When a patient asks ChatGPT about dental care, where does the answer come from?
Mostly from dentists’ own websites. Practice-owned sites are 53.8% of everything ChatGPT cites on dental questions – the single largest source, more than three times the next.
We ran 1,405 patient-realistic dental questions across 20 US metros and seven areas of dentistry, then categorized all 13,152 sources the model cited. This page is the dataset: the full source mix, the contrast with medicine, and how the share moves by sub-specialty and metro.
- Queries
- 1,405
- Citations
- 13,152
- Metros
- 20
- Areas of dentistry
- 7
- Model
- GPT-5.5
- Run
- June–July 2026
53.8% of ChatGPT’s dental citations land on a practice’s own website – spread across 2,497 distinct practice domains.
Everything ChatGPT cited, in thirteen buckets
No directory, association, or review site comes close. The next-largest bucket after practice websites is specialty associations at 17.6%.
Each citation is a URL the model pulled from while answering. We sorted all 13,152 of them into source categories using hand-audited rules. 1,390 of the practice domains were cited two or more times – no single chain or roster concentrates the share.
Share of all 13,152 citation events, 1,405 dental queries, June–July 2026. A single answer can cite several sources.
The share of all 13,152 cited sources that are some practice’s website. A statement about what the source pool is made of – not a figure for any single practice.
The share of the 1,405 questions whose answer cited at least one practice website. A statement about coverage – how widely practice sites show up across questions.
Two different denominators. Keep them apart: the first describes the pool’s composition, the second describes coverage.
In medicine, the practice site is a distant third
We ran the same study for physicians – same model, same prompt design, same markets. The top of the source pool inverts.
Medicine’s largest bucket is the hospital-system roster: “find a doctor” pages the physician does not own and cannot easily edit. Dentistry has no comparable roster, so the share hospitals capture in medicine lands instead on the surface the dental practice already owns.
Medicine
38,178 citations · May 2026
Dentistry
13,152 citations · June–July 2026
Both stacks are scaled to dentistry’s 53.8% so the two columns read on one scale. A physician is often described online by pages she does not hold; a dentist holds the main page herself. Explore the medical corpus in the companion studies below.
Specific questions send the model to practice sites
The share holds everywhere we tested – the floor is 45% – but it climbs when the question names a specialty.
What moves the number is whether the question names a specific dental specialty: vague questions invite directories, specific clinical ones send the model back to the practice’s own pages.
Bars scale to orthodontics’ 75.8%. Directories only show up in force on the generalist query (15.0%) – a question that names no specialty sends the model looking for a list. Pediatric’s remainder runs through regional “top dentists” magazine lists (16.9%).
The share moves by market, and the floor stays high
Practice-owned share runs from 38.5% in New York to 63.5% in Charlotte. Even the lowest metro sits above medicine’s corpus-wide 23.1%.
Bars scale to Charlotte’s 63.5%. Amber marks metros above 60%. Denser, brand-heavy metros (New York, Los Angeles, Boston) spread citations across more source types; the pattern mirrors the medical study’s metro gradient at a far higher level.
Audit your own practice in 5 minutes
Open ChatGPT with web search on and ask these four questions. Count how many answers cite your website.
- Who is the best dentist near me in [your city]?
- Who does Invisalign in [your city]?
- I need a periodontist for gum disease in [your city] – who should I see?
- How much do veneers cost, and who does them well in [your city]?
Does ChatGPT cite your practice?
Run a free scan to see your A–F visibility grade and how often ChatGPT cites your practice.
Questions about this research
Short answers to how the study was run and what it does and doesn’t claim.
What share of ChatGPT’s dental citations are practice websites?
53.8% of the 13,152 sources ChatGPT cited across 1,405 dental questions were practice-owned websites – the single largest category, more than three times specialty associations at 17.6%. The share describes the whole citation pool, spread across 2,497 distinct practice domains.
Does 53.8% mean my practice gets cited 53.8% of the time?
No. The 53.8% describes what the citation pool is made of, not any single practice’s rate. Coverage is a separate figure: 87.2% of the questions we ran cited at least one practice website. Whether yours is among them depends on whether your pages give the model something specific to quote.
Why is dentistry different from medicine?
Medicine’s largest citation source is the hospital-system roster at 27.1%, with specialty associations close behind – pages the physician does not own. The practice’s own website is a distant third at 23.1%. Dentistry has no comparable hospital roster, so that share lands instead on the surface the practice already owns.
Which areas of dentistry lean most on practice websites?
The more specific the specialty, the higher the share: orthodontics at 75.8% and periodontics at 72.5% lead, while generalist “best dentist” queries (45.9%) and cosmetic (45.6%) sit at the bottom. Generalist queries are also the one place directories show up in force, at 15.0%. No area we tested fell below 45% practice-owned.
Which AI model and time period does this cover?
All 1,405 questions ran against GPT-5.5 via the OpenAI API in June and July 2026, across 20 US metros, with multiple runs per question to absorb variation. The work is single-platform and US-only – behavior on Gemini or Google’s AI Overviews may differ.
Methodology & caveats
We ran 1,405 patient-realistic questions through GPT-5.5 via the OpenAI API across June and July 2026 – among them “best dentist in [city],” “who does Invisalign,” “I need a periodontist for gum disease” – spanning 20 US metros and seven areas of dentistry, with several runs per question to absorb run-to-run variation. A single answer can cite several sources, so the 13,152 figure counts citation events, not questions; each was sorted into a source category using hand-audited rules. The work is single-platform and US-only; behavior on Gemini or Google’s AI Overviews may differ. It is a snapshot rather than a trend and correlational rather than causal, and any study that begins from what the model already cites carries survivor bias. The medical comparison draws on the May 2026 corpus behind the two companion studies below, gathered the same way.