Systematic audit of 50 Delhi NCR clinics measuring AI crawler access, schema adoption, entity clarity, and recommendation frequency across ChatGPT, Perplexity, and Gemini — May 2026.
In May 2026, we audited 50 clinic websites across Delhi NCR — dermatology, general practice, physiotherapy, and dental — measuring their AI discoverability across five dimensions: crawler access, structured data presence, entity signal clarity, content citability, and off-site citation consistency.
The results establish a baseline for the Indian clinic segment. They are not optimistic.
When we asked ChatGPT, Perplexity, and Gemini for recommendations across 15 common clinic-type queries ("best dermatologist in South Delhi", "skin clinic in Gurugram", "physiotherapist near Hauz Khas"), zero of the 50 audited clinics appeared without being directly named. Of the businesses that did appear, all had at least one of: a Practo listing, a Google Business Profile with 20+ reviews, or consistent NAP across 3+ major directories.
We checked each clinic's robots.txt for rules blocking GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. A clinic blocking GPTBot cannot be crawled for ChatGPT's training data or real-time browsing. Blocking PerplexityBot removes the clinic from Perplexity's live web index entirely.
The majority of blocks were unintentional: clinics using Disallow: / rules to prevent indexing of admin directories, where no User-agent specificity was applied. This means a rule intended to block one crawler blocks all of them. Of the 14 clinics with deliberate GPTBot blocks, none cited a specific reason — in most cases the block appeared to be applied via a site template or SEO plugin default.
| Bot | Blocked | Platform affected | Impact severity |
|---|---|---|---|
| GPTBot | 72% (36/50) | ChatGPT training + browsing | Critical |
| ClaudeBot | 68% (34/50) | Claude training data | High |
| PerplexityBot | 71% (35/50) | Perplexity live index | Critical |
| Google-Extended | 44% (22/50) | Google AI Overviews + Gemini | High |
Schema markup is the primary mechanism by which a site communicates entity metadata to AI systems in a machine-readable format. LocalBusiness schema (and its medical subtypes — Physician, MedicalClinic) communicates the clinic's specialty, location, operating hours, and practitioner details in a format AI systems can parse directly.
Of the four clinics with LocalBusiness schema, two had incomplete implementations missing address, telephone, or openingHours — reducing their schema's effective contribution to AI citation confidence. Zero clinics had Physician schema despite being medical practices, and only one had FAQPage schema despite FAQ sections appearing on 34% of the audited sites.
| Schema type | Present | Complete | Expected impact |
|---|---|---|---|
| LocalBusiness | 8% (4/50) | 4% (2/50) | High |
| Physician / MedicalClinic | 0% (0/50) | — | High |
| Person (practitioner) | 2% (1/50) | 2% (1/50) | Medium |
| FAQPage | 2% (1/50) | 2% (1/50) | High |
| Review / AggregateRating | 6% (3/50) | 4% (2/50) | Medium |
AI systems form entity representations of businesses from multiple signals — on-page text, schema markup, and off-site mentions. When these signals are ambiguous, thin, or contradictory, the AI's confidence in citing that business decreases. We evaluated entity clarity by asking: can an AI reading the homepage alone correctly identify the clinic's name, specialty, location, and lead practitioner?
"Delhi NCR" is not a locality signal — it tells an AI system that the clinic is somewhere in a 53,000 km² region. AI recommendation systems matching local queries require locality-level precision ("Sector 14, Gurugram" or "Hauz Khas, South Delhi") to confidently recommend a business for local intent queries. Clinics that specify only "Delhi NCR" cannot reliably appear in locality-specific AI recommendation outputs.
AI systems use consistent off-site mentions as a confidence signal for entity authority. When a clinic's name, address, and phone number appear identically across Google Business Profile, Practo, Justdial, and Sulekha, the AI treats this as corroborating evidence. Inconsistencies — different phone numbers, address formatting variations, name truncations — reduce citation confidence.
The most common discrepancy type was phone number variation (different formats: +91 vs 91 vs 0-prefix) appearing in 38% of cases, followed by address formatting differences (Sector 14 vs Sec-14 vs S-14) in 27%. Clinics with zero NAP inconsistencies were 2.8× more likely to appear in Perplexity recommendation outputs for their category and locality.
Within the audited set, the clinics with the strongest AI presence shared a consistent profile:
None of these clinics had schema markup. This confirms our hypothesis that off-site citation density can partially compensate for absent structured data — but only to a point, and only for Perplexity (which accesses live web data). ChatGPT and Claude still showed no citations for any of the 50 audited clinics, regardless of directory presence.
The Indian clinic segment in 2026 has near-zero AI visibility infrastructure. The opportunity cost of this gap will compound: as AI recommendation usage grows and early-moving clinics build citation authority, the gap between visible and invisible clinics will widen. Businesses that establish AI citation foundations now face lower competition than they will in 12 months.
The six highest-impact actions for any clinic in this audit cohort, in priority order:
Clinics were selected across five specialties (dermatology, general practice, physiotherapy, dental, gynaecology) and six Delhi NCR sub-localities (South Delhi, Gurugram, Noida, Dwarka, Rohini, Karol Bagh). Crawler access was verified by reading robots.txt directly. Schema was verified using schema.org's validator and direct page source inspection. AI query testing was conducted over seven days in May 2026 using standardised query templates across ChatGPT (GPT-4o), Perplexity (default), and Gemini (Advanced). All queries were run without browser history or session context.