The mission

We're running the largest, most rigorous public study on AI visibility — and re-running it every quarter.

AI search is rewriting how businesses get found. ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews now mediate billions of buying-intent queries per month — and the rules for being recommended are completely different from classical SEO.

Most published GEO advice is built on case studies of N=1, hot takes from a single account, or rebranded SEO playbooks. We wanted to know what actually moves AI visibility — across thousands of businesses, multiple verticals and markets, with statistical controls strong enough to hold up under attack.

So we built a research program. Same sample, same methodology, re-run quarterly. Each study layer adds a deeper question on top of the last. Findings published whether they validate our product or not.

Data posture

Findings are public. Methodology is public. Per-slot statistics are public. Individual brand lookup is free. The 2,729-row underlying dataset is licensed to research partners under NDA — same posture as Pew, MIT Tech Review, Backlinko, GitHub Octoverse. Apply for full access →

The headlines so far

Six findings that will outlive the studies

Pulled from across the published studies. Click any card to read the full evidence.

Study 2 · n=2,545 · strict isolation

“Reddit's predictive power collapses from r=0.333 to r=0.000 once you control for general multi-platform presence.”

Read the study

Study 2 · n=76,457 · phi=0.167

“When a brand IS the URL Perplexity cites, it's 5.5x more likely to be mentioned in the response text.”

Read the study

Study 1 · n=1,004 · 5 models

“66% of businesses are completely invisible to AI — zero mentions across all five major models.”

Read the study

Study 2 · DA Q4 (54+) · n=250

“Directory presence delivers a +16 visibility-point lift specifically in the top Domain Authority quartile.”

Read the study

Study 2 · 32 industry × city slots

“Off-page leverage varies 30x across verticals — from 0% addressable in Med Spa to 30% in NYC plumbing.”

Read the study

Study 2 · n=66,918 responses · 0 sources

“The ChatGPT API doesn't return source URLs. Claims about 'what ChatGPT cites' built from API data are not supported by API data.”

Read the study

Published academic-format papers

White papers

The findings of each study, repackaged in formal academic register — abstract, methodology, results, discussion, limitations, references. Cite these versions in research, journalism, and analyst work.

Paper 1 · v1.1·April 10, 2026·~16 pages · 5,538 words

The AI Visibility Index: An Empirical Baseline of Business Visibility Across Five Generative AI Models, Ten Industries, and 1,004 Businesses

Phase 1 of the AI Visibility Index Research Program

Abstract excerpt

This paper reports the results of an empirical baseline study examining business visibility within five generative AI search engines: ChatGPT (gpt-4o), Perplexity (sonar-pro), Gemini (2.5-flash), Claude (Sonnet), and Google AI Overview (via SerpApi). Conducted between February and April 2026, the study evaluated 1,004 businesses across ten industry verticals, probing each business against 100 prompt-model combinations and producing 95,392 individual mention checks. Three findings are reported: 66% of the sample is completely invisible to AI; Domain Authority (Pearson r = 0.337) and Google review count (r = 0.333) are the strongest single predictors; and AI models exhibit substantial cross-model disagreement on which businesses to recommend.

Cite as: House, J. (2026). The AI Visibility Index: An empirical baseline of business visibility across five generative AI models, ten industries, and 1,004 businesses (Phase 1 white paper, v1.1). MentionLayer Research.

Download DOCX196 KB View web version

Paper 2 · v1.1·April 27, 2026·~38 pages · 13,510 words

The AI Visibility Index: A Cross-Market Empirical Study of Generative Engine Optimization Signals Across 2,729 Businesses, Five AI Models, Fourteen Industries, and Thirty-Two Industry-Market Slots

Phase 1 + Phase 2 Combined Findings, with Pre-Registration of Phase 3 Controlled Intervention

Abstract excerpt

This study examines the empirical predictors of business visibility within generative search engines, applying OLS-residual partial-correlation methodology to disentangle the independent contributions of 24 measurable off-page signals. Phase 2 (n = 2,729) finds that directory presence (r = 0.391) outranks Domain Authority (r = 0.338) as the top raw predictor, but no single off-page signal exceeds r = 0.10 in strict isolation; Reddit's predictive power collapses from r = 0.333 to r = 0.000 once general multi-platform presence is controlled for. The two AI models that return source URLs (Perplexity, Google AI Overview) exhibit a 5.5× lift in mention probability when the brand is the cited source. Phase 3 controlled-intervention design is pre-registered.

Cite as: House, J. (2026). The AI Visibility Index: A cross-market empirical study of generative engine optimization signals across 2,729 businesses, five AI models, fourteen industries, and thirty-two industry-market slots (Phase 1 + 2 combined white paper, v1.1). MentionLayer Research.

Download DOCX693 KB View web version

The study program

Phase 1 → 2 → 3 → 4 — and beyond

Each study layer answers a question the previous one couldn't.

1

Published·April 2026

Study 1 — AI Visibility Index

The baseline. Which businesses get recommended by AI models — and why.

66% of businesses are completely invisible to AI. Domain Authority (r=0.337) and Google reviews (r=0.333) are the strongest single predictors.

Study 2 — The Off-Page AI Visibility Index

What's INSIDE Domain Authority? Decomposing the strongest predictor into actionable off-page signals.

Reddit's predictive power collapsed from r=0.333 to r=0.000 once we controlled for general multi-platform presence. Directory presence (r=0.391) is the new #1 predictor — beating DA itself.

3

Recruiting now·May–July 2026

Study 3 — Layer 3: Controlled Intervention

Phase 2 was observational. Phase 3 is causal. The first controlled before-and-after experiment in GEO.

25–30 businesses · 60-day intervention · two dose groups · pre-registered success thresholds. Result published regardless of direction — including null results.

25–30 participants

60 days

6 metric deltas

Pre-reg thresholds

Apply for the trial

4

Planned·October 2026

Study 4 — Re-baseline + Layer 4

Quarterly re-run of the full Layer 1+2 pipeline against the same 2,729 businesses. Tracks how AI visibility distributions shift over six months. Plus Layer 4 — competitor delta tracking.

What changed in 6 months? Did the visible stay visible? Did Layer 3 interventions move the needle? Quarterly cadence — same methodology, same sample, fresh data.

Same sample

Q4 ’26 ships

Quarterly cadence

5+ AI models

On the roadmap. Subscribe to be notified when it ships.

Updates

What's new in the research

Every release, milestone, and operational note. Newest first. Updated monthly.

Apr 27, 2026
Press
Combined Phase 1 + 2 white paper published (v1.1)
Formal academic-format white paper combining both phases of the research program. 38 pages, 20 tables, 6 inline figures, full methodology, pre-registration of Phase 3, and complete reference apparatus. Released alongside the standalone Phase 1 paper.
Download papers
Apr 27, 2026
Study published
Study 2 shipped — Off-Page Decomposition
The contrarian sequel to Study 1. Reddit's predictive power collapses to r=0.000 once we control for general multi-platform presence. Directory count overtakes Domain Authority as the #1 raw predictor of AI visibility.
Read the study
Apr 27, 2026
New tool
Interactive explorer launched
Pick any of 30 industry × city slots. See top predictors, addressable citation share, top cited domains, and visible-vs-invisible profile for that exact slot. Browse-only.
Open the explorer
Apr 27, 2026
Milestone
Layer 3 recruitment opening soon
Recruitment for the controlled intervention trial opens in May. 25–30 spots across SaaS, home services, real estate, personal-finance apps, personal injury law. Email Joel for first access.
Express interest
Apr 10, 2026
Study published
Study 1 shipped — AI Visibility Index baseline
1,004 businesses. 10 industries. 95,392 mention checks across ChatGPT, Perplexity, Gemini, Claude, and Google AI Overview. Headline: 66% of businesses are completely invisible to AI.
Read the study

Three ways to plug in

Get on the research distribution list

Each new study, each monthly update, and every Layer 3 result lands first with people on the research distribution list. Early access, no spam, unsubscribe anytime.

Run your brand

Free · 60 sec · benchmarks live

Run free audit

Explore the data

Email · 30 industry × city slots

Open the explorer

Layer 3 trial

25–30 spots · 60-day intervention

Apply now