AI Scan Invisibility Study Methodology — How We Measure Local Business Invisibility

Status: methodology published — findings pending

Headline statistics are intentionally withheld until aggregate sample size reaches N≥100 completed scans. Until then, this page documents the measurement method, inclusion criteria, and privacy rules so linked research references do not lead to a 404 or unsupported claims.

Research question

When local businesses run buyer-intent prompts across major AI assistants, how often are they never mentioned while competitors are named?

Secondary questions:

1. Which platforms show highest invisibility rates?
2. How does invisibility correlate with review count bands (anonymized buckets)?
3. Do categories (home services vs healthcare vs legal) differ materially?

Why this matters

Anecdotes drive AI visibility hype. Aggregate scan data — collected with consistent prompt methodology — offers a reality check for SMB owners deciding whether to invest in answer-engine optimization programs.

We publish only above N=100 because sub-threshold samples produce confident-sounding percentages that are statistically fragile.

Data source

Primary: Anonymized results from AIrecommend.ai free scans completed between June 2026 and rollup date.

Inclusion criteria:

• Valid US market + category selection
• Six-platform sample completed without error
• At least three named competitor slots returned in market
• Outlier test accounts excluded (internal, demo, duplicate domain)

Exclusion:

• Incomplete scans
• Non-local categories (if miscategorized)
• Prompt sets flagged for API timeout partials

Methodology

Prompt design

Scans use buyer-intent prompt templates parameterized by:

• Service category (e.g., plumber, dentist, personal injury lawyer)
• Geography (city / metro)
• Optional urgency modifiers sampled rotationally

Templates mirror the production scan engine — not ad-hoc manual questions — so aggregates reflect product methodology.

Platforms sampled

1. ChatGPT
2. Gemini
3. Claude
4. Perplexity
5. Grok
6. Google AI Overviews (where available)

Metrics computed

Metric	Definition
Mention rate	Share of prompts naming the scanned business
Invisible flag	Zero mentions on ≥1 platform while ≥1 competitor mentioned
Share of AI voice	Scanned business mentions ÷ all business mentions in sample
Platform blind spot	Per-platform zero mention with competitor presence

Aggregation

• Businesses bucketed by review count deciles (from public signals at scan time)
• Category clusters: home services, healthcare, professional services, hospitality
• Geographic strata: top 50 US metros vs long tail

Privacy

• No business names in published charts
• Minimum cell size k≥5 for any reported slice
• Only team members with research role access row-level data

Planned findings format

When the publication threshold is met, this page will add headline statistics with confidence intervals and methodology version notes.

Table A — Invisibility by platform

Platform	% scans with zero mention
ChatGPT	Held until N≥100
Gemini	Held until N≥100
Claude	Held until N≥100
Perplexity	Held until N≥100
Grok	Held until N≥100
AI Overviews	Held until N≥100

Table B — Invisibility by review bucket

Review count bucket	% invisible ≥1 platform
0–25	Held until N≥100
26–100	Held until N≥100
101–300	Held until N≥100
300+	Held until N≥100

Interpretation guidelines (for future publish)

When findings publish:

1. Report confidence intervals on headline percentages
2. Note methodology version and date range
3. Compare to platform overlap research — low citation overlap suggests invisibility will be platform-specific
4. Avoid causal claims ("reviews cause mentions") — observational data only
5. Link readers to individual scans for actionable baselines

Limitations

• Selection bias — businesses running scans may be more marketing-aware
• US-centric early samples
• API variance — model updates mid-study create temporal noise
• Competitor set — auto-detected competitors may not match owner's mental list
• Not a forecast — past mention rates do not guarantee future movement after fixes

Relationship to product

This study informs AIrecommend.ai content and sales honesty — not algorithmic tuning of client delivery. Client programs use private monthly rescans, Growth Engine modules, and Super Pixel attribution.

Tracking overview: AI visibility tracking.

Reproduce or challenge our work

Researchers and journalists may request methodology exports after publication by contacting the team through the site. We will share:

• Prompt template redacts (no client PII)
• Aggregation code references
• Exclusion rule changelog

We welcome third-party replication with disclosed prompt parity.

For business owners

You do not need to wait for this study to act.

1. Run your free scan
2. Read your mention table — not industry averages
3. Fix NAP, reviews, entity per module map
4. Resample in 30 days

Public pricing now runs from AI Starter $97/mo to Category Authority $29,999/mo — see the full ladder on pricing.

Changelog

Version	Date	Notes
0.1-methodology	June 11, 2026	Methodology only; headline findings held until N≥100

---

Methodology note — findings publish when N≥100.