In Croatia, a three-star hotel and a five-star hotel get the same median guest score: 9.3 out of 10. Not close. Identical — 233 three-stars against 40 five-stars, every one of them with at least 100 reviews on the books.
That number came out of a bigger question we wanted to settle with data instead of vibes: what does the star plaque on the wall actually predict about your stay? So we took every hotel in the HotelScout catalog that carries an official 1–5 star classification, a guest score, and at least 100 reviews — 9,072 properties, 13,128,678 guest reviews between them — and lined the two systems up against each other.
They agree far less than the price difference suggests.
The whole ladder is worth 1.3 points
Climb from the bottom of the star system to the top and the median guest score moves from 8.0 to 9.3. That's it. Four full tiers of classification, 1.3 points of measured satisfaction.
| Tier | Hotels | Median guest score | Review-weighted average |
|---|---|---|---|
| 1★ | 168 | 8.0 | 7.69 |
| 2★ | 718 | 8.1 | 7.94 |
| 3★ | 2,977 | 8.5 | 8.27 |
| 4★ | 3,811 | 8.9 | 8.65 |
| 5★ | 1,398 | 9.3 | 9.04 |
Look at the right-hand column. Weight each hotel by its review count and every tier scores lower: five-stars drop from 9.3 to 9.04. The big properties drag their own class down. The 400-room property that dominates the booking results is statistically the weakest version of its own star class.
The overlap nobody prices in
If stars worked, the tiers would stack cleanly. They don't. 599 of the 2,977 three-star hotels — 20.1% — score at or above 9.3, the median five-star. Those 599 aren't obscure either: they carry 261,795 reviews between them.
It cuts the other way too. One in twelve five-star hotels (119 of 1,398) scores below 8.5 — the median three-star. And 37 five-stars sit below 8.0, which means they're rated worse than the average one-star hotel (7.96).
Even the bottom of the ladder joins in: 108 of the 886 one- and two-star properties in the sample meet or beat the five-star median.
A five-star plaque is a probability, not a promise. Roughly 8% of the time, the guests who actually slept there rate it like a mid three-star.
Croatia flattens the ladder completely
Back to that opening stat. In Croatia, the three-star median and the five-star median are both 9.3. The averages barely separate them: 9.15 versus 9.26. Whatever Croatian three-stars are doing (small, family-run, obsessively kept), the star system can't see it.
Dubrovnik makes it concrete. The city's 45 qualifying three-star hotels hold a median of 9.20. That single number beats the five-star medians of the United States (8.90), Japan (9.00), Spain (9.10) and Portugal (9.10), and ties the five-star median of France (9.20).
A Dubrovnik three-star, statistically, is a French five-star.
Where stars still mean something — and where the tier is a lottery
In some countries the plaque still earns its keep. The gap between tiers varies wildly by country:
| Country | 3★ median | 5★ median | Gap |
|---|---|---|---|
| Morocco | 7.90 | 9.30 | 1.40 |
| France | 8.20 | 9.20 | 1.00 |
| Japan | 8.20 | 9.00 | 0.80 |
| Thailand | 8.45 | 9.20 | 0.75 |
| United States | 8.20 | 8.90 | 0.70 |
| Italy | 8.60 | 9.30 | 0.70 |
| Spain | 8.40 | 9.10 | 0.70 |
| Portugal | 8.50 | 9.10 | 0.60 |
| Greece | 9.00 | 9.30 | 0.30 |
| Croatia | 9.30 | 9.30 | 0.00 |
(Countries with 200+ qualifying hotels, 50+ three-stars and 30+ five-stars.)
In Morocco the plaque carries real information — 1.4 points of it, the widest gap we measured. In Greece it's worth 0.3 points. In Croatia, nothing.
But Turkey's five-star tier is the widest lottery in the dataset: 23 of its 88 five-stars (26%) score below 8.5, below the global three-star median. Morocco is just behind at 23 of 94 (24%). And of the six lowest-rated five-star hotels with 1,000+ reviews anywhere in the catalog, four are Moroccan: Hotel Borj Rabat (6.5 across 3,129 reviews), Grand Mogador Agdal & Spa in Marrakech (7.2 / 3,769), Royal Mirage Deluxe in Marrakech (7.2 / 2,108) and Ramada by Wyndham Fes (7.6 / 1,294).
To be fair to Morocco: its five-star median is still 9.3, tied for the highest in the table. The tier isn't bad. It's wide. For a traveler paying five-star money, that's its own problem.
Why the plaque misleads: stars are self-reported
Here's the mechanism. Star classifications in booking data are declared by the properties themselves, and the declaring is done with different levels of restraint in different places.
Among the 14 countries with 200+ classified hotels, the five-star share of classified properties spans a near-eightfold range: 37.2% in Turkey, 4.8% in Croatia. Same plaque, wildly different scarcity.
The extremes are in the desert. In Wadi Rum, Jordan, 100% of starred properties (all 34) classify themselves five-star. In Merzouga, the dune town on the edge of the Moroccan Sahara, it's 79 of 93, or 84.9%. These are tent camps, and by their own declaration nearly every one of them out-ranks the Ritz.
That's not fraud, exactly. It's a rating system with no referee. And it's why a Croatian five-star and a Turkish five-star are different objects wearing the same badge.
The three-stars doing five-star numbers
59 three-star hotels hold a 9.3+ score across 1,000+ reviews each, sample sizes where a fluke can't hide. The top of that list: Kailash Boutique Hotel in Kathmandu (9.9 over 1,569 reviews), Villa Flavia Heritage Boutique Hotel in Plovdiv (9.9 over 1,367) and Muni Residence & Spa in Battambang (9.9 over 1,329).
And at the volume end, the mass guest experience is a three-star. None of the five most-reviewed hotels in the catalog is a five-star: the Excalibur in Las Vegas (3★, 30,613 reviews, scoring 7.6), Villa Fontaine Grand Haneda at Tokyo's airport (4★, 26,212), the APA towers in Osaka Umeda and Higashi Shinjuku (both 3★), and the Luxor in Las Vegas (3★, 23,914 reviews, 7.4). The most-reviewed five-star anywhere in the catalog, Marina Bay Sands in Singapore, has 17,315 reviews. The Excalibur beats it by more than 13,000 reviews.
The upgrade that buys the least
One more number for anyone hovering over the four-star filter: 1,141 of the 3,811 four-star hotels (30%) already meet or beat the five-star median. The step from four stars to five is the least informative move on the entire ladder.
So flip your filters. Sort by guest score first, set a review-count floor of 100 or so, and treat the stars as a tiebreaker — a rough index of facilities, not of whether you'll be glad you stayed. The 599 three-stars posting five-star numbers are cheaper to find than they've ever been. The plaque just won't point you to them.
How we ran the numbers
We analyzed 30,505 published hotel listings in the HotelScout catalog (snapshot: 14 July 2026) across 169 countries. To compare official star classification against guest satisfaction, we restricted the study to hotels with an official 1–5 star rating, a guest score (0–10 scale, aggregated from verified stays), and at least 100 guest reviews: 9,072 hotels backed by 13,128,678 reviews. Hotels with no star classification or fewer than 100 reviews were excluded.
Country-level claims require at least 200 qualifying hotels per country (and 50+ three-star / 30+ five-star properties for tier comparisons); city-level claims require at least 30.
We report medians rather than means to resist outliers, and compare star tiers only within the same catalog and, for country claims, within the same country. Cross-country "happiest guests" rankings are deliberately avoided because catalog composition differs by market. Star-share figures (Turkey 37.2%, Croatia 4.8%, Wadi Rum, Merzouga) are shares of the 13,829 hotels carrying a star classification, never of all hotels.
Those classifications are self-reported by properties to booking platforms, which is precisely the mechanism this study documents. Star ratings throughout are those displayed on booking platforms - not national hotel registers, which classify far fewer properties: 'hotels' here spans every lodging type carrying a platform star rating (aparthotels, guesthouses and similar included), and city figures cover the greater urban area rather than administrative city limits. The most-reviewed-hotel list is a census of this 30,505-hotel catalog, not a world-record claim. No price data was used or implied. Questions about the data or requests for country-level cuts: the full methodology and queries are reproducible on request.
