Hotelist Ratings Data: Yelp Tops Airbnb on Honest Distribution

What happened

Pieter Levels, the solo founder behind a long string of indie products, has followed up the launch of Hotelist.com with numbers meant to back up its premise. Hotelist uses AI to aggregate hotel reviews and counter the inflated scores common on booking platforms 2. Levels first framed it as a fix for Airbnb's rating scale, which he describes as running from 4.5 to 5.0 5.

His new analysis, published on a Hotelist stats page that breaks down ratings by site, makes three claims 3. First, Airbnb scores span only 4.5 to 5 with a median of 5 3. Second, Yelp is the most trustworthy of the platforms he examined because its ratings form an actual normal distribution 3. Third, Hotelist itself produces a normal distribution with a median of 3.5 3. Levels says that was a deliberate target, since he considers most stays average and worthy of about a 3 3.

The compression problem

The complaint is not new. Levels has been making it publicly. In late August he argued that Airbnb deletes negative reviews and that Google Maps strips them in many countries as defamation. He added that booking sites have every incentive to do the same, which leaves everything hovering around 4.7 4.

Independent figures circulating on X largely support the shape of his argument, though not every detail. One cross-city analysis of Airbnb listings found 65% sitting at a perfect 5.0 4. The median was 4.84, which displays as a rounded-up score 4. That is a slight divergence from Levels' flat "median of 5" claim, though the practical effect for a guest scrolling listings is much the same. The same analysis put TripAdvisor's ratings in a wider spread around 4.0, with only about 30% reaching 4.5 or higher 4. A 2021 study cited in that discussion found 95% of Airbnb listings at 4.5 or above 4.

The explanations also differ. Levels points to outright deletion 4. Others point to subtler mechanics, such as hosts and guests withdrawing positive reviews in retaliation for negative ones 4. Airbnb itself frames its removals as taking down feedback that violates policy in order to keep reviews authentic 4. Users have shared anecdotes of nasty surprises at top-rated properties, from black mold to meat stolen from a freezer 4.

Why it matters

Levels' core point is about trust. When negative reviews disappear, people feel anxious about booking because they can't verify a listing 2. A five-star scale that only uses its top half carries little information.

Researcher Ivan Vendrov raised the more interesting question. He called it an incentive design problem he would have expected e-commerce to solve after 25 years 2. Suppressing negative reviews clearly costs consumers over the long run, so why haven't profitable "honest info-brokers" emerged 2? He also suggested that collapsing a hospitality experience into a single number may not be necessary at all 2. That is a quiet challenge to Hotelist's own approach of re-scoring on one axis.

The distribution behind the distribution

The follow-up is also typical of how Levels operates. His playbook relies on shipping fast, sharing revenue and data openly, and posting relentlessly. One profile credits a decade of this, across more than 40 products, with building a following large enough to make each launch an instant event 1. That profile cites 600K followers in one place and 350K in another 1. It says Photo AI made $5.4K in its first week in 2023 and reached $132K in monthly recurring revenue by month 18 1. Levels' own post captions make the strategy explicit: "Never stop talking about hotelist.com until everyone finds out" 3. Hotelist is also just one item in a busy stream. Around the same time, he was building an AI live-stream experiment and adding census data to Hoodmaps 5.

Our reading

The Airbnb finding holds up. Several independent figures point to the same severe compression, and the Yelp observation is a useful reminder that platform design choices shape score distributions.

The Hotelist result deserves more caution. A median of 3.5 shows the model hits the distribution its creator aimed for 3. It does not show that those scores are more accurate. A normal curve tells you how ratings are spread out, not whether they are right. The analysis also comes from the product's owner and appears on the product's own stats page 3.

Vendrov's question remains open: whether honest ratings can sustain a business when every platform's revenue depends on bookings 2. Hotelist is a credible attempt at an answer. Its proof will come from travelers finding its 3s and 4s more useful than Airbnb's wall of 5s, not from its histogram.