Case study — design hotel

Mosaic House Design Hotel

Design / boutique hotel · Prague, Czechia

The best-performing hotel tracked on the full 5-engine panel was still named in fewer than 1 of every 8 checks, and completely missing from one major AI engine.

12.5%

Detected in 25 of 200 real, non-branded traveler questions

60.9%

AI readiness score (technical AI-crawler readiness)

Mosaic House Design Hotel in Prague had the highest AI detection rate among the three hotels tracked on the full 5-engine panel — 12.5% (25 of 200 checks) — but still scored 0% on Google AI Overviews.

Key takeaways

  • The strongest result among the three hotels tracked on the full 5-engine, ~200-300-check panel: 12.5% detection across 200 non-branded checks — meaning it was still only named about 1 in 8 times.
  • Zero mentions on Google AI Overviews — the same result 3 of the 4 hotels in this set got; only Bouda Mama had a single AI Overviews mention.
  • Mosaic House has a lower technical readiness score (60.87%) than Golden Well (77.14%) — and yet a meaningfully higher detection rate, the clearest evidence in this data set that readiness alone doesn't determine visibility.
  • Gemini was by far its best engine, mentioning it in over a quarter of checks; ChatGPT was strong too.
  • On ChatGPT specifically — the one engine every hotel in this set is tracked on, regardless of plan tier — Mosaic House still leads at 20%, ahead of Bouda Mama's 16.7%, Golden Well's 8.3%, and Olympia's 0%. That's the fairest like-for-like comparison across all four hotels, and Mosaic House wins it too.
  • It shows up reliably for "unique," "boutique," and "quiet" hotel searches, but disappears for straightforward searches like "hotels in Prague city centre."
  • Booking.com, TripAdvisor, Expedia, and a hotel-editorial site got cited by AI engines far more often than Mosaic House's own website.

Hotel Snapshot

Mosaic House is a design-led boutique hotel in Prague, positioned on personality and aesthetic rather than scale — the kind of property that travel bloggers and design-conscious travelers actively seek out by name. It's a useful example because it produced the best raw numbers of the three hotels tracked on the full 5-engine panel, while still landing well under a 1-in-8 mention rate and completely missing one major AI channel — proof that even a comparative "winner" here has real room to close.

The Challenge

Design-forward, independent hotels like Mosaic House often win on word-of-mouth and editorial coverage, but that reputation doesn't automatically transfer into what an AI engine says when a traveler asks a generic question. Mosaic House didn't have visibility into whether its strong offline reputation was actually showing up when guests asked AI tools where to stay in Prague.

What We Found

Across 200 real, non-branded traveler questions, Mosaic House was mentioned 25 times — a 12.5% detection rate, the best of the three hotels tracked on the full 5-engine, ~200-300-check panel (Bouda Mama also reaches 12.5%, but on a much smaller 24-check, 2-engine panel — see the note on comparing across plans below). Gemini was the standout engine, naming the hotel in 27.5% of checks, and ChatGPT followed at 20%. Perplexity mentioned it in 12.5% of checks, Claude in just 2.5%, and Google AI Overviews in none at all. The hotel showed up consistently for atmosphere- and identity-led searches — "unique hotels to stay in Prague," "boutique hotel near Prague city centre," "hotels in Prague recommended by travel bloggers" — but was absent from more generic, amenity-based searches like "hotels with a pool in Prague" or simple location searches like "hotels in Prague city centre."

Engine-by-Engine Breakdown

The table below breaks down Mosaic House's detection rate by engine — Gemini was its clear strongest channel (11 of 40), ChatGPT a strong second (8 of 40), and Perplexity a moderate presence (5 of 40). Claude was nearly absent (1 of 40), and Google AI Overviews returned zero mentions across 40 checks.

AI engineDetectedDetection rate
Gemini11 of 4027.5%
ChatGPT8 of 4020%
Perplexity5 of 4012.5%
Claude1 of 402.5%
Google AI Overviews0 of 400%

Who's Getting Cited Instead

When Mosaic House wasn't named, AI engines consistently pointed travelers toward Booking.com, TripAdvisor, Expedia, and the hotel-editorial site thehotelguru.com. These are the same kinds of intermediaries that dominated the Golden Well results too — a sign that in Prague specifically, AI engines default to large, well-established booking and editorial platforms over any individual hotel's own site, even a hotel with genuinely distinctive design credentials.

The Pattern This Reveals

Mosaic House is the anchor data point for one of the most important findings across this whole set: technical AI-readiness does not reliably predict how often AI engines actually recommend a hotel. Mosaic House's readiness score (60.87%) is notably lower than Golden Well's (77.14%) — yet Mosaic House was detected far more often (12.5% vs. 7.12%), a comparison that's fair because both hotels are tracked on the same full 5-engine panel. At the same time, even this best-in-set result still means AI engines named Mosaic House in fewer than 1 of every 8 checks, and it joins Golden Well and Olympia at a flat 0% on Google AI Overviews for generic questions — only Bouda Mama, on a much smaller panel, got a single AI Overviews mention. Bouda Mama's overall 12.5% detection rate technically matches Mosaic House's, but the two aren't directly comparable — Bouda Mama's number comes from 24 checks on 2 engines, Mosaic House's from 200 checks on 5. On the one comparison that is fair to all four hotels — ChatGPT detection, the single engine every plan tier includes — Mosaic House still leads outright (20% vs. Bouda Mama's 16.7%, Golden Well's 8.3%, Olympia's 0%). The lesson isn't "readiness doesn't matter" — it's that readiness is table stakes, not the whole game.

What This Means For Your Hotel

Even the best-performing hotel on the full 5-engine panel has real, specific gaps — a whole major AI engine where it's invisible, and clear categories of guest question it never appears for. If that's true for the strongest result in this data set, it's worth finding out exactly where your own hotel stands rather than assuming a strong offline reputation is carrying over automatically. Run a free check on your hotel to see your real numbers, question by question.

At a glance

Plan tier tracked
Pro
Overall detection rate
12.5% (25 of 200)
Engine coverage
5 of 5 engines (Pro plan)
Snapshot type
Single scan, one point in time — not a trend

Panel size and engine coverage vary by plan tier, so headline detection rates across hotels in this data set aren't a precise ranking — see the note on comparing across plans for a fair, same-engine comparison.

Questions

Does a high AI readiness score guarantee AI visibility?

No — and Mosaic House is the clearest proof of that in this data set. Golden Well Hotel has the higher technical readiness score (77.14% vs. Mosaic House's 60.87%), meaning its site is more thoroughly built for AI crawlers to read. But Mosaic House was actually detected more often by AI engines (12.5% vs. 7.12%). Readiness is a real, useful signal, but it's one input among several — it doesn't determine the outcome on its own.

Why do OTAs like Booking.com get cited by AI engines more than a hotel's own website?

Booking sites and travel platforms tend to have extensive, structured, frequently updated content covering thousands of hotels in a consistent format, which makes them easy and reliable sources for AI engines to pull from. An individual hotel's own website usually covers just one property, updated less often and in a less standardized way — so even when a hotel is the right answer, the AI engine may find it easier to cite the platform that lists it.

How does a boutique hotel compete with big chains in AI-generated recommendations?

Mosaic House shows it's possible: it out-performed the other hotels in this set on identity- and atmosphere-driven searches, where its distinct design positioning clearly helped it get named. The gap is on broader, generic searches, where AI engines default to platforms rather than any individual property, chain or independent. The path forward is winning the specific-question categories a boutique hotel is naturally suited for, while working to close the gap on the generic ones.

How was this measured?

We tracked how AI engines answered a large set of real, non-branded traveler questions about hotels in the property's destination — the kind of question a guest asks before they know which hotel they want — and recorded whether and how often each hotel was actually mentioned in the response, engine by engine.

Bouda Mama's overall rate is also 12.5% — is it directly comparable to Mosaic House's?

No, and this is worth being precise about. Mosaic House's 12.5% comes from 200 checks across all 5 AI engines; Bouda Mama's identical 12.5% comes from just 24 checks across 2 engines (it's tracked on Staylight's entry-level Radar plan). A smaller panel means far more statistical noise — on Bouda Mama's panel, one extra or missing detection swings the rate by roughly 4 percentage points; on Mosaic House's, one detection only moves it by about 0.5 points. The one comparison that is fair across all four hotels regardless of plan tier is ChatGPT detection alone, since every hotel is tracked on it: there, Mosaic House leads outright at 20%, ahead of Bouda Mama's 16.7%.

See where your own hotel stands.

The same check, on your property — free, no credit card, results in minutes.