Same chain, different categoryHow Google Maps’ primary_type field quietly filters hotels out of AI answers

The Generator portfolio: four properties, four different category outcomes

Generator Hostels is a single chain operating design-forward properties across European capitals. Same brand, same product playbook, broadly comparable guest ratings. Yet here’s how Google Maps tags four of them this week:

The signal isn’t star count — it’s the type

We initially expected the “1-star” numeric field (hotel_classification) to be the smoking gun. It isn’t. Spot-check from this week’s data:

Burj Al Arab as a 1-star is a clear giveaway: the field isn’t a credible quality tier. It’s sparsely populated, inconsistently filled, and not the input AI assistants reliably filter on.

The categorical tag — primary_type in Google Places, the “Hotel / Hostel / Resort hotel / Bed & breakfast” chip you see on a Maps listing — is the more decisive filter. It’s the attribute every AI assistant we’ve looked at gets back from Google Places, and it’s used both as a categorical tag in answers and as a filter when users constrain by type.

And: most hotels have no class at all

Of the 348 hotels mentioned by AI assistants this week, 81.3% have no hotel_classification value at all in our index. Only 1.1% are classed 1 or 2. The 19% that do have a class span every value 1–5. So if any AI model treats the numeric class as a hard quality gate, it’s either filtering out four-fifths of the inventory or quietly ignoring the field. Either way, the categorical primary_type is doing more work than the numeric class.

Other cases we looked at

Beyond the Generator family, six other properties from this week’s sample. Some are correctly classed (Hôtel Amour at class 4, The Standard London at class 5). Some are clearly under-typed (The Hoxton Paris at class 1; 25hours Dubai with no class despite 4.6☆). The pattern isn’t consistent — which is part of the point.

Top under-classed candidates (25)

Hotels in our index with class 1 or 2 but a 4.0☆+ guest rating and 4,000+ reviews. Sort signal: google_reviews_count desc. The first row tells the story: Atlantis — The Palm Dubai, classed 1 in our data, 4.7☆ from 100,000+ guests. If anything filters on the numeric class, this is what gets dropped.

What the full snapshot will cover

This is a draft. The full piece will widen the sample, run prompts, and observe whether category mismatches actually move AI answers. Specifically:

What to do in the meantime

If you run a hotel and you suspect you’re mis-typed:

1. Open your Google Business Profile and look at Category. If the primary category is “Hostel,” “Bed & breakfast,” or anything narrower than what you actually offer, change the primary to Hotel (or Boutique hotel / Resort hotel). Add the rest as secondary categories. Generator’s portfolio split is the test case for what happens when this isn’t consistent across a chain.
2. Publish your own schema.org/Hotel JSON-LD. Use the explicit @type: Hotel (or Resort, BedAndBreakfast) and fill in starRating if you have an official one. When AI assistants cross-check with on-site schema, you want to be saying the same thing as Google. Our free schema generator produces this in 30 seconds.
3. Test it yourself. Ask three AI assistants “best hotels in [your city]” and see whether your property appears. If it doesn’t, the type mismatch is one of the likeliest reasons. Worth checking roughly monthly.

Caveats

• This is a draft. Numbers are from a single weekly snapshot (2026-04-27). The framing — that primary_type is the decisive filter — is plausible but not yet verified by running prompts and observing answer-set shifts.
• hotel_classification is inconsistent. Google’s own data has Aman Tokyo, Atlantis Palm Dubai, and Burj Al Arab labeled “1.” The numeric class field is a useful headline but not a credible quality tier. The real underlying signal is the categorical primary_type.
• Some properties are deliberately positioned this way. Generator markets itself as a hostel chain in some markets and as a hotel/lifestyle brand in others. Not every type mismatch is an error — some are intentional segment choices. Worth checking which is which before recommending changes.
• Star ratings are messy by design. Different countries have different official rating bodies; some boutique hotels avoid the system entirely; many tourist boards don’t hand out ratings to hostels at all. The primary_type field is a categorical bin, not a quality score.