatGPT to recommend a healthcare company or product.

Commercial Action Matrix: What Publishers and CMOs Can Do With the Findings

The daily operating model is therefore four-layered:

1. Current ChatGPT Health core: maintain the sources that matter now, with the collection date attached.
2. Vertical and decision-cluster layer: identify which sources dominate the exact hearing aid, therapy, pharmacy, medical alert, fertility or other consumer decision.
3. Longitudinal check: monitor whether priority sources persist, decline, rotate or change rank.
4. Cross-platform check: determine whether the same sources also matter in Google AI Overviews, Google AI Mode, Gemini, Perplexity and Microsoft Copilot.

How This Study Fits the High-Stakes Consumer Decisions Research Corpus

This Article is the intersection of two parent studies:

• The ChatGPT Platform Study
• The Health and Medical Citation Authority Study

It should also be read with:

• the 2026 AI Citation Authority Study;
• the Persistence-Portability Gap;
• Do AI Platforms Cite the Same Websites?;
• How Stable Are AI Citations?;
• Reddit's Decline in AI Citations for Health and Medical Decisions; and
• How Reddit Citations Changed in ChatGPT for Health and Medical Decisions.

What We Studied

This study isolates ChatGPT responses from the larger LLM Authority Index High-Stakes Consumer Decisions research corpus and restricts analysis to the Health and Medical family.

The 14 source vertical labels are:

1. Addiction Treatment Centers
2. Assisted Living Facilities
3. Dental Implants
4. Fertility Clinics
5. Hearing Aids
6. Home Health Care
7. IVF Clinics
8. Medical Alert Systems
9. Mental Health Treatment Centers
10. Online Doctors
11. Online Pharmacies
12. Online Therapy
13. STD Tests
14. Weight Loss and Metabolic Health

These are originating source labels from the research archive, not 14 statistically independent industries. Fertility Clinics and IVF Clinics overlap heavily. Some Home Health Care prompts are employment or franchise adjacent. Dental Implants has a particularly small ChatGPT response count in this slice and should be interpreted cautiously.

Study size

This section answers

How large is the ChatGPT health dataset, and what exactly is being ranked?

The primary unit is an eligible ChatGPT response. A domain can count at most once per response for the primary metric, regardless of how many URLs from that domain appear in the response.

The 25 Most-Cited Websites in ChatGPT for Health and Medical Decisions

The top of the table contains three different forms of source authority.

Healthline has category breadth and repeated strength across multiple consumer health verticals. Forbes and Reddit are broad cross-category sources that remain visible in this high-stakes health panel. HearingTracker is a specialist: it ranks #4 overall even though its category breadth is narrow because its Hearing Aids coverage is so high.

Consumer Reports is another useful hybrid. It has broad consumer authority, but its health visibility is disproportionately strong in decision-heavy product categories such as hearing aids.

The Full Top 100 ChatGPT Health and Medical Citation Sources

The full table is intentionally included because the commercially useful source market often sits below the first ten domains. A specialist publisher can rank outside the global top tier while still being one of the most important sources in a specific high-intent decision cluster.

ChatGPT Health Has a Flatter Head but a Shorter Tail Than Broad ChatGPT

The category behaves differently depending on where the concentration curve is measured.

The Top 10 are less dominant in health than they are in broad ChatGPT. But by the Top 100 cutoff, the health list captures more of the domain-response market.

That suggests a flatter head with a more bounded commercially relevant source set. A few universal domains do not monopolize the category, yet a finite group of health publishers, specialist reviewers, providers, government sources and consumer platforms account for much of the recurring citation visibility.

This is useful for outreach. The practical target universe may be larger than a simple Top 10, but it is still meaningfully smaller than the full set of 858 domains observed in the category.

Which Sources Gain the Most When the Analysis Is Restricted to Health and Medical Decisions?

Several domains become far more important in the health slice than in broad ChatGPT.

The reverse pattern matters just as much. Forbes appears in 14.8% of the health slice but 33.5% of broad ChatGPT responses. Reddit appears in 14.0% of health responses versus 17.1% broadly.

A domain can therefore be highly important to ChatGPT overall and still lose relative share when the prompt universe narrows to commercial health decisions. Conversely, a specialist can be nearly invisible in a broad platform ranking and become central inside one health market.

The Most-Cited Sources by Individual Health and Medical Vertical

The vertical splits are more actionable than the category-wide Top 100 for many commercial teams.

• Hearing Aids: HearingTracker reaches 55.7% response coverage, followed by Consumer Reports and Audiologists.org.
• Online Therapy: Healthline reaches 64.1%, followed by Forbes, Talkspace, Reddit and Psych Central.
• STD Tests: Healthline reaches 52.5%, with Medical News Today and Verywell Health next.
• Online Doctors: Healthline leads at 44.0%, followed by Forbes, Reddit, PlushCare and Sesame Care.
• Medical Alert Systems: Forbes leads at 46.1%, followed by SeniorLiving.org and ConsumerAffairs.
• Online Pharmacies: Amazon leads at 18.7%, followed by CVS and GoodRx.
• Fertility Clinics and IVF Clinics: Yelp leads at 26.9% in the captured slice. These two source labels overlap heavily and should not be treated as independent samples.
• Dental Implants: Aspen Dental leads, but the vertical has only 16 eligible ChatGPT responses in this slice, so rank differences are fragile.

This section answers

Can a company use the category-wide Top 100 as its PR target list?

Not by itself. The vertical table shows why. A domain can be modest at the family level and still be one of the most frequently cited sources for a specific buying or provider-selection decision.

Broad Publisher Authority and Specialist Authority Are Different Assets

The health data separates broad publishing reach from decision-specific authority.

Healthline is the strongest example of broad category authority. It leads the overall health slice and four individual decision verticals.

HearingTracker is the clearest specialist example. Its 7.4% category-wide coverage is generated primarily from one vertical, Hearing Aids, where it appears in 55.7% of eligible ChatGPT responses.

Forbes shows another pattern. It has broad platform authority and is still a major health source, but its health coverage is less than half its broad ChatGPT coverage. Its most visible health niche in this study is Medical Alert Systems, where it appears in 46.1% of responses.

For earned-media planning, these are different assets. Broad publishers can help across multiple prompt clusters. Specialist publishers can have disproportionate influence in a narrow commercial decision market.

ChatGPT Health Rankings Were Much Less Stable Than the Two Google Health Lists

Only 45 domains remained in the ChatGPT health Top 100 in all three months. The median of the three monthly Top 100 Jaccard comparisons was 40.9%.

The leading domains were more persistent than the lower ranks:

• Healthline moved #2 to #1 to #1.
• Forbes moved #3 to #2 to #2.
• Reddit moved #1 to #3 to #7.
• HearingTracker moved #6 to #4 to #4.
• Consumer Reports moved #9 to #5 to #3.
• GoodRx moved #5 to #6 to #27.
• Wikipedia moved #4 to #22 to #35.

The July-to-September Jaccard similarity was only 32.5%. That is substantially lower than the two Google health studies, which had a much larger shared core.

The monthly response counts also differ, with 267 eligible responses in July, 403 in August and 446 in September. The stability statistics describe the captured corpus and should not be interpreted as a controlled experiment with identical monthly sample sizes.

This aggregate Top 100 movement is separate from same-prompt citation drift. The citation drift study uses matched prompt-platform scenarios and is the better source for exact scenario-level turnover.

ChatGPT Health Sources Are Highly Platform-Specific

The final Health and Medical child-study Top 100 lists show that ChatGPT's source market differs substantially from every other measured engine.

No other Health platform shares even half of ChatGPT's Top 100 domains. Perplexity is the closest of the five sibling engines at 44 shared domains.

The two Google Health surfaces are a useful control. They share 76 domains with each other, while ChatGPT shares only 39 with Google AI Overviews and 41 with Google AI Mode.

That difference matters commercially. A publisher that performs well on Google, Gemini, Perplexity or Copilot cannot assume the same citation visibility in ChatGPT. A brand measuring third-party source opportunities should maintain a ChatGPT-specific source map rather than one blended "AI authority" list.

The source types also differ. ChatGPT's Top 10 includes Healthline, Forbes, Reddit, HearingTracker, Consumer Reports, GoodRx, Wikipedia, Reuters, Verywell Health and Medical News Today. Other platforms elevate different combinations of retailers, institutions, marketplaces, government sources and specialists.

The operating implication is to maintain three layers: a cross-platform health core, a ChatGPT-specific source layer, and a vertical-specific specialist layer.

How This Study Relates to Other Health and AI Search Research

BrightEdge: ChatGPT and Google use different healthcare source mixes

BrightEdge analyzed healthcare citations across ChatGPT, Google AI Mode and Google AI Overviews over 14 weeks from October 2025 through January 2026. It reported that 27% of ChatGPT healthcare citations came from government sources and only 1% from elite hospital systems in its source-type framework. (BrightEdge healthcare citation study)

That is not directly comparable to this ranking. BrightEdge groups sources by type, while LLMAI ranks registrable domains by the share of eligible commercial responses that cite them. A government-source share can be substantial even when no single government domain ranks #1.

The direction is nevertheless relevant. FDA and NIH both appear prominently in LLMAI's ChatGPT health source set, while specialist publishers and consumer-health domains also occupy major positions.

Ahrefs: broad ChatGPT is dominated by a different source market

Ahrefs' September 2026 broad US ChatGPT study ranks Reddit #1 by mention share at 16.8%, followed by Wikipedia at 7.0%, Consumer Reports at 3.7% and Forbes at 3.1%. (Ahrefs ChatGPT Top 50)

LLMAI's commercial health slice ranks Healthline #1, Forbes #2, Reddit #3, HearingTracker #4 and Consumer Reports #5 by Response Citation Coverage.

The studies use different metrics, denominators and prompt populations. Ahrefs uses broad all-topic queries and mention share. LLMAI uses high-stakes commercial health prompts and counts a domain at most once per eligible response for the primary metric.

The difference is the point: broad ChatGPT source rankings are not universal category maps.

Authority Signals preprint: institutional sources dominate a general consumer-health sample

A 2026 medRxiv preprint, Authority Signals in AI Cited Health Sources, analyzed 100 questions sampled from HealthSearchQA and coded 615 ChatGPT 5.2 Pro cited sources. The authors reported that more than 75% of cited sources came from established institutional sources such as major health systems, Wikipedia, the NHS and PubMed. (medRxiv preprint)

LLMAI studies a different question population. Its prompts are concentrated on high-stakes commercial consumer decisions, including product comparisons, providers, pharmacies, therapy platforms and senior-care services. That can elevate publishers, review sites, retailers and specialist industry sources that may be less prominent in a general health-information sample.

The medRxiv paper is a preprint and had not been peer reviewed at publication. It is useful as a source-composition comparison, not as a clinical validation of LLMAI's ranking.

Conductor: a massive healthcare benchmark surfaces a broader institutional market

Conductor's 2026 Health Care AEO/GEO benchmark is based on 13,770 domains, 3.5 million prompts, 17 million AI-generated responses and more than 100 million citations from a May-September 2025 index. It reports that ChatGPT generated 83.8% of measured AI referral traffic in its health-care domain sample and highlights Mayo Clinic and Cleveland Clinic as citation leaders in health care equipment and services. It also reports Healthline as the leading AIO share-of-voice domain across its Health Care industry. (Conductor Health Care benchmarks)

Conductor's unit of analysis is industry-classified domains and a very large cross-engine benchmark. LLMAI's unit is a response-level commercial decision panel. Those different frames explain why a consumer-health publisher such as Healthline can lead the LLMAI ChatGPT category while institutional health systems are more prominent in broader industry benchmarking.

Tinuiti: commercial-intent prompt tracking also shows category and platform effects

Tinuiti's 2026 AI Citation Trends research tracks commercial-intent prompts across multiple answer engines and categories, including OTC health. Tinuiti explicitly notes that prompt sets, platform citation volume and category mix affect cross-platform citation shares. (Tinuiti Q3 2026 report)

That methodological caution aligns with the central conclusion here: citation visibility should be measured against the exact platform and commercial question set a brand cares about.

How the LLM Authority Index Study Is Different

This study is not presented as the only valid way to measure health citations. Its contribution is the combination of:

• a high-stakes consumer-decision focus,
• 14 health and medical source verticals,
• a three-month July-September 2026 window,
• a response-level primary metric,
• visible citation-event counts as a secondary metric,
• registrable-domain consolidation,
• explicit extraction-failure handling,
• exact-repeat anti-double-counting,
• infrastructure exclusion for image and map endpoints,
• monthly persistence analysis,
• vertical-level source leaders,
• category lift against broad ChatGPT,
• and direct comparison with matching Google health studies.

The study is designed for publishers and marketing teams asking a specific commercial question: which third-party websites repeatedly appear in AI citations when consumers are making consequential health-related choices?

Why AI Citation Studies Can Disagree

Two citation studies can both be correct and publish very different domain rankings.

Prompt population

A symptom or definition prompt is not the same market as "best online therapy," "best hearing aids," "medical alert system cost," "online pharmacy" or "best fertility clinic."

Platform surface

ChatGPT, ChatGPT Health, Google AI Mode and Google AI Overviews are different products and surfaces. This study analyzes the captured ChatGPT platform family in the LLMAI archive. It should not be read as a study of the separate ChatGPT Health product.

Metric

This article ranks Response Citation Coverage. Other studies may rank total links, mention share, source-type share, pages cited or referral traffic.

Domain normalization

Subdomains are consolidated to registrable domains. A study that keeps subdomains separate will produce different rankings.

Infrastructure handling

This public table removes technical image and map infrastructure that does not function as a substantive publisher source in the analyzed response. Specifically, this slice excludes 198 images.openai.com entries, 88 mapbox.com entries and 32 openstreetmap.org entries.

Time period

AI citation behavior changes. This article uses July through September 2026. Studies from 2025 or early 2026 measure a different product state.

Geography and personalization

The archived corpus is not a census of every possible user session. Geography, account state, personalization, model version and live retrieval conditions can affect outputs.

Clinical research citations versus consumer web citations

A study that asks ChatGPT to write a medical literature review and verifies references against PubMed is answering a different question from a study of visible web citations in consumer decision responses. Citation accuracy research remains important, but it should not be conflated with citation visibility.

What the Results Mean for Health and Medical Publishers

1. Prove the exact commercial footprint

A publisher should know which prompt clusters generate citations, not just its overall AI visibility score. Healthline's value is different from HearingTracker's because the former has broad category reach while the latter has extreme specialist depth.

2. Preserve pages that repeatedly earn citations

If a page or URL pattern repeatedly appears in AI citations, treat it as distribution infrastructure. Avoid unnecessary URL changes, thin rewrites or migrations that remove the cited evidence without a plan.

3. Package independent citation data for advertisers

A publisher can use third-party citation coverage to strengthen media-kit claims, especially when the data is tied to a buyer's exact commercial category. Keep the metric precise. "Cited in 55.7% of captured Hearing Aids ChatGPT responses" is more defensible than "ChatGPT trusts us."

4. Separate platform inventory

ChatGPT health and the two Google health surfaces have limited Top 100 overlap. Publishers should track each platform independently.

5. Distinguish citation visibility from clinical authority

A citation frequency metric does not establish medical accuracy, clinical quality, treatment efficacy or safety. Publishers should not turn AI-search visibility into a medical-quality claim.

What the Results Mean for Healthcare CMOs and Brands

Start with the exact commercial decision cluster

Build prompt clusters around the decisions that drive revenue: provider selection, pricing, comparison, safety, features, eligibility and alternatives.

Identify recurring third-party sources

Separate company-owned citations from independent publishers, specialist review sites, government sources, retailers and community platforms. The outreach strategy should depend on the source type.

Prioritize specialist domains when the category data supports it

A specialist such as HearingTracker can be more valuable for a hearing-aid campaign than a much larger general publisher.

Measure ChatGPT separately from Google

Only 39 to 41 domains overlap between ChatGPT's health Top 100 and the matching Google health Top 100 lists. One platform's source market should not be used as a proxy for another.

Track citations and recommendations separately

A domain can be cited without the associated company being recommended. A company can also be mentioned without the source domain being part of the visible citation set. Measure both layers.

Re-run after material market changes

The low July-to-September Top 100 similarity shows why a one-time benchmark becomes stale. Re-run after major product launches, pricing changes, site migrations, regulatory events or major new earned media.

Keep compliance and medical review separate from AEO strategy

AI-search optimization does not reduce the need for medical, legal or regulatory review. High-stakes content should remain evidence-grounded and appropriately reviewed.

Methodology

Parent dataset

This article uses the frozen LLM Authority Index archive underlying the 2026 High-Stakes Consumer Decisions research series.

The parent project contains 153 unique stage-0 datasets after exact-file deduplication, 53 originating vertical labels and six platform families in the primary July-September analysis window.

Platform and category restriction

This article keeps records where:

• platform_family = ChatGPT
• proposed_family_id = F4, Health and Medical
• report month is July, August or September 2026

Explicit extraction failures

Rows marked as explicit extraction failures are excluded before ranking. This ChatGPT health slice contained zero explicit extraction failures.

Exact-repeat anti-double-counting

An exact repeat is collapsed when the following are identical:

1. report month,
2. exact raw platform/source surface,
3. normalized prompt text,
4. exact set of original citation URLs.

One record is retained. This is an analytical anti-double-counting rule, not proof that the prompt ran only once.

Prompt normalization

Prompt text is normalized using Unicode NFKC, case folding, whitespace collapse and trimming for duplicate detection.

Domain normalization

Citation hostnames are lowercased, www. is removed, and subdomains are consolidated to registrable domains using the Public Suffix List. For example, a citation to a publisher's finance or health subdomain is attributed to the registrable parent domain.

Public infrastructure handling

Technical entries that do not function as substantive publisher sources are excluded from the public ranking.

In this ChatGPT health slice, the exclusion removed:

• 198 images.openai.com entries,
• 88 mapbox.com entries,
• 32 openstreetmap.org entries.

That leaves 4,504 visible citation events across 858 registrable domains in the public analysis.

Primary ranking metric: Response Citation Coverage

For a domain d:

Response Citation Coverage(d) = eligible responses citing d / all eligible responses

A domain counts at most once per response for the primary metric.

Secondary metric: visible citation events

Raw visible citation events are reported as a secondary measure. Multiple URLs from the same domain can generate multiple events in one response.

Vertical membership

The source archive labels are used to calculate vertical breadth. Because exact-repeat deduplication can merge identical observations associated with overlapping source labels, vertical counts should be interpreted as corpus membership rather than independent experiments.

Monthly persistence

For each month, domains are ranked by response-level citation count. The study reports pairwise Top 100 overlap, Jaccard similarity and the number of domains present in all three monthly Top 100 lists.

Category lift

For domains with a broad ChatGPT baseline, category lift compares health/medical response coverage with broad ChatGPT response coverage.

Category Lift = Health ChatGPT Coverage / Broad ChatGPT Coverage

A lift above 1 means the domain is proportionally more visible in the health slice than in broad ChatGPT. It is descriptive, not causal.

Limitations

• This is a captured research corpus, not a census of all ChatGPT sessions.
• The 14 source labels are not equally sized or statistically independent.
• July has fewer eligible ChatGPT health responses than August or September.
• Fertility Clinics and IVF Clinics overlap heavily.
• Dental Implants has only 16 eligible ChatGPT responses in this slice.
• Some Home Health Care prompts are employment or franchise adjacent.
• Domain-level aggregation can hide page-level differences.
• Citation visibility does not prove recommendation, conversion, clinical quality or causal influence.
• Model updates, retrieval changes, geography, personalization and account state can change results.
• Cross-study comparisons are sensitive to prompt population, metric, source normalization and infrastructure rules.

Commercial Relationship Disclosure

LLM Authority Index is part of a business ecosystem that provides AI visibility, research, and marketing services. Affiliated businesses may have current or historical commercial relationships with companies or publishers that appear in the dataset. Commercial relationships are not inputs to ranking methodology. Domains are included based on observed citation data. A ranking is not an endorsement. A commercial relationship is not evidence of causation.

References

1. LLM Authority Index: The 2026 AI Citation Authority Study
2. LLM Authority Index: Most-Cited Websites in AI for High-Stakes Health and Medical Consumer Decisions
3. LLM Authority Index: The 100 Most-Cited Websites in ChatGPT for High-Stakes Consumer Decisions
4. LLM Authority Index: Google AI Overviews Health and Medical Top 100
5. LLM Authority Index: Google AI Mode Health and Medical Top 100
6. LLM Authority Index: Citation Drift From July to September 2026
7. BrightEdge: Healthcare AI Citations, ChatGPT vs. Google
8. Ahrefs: The 50 Most-Cited Websites in ChatGPT, September 2026
9. medRxiv: Authority Signals in AI Cited Health Sources
10. Conductor: Health Care Industry 2026 AEO / GEO Benchmarks
11. Tinuiti: Q3 2026 AI Citation Trends Report

Related LLM Authority Index Research

Parent studies and measurement framework

• The 2026 AI Citation Authority Study
• The Health and Medical Citation Authority Study
• The ChatGPT Platform Study
• The Persistence-Portability Gap
• Do AI Platforms Cite the Same Websites?
• How Stable Are AI Citations?

Sibling Health and Medical platform studies

• Google AI Overviews
• Google AI Mode
• Perplexity
• Gemini
• Microsoft Copilot

Health and Medical Reddit trend studies

• Health and Medical Reddit parent study
• Google AI Overviews Reddit trend
• Google AI Mode Reddit trend
• ChatGPT Reddit trend
• Perplexity Reddit trend
• Gemini Reddit trend
• Microsoft Copilot Reddit trend