Most AI search tools track the businesses AI names. ARO Index™ defines the full market cohort first, then records which businesses AI selects, and which it never does. In Charleston, that meant 688 businesses audited before a single ranking query ran. 89 were ever selected in a cold buyer query. The other 599 are in the dataset too. That denominator is the research.
Ranking tools build their universe from whatever names AI produces. ARO Index locks the market cohort before querying. You can't measure who's left out if your dataset starts with who's included.
Charleston 2026: 688 audited, 89 selected, 12.9%. The not-selected count is a first-class number, not a blind spot. Every figure traces to a locked cohort and a published matching query anyone can rerun.
A business counts as recommended only when a model places it top 3 for a real buyer query, 2 of 3 phrasings, per model. Being named in a list of twenty is not a recommendation.
The local AI ranking category is real now. Several products publish local rankings. The question is no longer whether these datasets exist. It's which one is built like research.
| Feature | ARO Index™Market research dataset | Courtyard IndexWeekly public rankings | MentionedOnPer-business monitoring | AIVisibilityIndex.netOne-time reports | ProfoundEnterprise brand tracking |
|---|---|---|---|---|---|
| Market cohort defined before querying The full market is enumerated first, so zero-selection businesses are in the dataset | ✓ Yes | ✕ NoUniverse built from businesses AI names | ✕ NoTracks the subscribing business | ✕ No | ✕ No |
| Published not-selected denominator Reports state how many businesses in the market were never selected | ✓ Yes | ✕ No | ✕ No | ✕ No | ✕ No |
| Versioned methodology with disclosed model access Method versions, model strings, retrieval state, and reproducible matching queries published | ✓ Yes | ~ PartialMethodology published, not versioned with reproducible queries | ~ PartialMethodology page, monitoring-oriented | ✕ Not published | ✕ No |
| Selection standard Counts top-3 choice, not any mention | ✓ Yes | ~ PartialShare-of-voice weighting across mentions | ~ PartialMention rate + score | ? UnclearUnclear from public info | ✕ NoMention/citation share |
| Local rankings by city and category Ranked leaderboards by city and category based on actual AI recommendation behavior | ✓ Yes | ✓ Yes10 metros, 5 industries, weekly | ✓ YesCity leaderboards | ✓ YesReport-based | ✕ NoNational/enterprise |
| Daily monitoring for a single business Continuous per-business alerting rather than a market-level cadence | ✕ NoMarket cadence is monthly; higher frequency is a paid research tier | ✕ No | ✓ YesDaily checks from $99/mo | ✕ NoOne-time | ✓ YesEnterprise |
*Based on each product's public site as of September 2026. Corrections welcome: therese@aroindex.com
Two years ago, no one published local AI recommendation rankings. Now several products do, and that's good news. It means the question every local business is starting to ask ("does AI recommend me?") has been validated as a market.
But a ranking and a market index are different instruments. A ranking starts from the names AI produces and orders them. A market index starts from the market, every qualifying business, enumerated and locked before the first query, and records what AI does with all of them. Only the second design can tell you the number that matters most: how many businesses AI never selects at all.
In Charleston, that number was 87 in 100. No dataset built from AI's own answers can contain it.
What's the difference between ARO Index and tools like Courtyard or MentionedOn?
Courtyard publishes weekly rankings of the businesses AI names across 10 metros. MentionedOn monitors one business's AI answers daily and publishes city leaderboards. Both are built from AI's outputs. ARO Index defines the full market cohort first, then measures selection against it, so businesses AI never names are measured, not missing. Different designs, different questions answered.
Why does the denominator matter?
Because "ranked #4" means something different in a market of 12 selected businesses than in a market of 688 audited ones. Without a locked cohort, a ranking can't tell you selection rate, can't track who fell out, and can't compare across time without survivorship bias. ARO Index cohorts are frozen snapshots with published membership rules.
What is an ARO Score™?
ARO Score™ is the 0-100 diagnostic composite inside an ARO Report: recommendation rate (40%), position strength (20%), confidence (20%), and signal quality (20%). It explains what may be driving a business's public ARO Index Score. Full formulas and model access details are on the methodology page.
Is ARO Index a monitoring tool?
No. Monitoring tools check one business's answers daily and alert on change. MentionedOn does that well at $99/mo. ARO Index is market research infrastructure: locked cohorts, versioned methods, published denominators, monthly market cadence. Agencies use both for different jobs.
Do other tools publish local AI rankings?
Yes, several do now. What ARO Index publishes that rankings built from AI outputs cannot: the businesses AI never selects, measured against a market cohort defined before the queries ran.
Run a free audit and see how ChatGPT, Claude, Gemini, and Perplexity currently select, or skip, your business. No credit card. No pitch. Just the data.