What determines whether a business appears in ChatGPT, Google AI Overviews and Gemini — a controlled study of 216 New Jersey health practices.
📄 As featured on Academia.edu →This paper examines the factors that determine whether an independent local health practice appears in AI-generated recommendation responses across ChatGPT, Claude and Gemini. Drawing on a dataset of 216 New Jersey health practices tested using standardised local search prompts, we identify the entity visibility gap as the primary barrier to AI citation for local businesses. Our dataset reveals that 98% of practices scored zero across all three AI models, with no practice achieving cross-model visibility. We conclude that a single well-indexed third-party publication placement may be sufficient to generate initial AI citation presence.
The emergence of AI-generated search results has fundamentally altered how local businesses are discovered by potential customers. Google AI Overviews now appear on 88% of health-related searches (BrightEdge, 2026). ChatGPT processes over 5.7 billion monthly visits (Exploding Topics, 2026), and Gemini is integrated directly into Google's search interface. When a patient asks an AI system for a local health practice recommendation, the system generates a direct answer naming one to three providers — before any website is visited.
The stakes are significant. Brands cited in AI Overviews experience a 35% higher organic click-through rate and 91% higher paid click-through rate compared to uncited competitors on the same query (Seer Interactive, 2026). AI search traffic converts at 14.2% compared to 2.8% for traditional search — a 5x differential (Exposure Ninja, 2026). Yet 26% of brands have zero mentions in AI Overviews.
Existing literature on AI citation factors has focused primarily on enterprise brands, SaaS companies, and national publishers. The local health vertical — independent dental practices, physical therapists, chiropractors, medical spas, and specialty clinics — has received minimal attention. This paper addresses that gap through empirical testing of 216 independent New Jersey health practices across three major AI systems.
Goralewicz (Onely, 2026) identifies five primary factors that determine AI search ranking: brand mention volume, content freshness, structured format, schema markup, and traditional ranking signals. The most significant finding is that brand mentions correlate 3x more strongly with AI citations than backlinks (0.664 vs 0.218 correlation coefficient). Content freshness is substantial: 76.4% of ChatGPT's most-cited pages were updated within the last 30 days. Structured format carries significant weight: listicles achieve a 25% citation rate versus 11% for narrative blog posts.
XFunnel's analysis of 250,000 citations reveals significant platform-level differences. Perplexity cites an average of 6.61 sources per answer; Google Gemini cites approximately 6.1; ChatGPT cites only 2.62. This means ChatGPT represents the most competitive citation environment — fewer slots, higher competition. Commercial domains account for over 80% of all AI citations.
Pol and Ali (Semrush, 2026) provide a structured framework emphasising brand mention acquisition, content restructuring for AI extraction, and technical accessibility for AI crawlers. Critically, 76.1% of URLs cited in Google AI Overviews also rank in the top 10 of traditional Google search results — establishing a strong correlation between conventional SEO authority and AI citation.
None of the above studies address the specific conditions of independent local health practices — businesses with limited third-party coverage, restricted access to high-authority publication networks, low brand mention volume, and minimal structured data implementation. This paper fills that gap.
Each practice was tested using a standardised query format designed to replicate real-world patient search behaviour:
"best [simplified category] in [city] NJ"
| Score | Meaning |
|---|---|
| 0 | No mention detected in AI response |
| 1–39 | Weak or incidental presence |
| 40–59 | Partial inclusion |
| 60–79 | Consistent inclusion |
| 80–100 | Strong inclusion — none observed in this dataset |
| Parameter | Value |
|---|---|
| Sample size | 216 practices |
| Geography | New Jersey — Bergen County and South Jersey Shore |
| AI models | ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google) |
| Test period | June 2026 |
| Rate control | 4-second delay per query |
| Tooling | AEOGeoAI visibility scoring system |
Limitations: AI outputs are probabilistic and may vary across time and model updates. Results represent a snapshot of observed behaviour in June 2026.
| Score | Meaning | Practices |
|---|---|---|
| 0 | No citation presence | 212 (98.1%) |
| 1–39 | Minimal presence | 0 |
| 40–59 | Weak presence | 3 (1.4%) |
| 60–79 | Moderate presence | 1 (0.5%) |
| 80–100 | Strong presence | 0 |
Only four practices appeared in any AI-generated recommendation. All four scored on a single model only.
| Practice | Category | City | ChatGPT | Claude | Gemini |
|---|---|---|---|---|---|
| Dental Arts of Hackensack | Dental | Hackensack | 50 | 0 | 0 |
| Fort Lee Physical Therapy | PT | Fort Lee | 50 | 0 | 0 |
| Fort Lee Orthodontics | Orthodontics | Fort Lee | 50 | 0 | 0 |
| New Jersey Eye Center | Ophthalmology | Bergenfield | 0 | 0 | 75 |
| Category | Tested | Zero score | Zero rate |
|---|---|---|---|
| Dentistry | 38 | 37 | 97% |
| Physical therapy | 21 | 20 | 95% |
| Chiropractic | 18 | 18 | 100% |
| Medical spa / Aesthetics | 19 | 19 | 100% |
| Orthopaedics / Sports medicine | 16 | 16 | 100% |
| Plastic surgery | 9 | 9 | 100% |
| Mental health / Psychiatry | 8 | 8 | 100% |
| Ophthalmology / Optometry | 8 | 7 | 88% |
| Paediatrics | 7 | 7 | 100% |
| Other specialties | 72 | 71 | 99% |
Goralewicz (2026) identifies brand mention volume as the factor most strongly correlated with AI citation presence (0.664). Our dataset is consistent with this finding. The 98% zero-score rate reflects the near-complete absence of third-party brand mentions for independent local health practices — practices that are well-known locally but effectively invisible to AI training data and retrieval systems.
Signal gap vs ranking gap. This is not a ranking problem. It is an entity visibility problem — AI systems surface practices with sufficient external confirmation signals. A practice can rank Page 1 on Google and score zero across all three AI models.
The entity visibility gap describes the absence of sufficient third-party indexed signals for AI systems to confidently include a business in recommendation outputs. For local health practices, this gap is structural: independent practices have limited access to the publication networks, PR infrastructure, and community presence that enterprise brands use to build brand mention volume.
No practice achieved cross-model visibility — no practice scored above zero on more than one AI model simultaneously. This suggests that AI citation presence at the local health level is model-specific rather than systemic. Cross-model visibility likely requires a higher volume of consistent third-party signals — the kind produced by multiple indexed publication placements rather than a single source.
Based on observed patterns, a single well-indexed third-party publication placement may be sufficient to shift a practice from zero to measurable citation presence in at least one model. The distribution — 212 at zero, 3 in the 40–59 range, 1 in the 60–79 range — suggests a threshold effect: practices are either invisible to AI systems or partially visible, with no middle ground in this dataset.
The local health vertical represents a substantially underserved market for AI visibility services. 98% of practices in the sample have zero AI citation presence, creating a large addressable market with a clear, reproducible solution.
This study establishes a reproducible methodology for measuring AI citation presence at the local business level. The query format, scoring system, and multi-model simultaneous testing protocol provide a framework applicable to other geographies, verticals, and markets.
Our study of 216 New Jersey health practices across ChatGPT, Claude and Gemini confirms that most independent local health practices are completely invisible to AI-generated local recommendation systems. The 98% zero-score finding reflects the structural absence of third-party entity signals that AI systems require before citing a local business.
The entity visibility gap is the defining barrier for local health practices in the AI search era. It is addressable: a single well-indexed third-party publication placement on a trusted local source may be sufficient to generate initial citation presence in at least one AI model.
As 74% of New Jersey residents are already using AI tools (Rutgers, 2026), practices that build AI citation presence now will compound that advantage as the market continues its shift from search ranking to AI recommendation eligibility.