Expert Advice

Your 5-Star Rating Is Costing You

You’ve spent the last year pushing your rating from 4.7 to 5.0. Responding to every review within 24 hours.

By Conor Healy

April 13, 2026

You’ve spent the last year pushing your rating from 4.7 to 5.0. Responding to every review within 24 hours. You train your staff on service recovery. Following up with unhappy customers before they post. You finally hit that perfect score.

And then you notice something strange. Your traffic from AI-powered search has dropped.

Most businesses assume 5 stars is the goal. The research says otherwise. PowerReviews analysed 20 million product pages and found that conversion rates peak between 4.75 and 4.99 stars. Products with perfect 5.0 ratings convert at roughly the same rate as those sitting between 3.0 and 3.49. Nearly half of all shoppers, and more than half of Gen Z, actively distrust perfect scores.

Consumers learned this years ago. Now AI has learned it too. New research from the Scientific Institute for Generative Intelligence (SIGI) shows exactly how.

The 4.9 Effect

A 4.9 rating generates more favourable AI language than a perfect 5.0.

The Scientific Institute for Generative Intelligence (SIGI) studied how large language models respond to different star ratings when asked to recommend businesses. At 4.9 stars, AI systems produce what researchers call “credentialing language”: phrases emphasising quality verification, earned trust, and consistency of service. The AI treats the near-perfect score as evidence that needs explaining.

At 5.0, the AI shifts to “brand-naming language.” It focuses on identity and recognition. It stops discussing quality evidence. The perfect score becomes a label rather than a credential.

Credentialing language sells. Brand-naming language identifies.

Trained on Human Scepticism

AI systems are trained on human-generated text: decades of consumer research, marketing literature, and purchasing behaviour data. They absorbed what humans already knew.

Northwestern University’s Spiegel Research Center analysed review behaviour across 40 product categories. Purchase likelihood peaked between 4.2 and 4.5 stars, then declined as ratings approached 5.0. The researchers found that consumers perceive ratings close to a perfect 5 as “too good to be true.”

PowerReviews confirmed this with harder numbers. Their analysis of 20 million product pages found that 46% of shoppers are suspicious of perfect 5 star reviews. Among Gen Z shoppers, that figure rises to 53%. The sweet spot for conversion sits between 4.75 and 4.99 stars.

AI reproduces the scepticism.

How AI Reads the Rating Spectrum

SIGI mapped how AI sentiment shifts across the full rating spectrum. The patterns are consistent and specific.

Ratings below 3.8 stars trigger warning language. The AI hedges, suggests alternatives, flags concerns. At 3.8, sentiment turns neutral. Between 4.0 and 4.6, recommendations become moderately positive. At 4.7 to 4.9, AI language hits peak enthusiasm with strong credentialing. At 5.0, the enthusiasm flattens. Quality discussion fades. The AI names you but stops vouching for you.

SIGI tested these thresholds repeatedly. The 3.8 floor and 4.7 ceiling held across different prompt framings and business categories. The 5.0 flattening was not an anomaly.

The drop from 4.9 to 5.0 is not catastrophic. But businesses fighting for AI visibility are competing on margins, and this margin favours the almost-perfect.

Review Count Moderates Rating

Star ratings don’t exist in isolation. SIGI found that review count changes how AI processes them.

A 5.0 rating with fewer than 25 reviews triggers scepticism modifiers. The AI questions whether the score means anything. A 4.9 rating with 100 or more reviews triggers the strongest credentialing response. High volume plus near-perfection reads as the most reliable quality signal.

SIGI’s tests showed the interaction clearly: a 3.8 rating with 1,000 reviews gets processed as neutral. A 5.0 with 3 reviews gets processed as negative. Volume rescues mediocre ratings and undermines suspicious ones.

New businesses face a hard trade-off. Chase perfect scores early, when volume is low, and you may be hurting your AI visibility when you most need traction. A 4.7 with 50 genuine reviews may outperform a 5.0 with 8.

Negative Reviews as Trust Signals

A few negative reviews might help more than hurt.

82% of shoppers actively seek out negative reviews. Mixed feedback signals authenticity. A business with nothing but praise looks curated. A business with occasional criticism looks real.

AI has absorbed the same logic. 5 star reviews in quantity start to look like a filter rather than a fact. Near-perfect ratings with visible imperfection look like the truth.

This is not an argument for tanking your own reviews. It is an argument against obsessing over 5.0. The effort spent pushing from 4.8 to 5.0 might be better spent elsewhere.

AI Citation Reinforces Itself

AI citation behaviour creates reinforcement effects. Businesses that AI recommends get more traffic, more customers, and more reviews. Those reviews feed back into AI training data. The cycle builds on itself.

If AI favours 4.9 over 5.0, that preference accumulates. Businesses stuck at perfect ratings may find themselves slowly edged out by competitors with more believable scores.

SIGI observed this pattern directly in their citation analysis: entities that get cited tend to get cited more. Those that don’t fall further behind. The gap widens over time, not narrows.

Early AI visibility matters. Getting the signals right now sets the trajectory.

5 Stars Still Has a Place

Perfect ratings are not worthless. They still impress humans scanning a Google listing. They still matter for traditional search. Usually they are not actively harmful in most contexts.

But in the growing slice of discovery that runs through AI, 5 star reviews carry less weight. The machines learned our scepticism. They apply it at scale.

Credible beats flawless. Real beats perfect. 4.9 beats 5.0.