Original Data Study · v1.0

54%of Sampled NYC-Area Businesses Are Invisible to AI Search.

We measured 99 real NYC-area business IDs across 4 verticals and 5 sampled areas against two major AI engines (Perplexity and Gemini). 40% were named by an AI engine. 1% had their own website cited as a source.

See the results

Data collected 2026-07-17 · 198 primary measurements (n) · Published 2026-07-18 · CC BY 4.0

99
Businesses measured
4 verticals · 5 sampled areas
198
Primary measurements
the study's n — usable engine observations
2
AI engines tested
Perplexity, Gemini
1.1
Avg position when mentioned
1 = first recommendation

The data

The headline numbers

Every percentage below carries its denominator. This is an exploratory first measurement — see Limitations.

54% of sampled NYC-area businesses are invisible to AI search engines, and only 1% had their own website cited as a source.

— summary of 198 measurements across 99 NYC businesses, 2026-07-17.

Findings

Five findings worth citing

In this sample, the two tested engines named fewer than half of the businesses — and exposed the businesses’ own websites as citations in only 1% of measurements.
40% mentioned, 1% cited (n=198)
Over half of the sampled NYC-area businesses were entirely invisible to AI search across both engines tested.
54% invisible (46 of 99 distinct businesses)
The sample’s mention rate varied by vertical: med-spa and dental led, while legal and CPA/trades trailed.
med-spa 57% · dental 51% · legal 26% · other 26%
Even when a business was recommended, its own website almost never appeared in the answer or exposed source-citation list.
1% cited (2 of 198 measurements)
Both tested engines had below-half mention rates in this sample.
Perplexity 42% · Gemini 37% mentioned

Results

Full results

Rates are computed on the primary set (rows that successfully captured an answer; blocked rows are excluded and reported separately, never scored as “absent”).

Deep dives

By vertical, in depth

Real measured businesses, not invented. Names appear where the engine surfaced them in response to a generic local query.

Med-spa & dermatology57% mentioned

Medical spas, aesthetics clinics, and dermatology practices — high-review-count, visually-driven businesses.

Med-spas were the most visible vertical: 57% were named by an AI engine (28/49, n=49). The category maps cleanly to consumer "best of" queries.

Named by both engines
  • Glowbar Jersey City — Newport, Jersey City
  • Bared Monkey Laser Spa — Long Island City, Queens
Named by one engine
  • Elan Salon & Day Spa — Park Slope, Brooklyn (Perplexity)
  • Luxe Haven Medspa — Riverdale, Bronx (Perplexity)

Dental51% mentioned

General dentists, pediatric, oral surgery, and orthodontics across four NYC boroughs plus Jersey City.

Dental practices were named 51% of the time (25/49, n=49) — but zero had their own website cited as a source by either engine.

Named by both engines
  • Midtown Dental Group Park Slope — Park Slope, Brooklyn
  • Midtown Dental Group Jersey City — Downtown Jersey City
Named by one engine
  • Nuance Dental Studio — Upper East Side, Manhattan (Perplexity)
  • Charming Smile Dental — Downtown Jersey City (Perplexity)

CPA, real estate & home services26% mentioned

CPAs/accountants, real-estate agents, HVAC, plumbers, and electricians — grouped as "other professional & home services."

CPAs and trades were tied with legal as the least visible (26%, 13/50, n=50). Despite high review counts and mature Google Business Profiles, AI engines rarely surfaced them.

Named by both engines
  • Miller & Company LLP, CPA of NYC — Midtown, Manhattan
  • Sal Merante Tax & Consulting — Pelham Bay, Bronx
Named by one engine
  • Isaac Ostreicher, CPA — Downtown Brooklyn (Perplexity)

Method

Methodology

Sample

100 locations were sampled deterministically (seeded RNG, stratified by area) from 1,334 real businesses scraped from Google Maps across 40 query templates spanning Manhattan, Brooklyn, Queens, the Bronx, and Jersey City. The sample has 25 locations per vertical and represents 99 unique business IDs because one brand appears at two locations. 92 of 100 locations had a website.

Query design

Each sampled location was measured with a localized consumer-discovery query templated from its vertical and neighborhood (e.g. "best dentist near Park Slope Brooklyn", "best immigration lawyer in Downtown Brooklyn"). The query never contained the business’s own name — this measures whether the engine discovers the business, not whether it can recall a named business.

Scoring

A business was counted "mentioned" if its name appeared in the engine’s rendered answer (fuzzy, tokenized match with confidence tiers). It was counted "cited" if its website domain appeared in the answer body or in the engine’s source-citation list. Position is the 1-based rank among named businesses. Blocked rows (CAPTCHA/login/timeout) were excluded from all rates — never scored as "absent".

Engines & data collection

Perplexity (perplexity.ai) and Gemini (gemini.google.com/app), both via their logged-out consumer surfaces, driven by Playwright with a fresh browser context per query to avoid per-session anti-bot detection. Data collected 2026-07-17.

Reproducibility

The sampler uses a seeded RNG, so the 100-location sample is bit-identical across re-runs. Re-running measurement can produce different per-business results because the engines are non-deterministic. The full local harness, raw JSONL, and screenshots are retained for review; this public page publishes the methodology, denominators, and limitations rather than a downloadable per-business dataset.

Honesty

Limitations & honest disclosures

Transparency increases credibility; hiding limitations destroys it. This is an exploratory first measurement, not a definitive study.

  • Small sample per vertical (n≈25). Findings are descriptive, not statistically significant.
  • Only two engines (Perplexity, Gemini) were measured. ChatGPT and Google AI Overviews were not included in v1.
  • Single point in time (2026-07-17). AI engine outputs shift weekly; this is a snapshot, not a trend.
  • Single market (NYC + Jersey City). Findings may not generalize to other metros.
  • Logged-out consumer experience only. Logged-in (Plus/paid) experiences may differ.
  • Discovery queries, not name-recall. A business that appears when you name it may still be invisible to consumers who don’t.

What v2 will add

  • Add ChatGPT (logged-in persistent session) and Google AI Overviews (throttled / warmed profile).
  • Expand the sample to 200 businesses (50/vertical) for tighter per-vertical intervals.
  • Re-measure quarterly to track how the visibility gap shifts as engines evolve.
  • Cross-validate browser-captured answers against official APIs where available (fidelity check).

FAQ

Frequently asked questions

46% of the sampled NYC-area businesses were mentioned by at least one AI engine; 54% were entirely invisible. Of 198 measurements across Perplexity and Gemini, 40% named the business by name and only 1% cited its website as a source (n=99 distinct businesses, data collected 2026-07-17).

Med-spa and dermatology had the highest AI mention rate at 57% (28/49), followed by dental at 51% (25/49). Legal (immigration and personal injury) and CPA/trades tied for lowest at 26% each (13/50).

Only 1% — 2 out of 198 measurements had the business’s own website domain appear in the answer or exposed citation list. The metric records what the engine displayed; it cannot identify every source the engine used.

Perplexity and Gemini (99 measurements each). ChatGPT and Google AI Overviews were not included in v1 — ChatGPT began gating anonymous queries behind a login wall during data collection, and Google’s search endpoint was IP-rate-limited. The study commits to adding both in v2.

100 NYC-area business locations across 4 verticals and five sampled areas (four NYC boroughs plus Jersey City) were sampled from Google Maps. They represent 99 unique business IDs because one brand appears at two locations. Each was queried against Perplexity and Gemini using a localized discovery query (e.g. "best dentist near Park Slope Brooklyn") — never the business’s own name, to measure discovery rather than recall. Each answer was scored for whether the business was named (fuzzy name match) and whether its domain appeared as a source. Full methodology is on the study page.

No — with n≈25 locations per vertical, the study reports descriptive percentages, not significance testing. It is framed as an exploratory first measurement, not a definitive one. The pattern appears across two engines and five sampled areas, but a larger v2 sample is needed before drawing broader conclusions.

Cite

Cite this study

Findings may be cited and quoted with attribution to EnovaCreations. The headline stat and full methodology above are the citable record.

A DOI (Zenodo deposit) is pending — the page is structured for it; the deposit is a manual step.

Headline stat (copy-paste)

54% of sampled NYC-area businesses are invisible to AI search engines, and only 1% had their own website cited as a source.

EnovaCreations, NYC-Area Local Business AI Visibility Study (2026)

Researchers who want the underlying data can request it via the contact page.

Sources

References & sources

GEO: Generative Engine Optimization (Princeton / IIT Delhi, KDD 2024)The peer-reviewed study that introduced GEO and tested content changes including cited sources, statistics, and fluency on its benchmark.
Google: Optimize your content for AI search experiencesOfficial guidance on technical prerequisites for AI-search visibility.
Schema.org Dataset typeThe structured-data standard this study’s data is published under (see the JSON-LD on this page).
PerplexityEngine measured in this study.
Google GeminiEngine measured in this study.

Is your business one of the 54%?

We measure and improve how AI engines represent your business — across ChatGPT, Perplexity, Gemini, and Google AI Overviews. Find out where you stand.

AI Search Visibility serviceCheck your AI visibility