Star ratings don't get you recommended by AI. Review volume does.

1. The rating race is already over — everyone won

Among the businesses that rank, ratings barely vary: the 10th percentile is 4.6 and the 90th is 4.9. When 74% of your competitors share the same 4.8 stars, the rating stops being a tiebreaker. AI engines and the Map Pack have to look at something else to decide who to name — and that something is how many people have weighed in.

2. Review volume predicts visibility ~2× better than rating

We correlated each signal against our composite 14-point visibility score (the same score that drives our city rankings). Review count — measured on a log scale, because it spans from a couple hundred to 17,000 — is the stronger predictor by a wide margin.

3. The counterintuitive part: more reviews, slightly lower stars

Rating and review count are negatively correlated in our data (-0.33). The businesses with thousands of reviews tend to carry a 4.7–4.8, not a 5.0 — high volume pulls an average toward the mean, and a real business serving thousands of customers will collect some unhappy ones. That's fine. A near-perfect rating with a tiny review count reads as "new and unproven" to both customers and the engines that rank them. Volume is the trust signal; chasing a flawless 5.0 by suppressing volume is the wrong trade.

4. How many reviews it takes to be a top pick, by industry

Median review count of the top-ranked businesses in each category — a rough "table stakes" bar for being in the conversation.

Median Google review count per category · n = 630 businesses with public Google data.

What this means for your business

• Stop optimizing for a perfect score. If you're above ~4.6 stars you're already where you need to be. Don't suppress reviews to protect a 5.0.
• Build a review-velocity habit. The gap between you and the top pick in your city is almost always volume, not stars. Find your category's bar above and close the distance.
• Make the reviews legible to AI. Volume gets you considered; structured profile data, citations, and listicle presence get you named. That's what our 14-point audit measures.

Methodology & limits

The dataset is every business featured in our published city rankings — 730 businesses, 630 with public Google rating and review data, 229 carried through the full 14-point audit. Google data is pulled via Bright Data Maps. Correlations use Pearson's r with review counts on a log scale. Every number on this page is regenerated from source by scripts/build-research-stats.mjs.

What this is not: a controlled experiment. This is observational data about businesses that already rank, so it describes what top performers have in common — not a proven causal lift from adding reviews. It's also a deliberately curated sample (top picks, not a random draw), which is why ratings look so uniformly high. We're publishing the shape we see and the method behind it; treat it as a strong directional signal, not a law.