ChatGPT Is Diversifying
Its Hotel Data Sources
We tracked 100K map entities over 90 days. Google dropped from 100% to 70.3%. TripAdvisor appeared with descriptions 8.8x longer. The multi-source era has begun.
TL;DR
ChatGPT's hotel search is no longer Google-only. In 90 days, Google Places dropped from 100% to 70.3% of map entities as three new providers entered: Yelp (10.2%), Foursquare (0.2%), and TripAdvisor (0.1%). TripAdvisor descriptions are 8.8x longer than Google's (855 vs 97 chars). And critically: TripAdvisor is the first provider where ChatGPT links to an intermediary instead of the hotel's own website. Google and Yelp both link to hotel direct sites — TripAdvisor links to TripAdvisor pages.
Summary
We tracked 99,538 map entities shown in ChatGPT Search results across 25,645 hotel queries from December 2025 to March 2026. The findings reveal a dramatic shift: Google Places' share dropped from 100% in December 2025 to 70.3% in the last week of March 2026, as OpenAI integrated three new data providers.
Yelp is the established second source at 10.2% of all entities, concentrated in US cities (Las Vegas 40.8%, San Francisco 34.9%). TripAdvisor first appeared February 20 and is accelerating — reaching 1.0% in the final week with descriptions 8.8x longer than Google's. Foursquare emerged in late February and hit 3.9% in the final week.
Crucially, TripAdvisor also introduces a new traffic model: it is the first provider where ChatGPT links to an intermediary rather than to the hotel's own website. Google and Yelp both send users to hotel direct sites. TripAdvisor sends them to TripAdvisor pages — a fundamental shift for hotel direct bookings.
This is the second chapter in our tracking of ChatGPT's hotel data strategy. The Yelp study documented the first non-Google provider in January. This study documents the transition from "Google plus one" to a genuine multi-source data layer.
Provider Market Share
Google still dominates at 88.8% overall, but this aggregate masks the trend. In December 2025, Google was 100%. By the last week of March 2026, it had dropped to 70.3%.
| Provider | Entities | Share |
|---|---|---|
| Google Places | 88,425 | 88.8% |
| Yelp | 10,148 | 10.2% |
| SerpAPI | 741 | 0.7% |
| Foursquare | 156 | 0.2% |
| TripAdvisor | 66 | 0.1% |
The aggregate hides the trend.
Google's 88.8% overall share includes 3 months of data where it was 100%. The weekly trend below tells the real story — Google is at 70.3% and falling in the latest data.
Weekly Trend: Google's Decline
The headline finding. Google Places went from 100% of ChatGPT map entities in December 2025 to 70.3% by the last week of March 2026 — a 30-point drop in 90 days. Three distinct phases emerge.
100% Google Places. ChatGPT map cards are exclusively powered by Google.
Yelp appears and stabilizes at ~10-12%. Google drops to ~88%. Destination-dependent.
Foursquare and TripAdvisor enter. Google drops to 70.3%. Diversification accelerates.
| Week | Entities | Yelp | Foursquare | TripAdvisor | |
|---|---|---|---|---|---|
| Dec 22 | 376 | 100% | 0% | 0% | 0% |
| Dec 29 | 996 | 100% | 0% | 0% | 0% |
| Jan 5 | 1,184 | 100% | 0% | 0% | 0% |
| Jan 12 | 4,135 | 88.6% | 11.4% | 0% | 0% |
| Jan 19 | 7,125 | 87.5% | 12.1% | 0% | 0% |
| Jan 26 | 6,119 | 88.5% | 10.8% | 0% | 0% |
| Feb 2 | 19,059 | 88.4% | 10.8% | 0% | 0% |
| Feb 9 | 22,737 | 89% | 10.4% | 0% | 0% |
| Feb 16 | 6,578 | 91.2% | 8.4% | 0% | 0% |
| Feb 23 | 8,068 | 89.2% | 10% | 0.1% | 0% |
| Mar 2 | 10,292 | 88.8% | 10.2% | 0.3% | 0.1% |
| Mar 9 | 5,795 | 90.6% | 8.3% | 0.4% | 0.2% |
| Mar 16 | 6,566 | 83.9% | 13.6% | 1.1% | 0.5% |
| Mar 23 | 508 | 70.3% | 22.4% | 3.9% | 1% |
The last week (508 entities) is a smaller sample.
The Mar 23 week has fewer entities because the data was cut mid-week. But the trend is consistent with the preceding weeks — Google has been declining since mid-March. The 70.3% figure is directionally strong, even if the exact number may shift with a full week of data.
TripAdvisor: Small But Rich
TripAdvisor first appeared on February 20, 2026 with just 66 entities total. But what it lacks in volume, it makes up in content quality. This is a fundamentally different kind of data provider than what ChatGPT has had before.
66 total (0.1%)
1.0% share (up from 0.1%)
4.29 / 5 (501 reviews)
855 chars avg (8.8x Google)
Unlike Google Places, TripAdvisor integration changes what users see in the entity card. Map links point to TripAdvisor pages instead of Google Maps. Reviews are TripAdvisor reviews, not Google reviews. And descriptions are dramatically longer and more editorial.
TripAdvisor URL Format in ChatGPT
tripadvisor.com/Hotel_Review-g{geo}-d{id}-Reviews-{name}-{city}.html?m=69767The ?m=69767 parameter is likely an OpenAI affiliate or tracking ID. Official TripAdvisor logo is hosted on OpenAI's CDN, confirming a formal partnership.
TripAdvisor is destination-agnostic — unlike Yelp.
TripAdvisor appears at similar low rates across all cities (Paris, Rome, Berlin, New York). This is a global rollout, not market-specific. Yelp, by contrast, only appears where it has strong local market presence (US cities + Berlin). As TripAdvisor grows, it could become the first truly global alternative to Google in ChatGPT.
The Link Model Shift: Who Gets the Click?
TripAdvisor introduces a fundamentally different traffic model to ChatGPT. Until now, every provider linked users to the hotel's own website. TripAdvisor is the first to route traffic to an intermediary.
| Provider | website_url points to | Example | Who gets the traffic? |
|---|---|---|---|
| Google Places | Hotel's own website | bestwestern.com/... | Hotel direct |
| Yelp | Hotel's own website | marriott.com/... | Hotel direct |
| Foursquare | Hotel's own website | hyatt.com/... | Hotel direct |
| TripAdvisor | TripAdvisor listing page | tripadvisor.com/Hotel_Review-...?m=69767 | TripAdvisor |
When a user clicks a Google or Yelp entity card in ChatGPT, they land on the hotel's own website — a direct booking opportunity. When they click a TripAdvisor entity, they land on a TripAdvisor listing page where TripAdvisor monetizes the visit through its own metasearch and affiliate links.
The ?m=69767 tracking parameter on every TripAdvisor URL confirms this is a commercial arrangement — TripAdvisor can measure exactly how much traffic OpenAI sends. The utm_source=chatgpt.com parameter makes attribution explicit.
This is a direct booking problem.
For the first time, ChatGPT is sending hotel traffic to an intermediary instead of to the hotel. As TripAdvisor's share grows from 0.1% to potentially much more, an increasing portion of AI-driven hotel discovery will bypass hotel direct websites entirely. Hotels that relied on ChatGPT as a "direct channel" need to reconsider — it depends on which provider powers the entity card.
Why this matters at scale
Today, TripAdvisor is 0.1% of entities (1.0% in the latest week). But the trajectory is clear: every new provider so far has grown from near-zero to meaningful share within weeks. If TripAdvisor reaches 5-10% — which Yelp already has — that's a significant share of ChatGPT hotel traffic flowing to TripAdvisor instead of to hotels. For an industry fighting to increase direct bookings, TripAdvisor's ChatGPT integration moves the needle in the wrong direction.
Description Richness: TripAdvisor's Edge
The strongest finding in the dataset. TripAdvisor descriptions are 8.8x longer than Google's (855 vs 97 characters). And Google only provides descriptions for 1.3% of entities — TripAdvisor provides them for 69.7%. This is a massive content gap that TripAdvisor fills.
| Provider | Entities | Has Description | Avg Length | Median Length |
|---|---|---|---|---|
| TripAdvisor | 66 | 69.7% | 855 chars | 870 chars |
| SerpAPI | 741 | 100% | 145 chars | 142 chars |
| Foursquare | 156 | 44.7% | 136 chars | 140 chars |
| Yelp | 10,148 | 0.9% | 140 chars | 142 chars |
| Google Places | 88,425 | 1.3% | 97 chars | 95 chars |
Google's description gap is ChatGPT's content problem.
Only 1.3% of Google Places entities have any description at all. For the other 98.7%, ChatGPT gets a name, rating, and address — but no narrative context. This is why TripAdvisor matters disproportionately to its 0.1% share: its descriptions give ChatGPT the editorial depth needed for nuanced hotel recommendations.
Non-Google Providers by Destination City
Yelp presence is heavily destination-dependent. In Las Vegas, Yelp accounts for 40.8% of entities. In Paris, it's 0%. TripAdvisor, by contrast, shows a small but consistent presence across all destinations — the chart below excludes Google to make these smaller providers visible.
| City | Entities | Yelp | Foursquare | TripAdvisor | |
|---|---|---|---|---|---|
| Las Vegas | 5,366 | 59% | 40.8% | 0.1% | 0% |
| San Francisco | 4,237 | 64.5% | 34.9% | 0% | 0% |
| New York | 4,358 | 69.6% | 29.9% | 0% | 0% |
| Los Angeles | 1,701 | 70.4% | 29.1% | 0% | 0% |
| Berlin | 5,652 | 73.1% | 25.8% | 0.1% | 0.1% |
| Amsterdam | 6,437 | 93.9% | 5.7% | 0% | 0.1% |
| Saint-Tropez | 5,274 | 95.7% | 3.2% | 0.2% | 0.1% |
| Dubai | 5,514 | 98.3% | 1% | 0.1% | 0.1% |
| Rome | 6,495 | 98.7% | 0% | 0.2% | 0.2% |
| Istanbul | 5,744 | 98.7% | 0% | 0.5% | 0% |
| Paris | 9,396 | 99.6% | 0% | 0.1% | 0.1% |
| Barcelona | 5,848 | 99.1% | 0% | 0.2% | 0.1% |
| London | 4,664 | 99.2% | 0% | 0% | 0% |
European cities are still Google-dominated.
Paris (99.6%), Barcelona (99.1%), and London (99.2%) remain almost entirely Google-powered. Yelp has no presence in these cities. TripAdvisor's small but global footprint makes it the only current alternative provider for European hotel visibility in ChatGPT.
Provider Share by Proxy Country
Unlike destination city, the user's location (proxy country) has minimal effect on provider distribution. TripAdvisor and Foursquare show no geographic bias — they appear at similar rates across all proxy countries.
| Country | Entities | Yelp | Foursquare | TripAdvisor | |
|---|---|---|---|---|---|
| US | 55,627 | 89.9% | 9.4% | 0.1% | 0.1% |
| ES | 17,608 | 87.7% | 11.5% | 0.2% | 0.1% |
| DE | 17,332 | 87.2% | 12% | 0.2% | 0.1% |
| GB | 8,971 | 87.6% | 11.5% | 0.2% | 0.1% |
Provider selection is destination-based, not user-based.
Whether you're searching from the US, UK, Germany, or Spain, you get essentially the same provider mix for the same destination. ChatGPT picks data sources based on where the hotel is, not where you are. This confirms the finding from our Yelp study.
Provider Share by Map Position
Is Google more likely to hold the #1 spot? Slightly — 91.5% vs 88-89% at other positions. But the differences are small. There is no strong position bias by provider.
| Position | Entities | Yelp | Foursquare | TripAdvisor | |
|---|---|---|---|---|---|
| #1 | 11,093 | 91.5% | 8.1% | 0.2% | 0.1% |
| #2 | 11,073 | 88.7% | 10.6% | 0.2% | 0.1% |
| #3 | 11,045 | 88.9% | 10.2% | 0.2% | 0.1% |
| #4 | 10,770 | 88.5% | 10.7% | 0.3% | 0.1% |
| #5 | 10,216 | 88.5% | 10.7% | 0.2% | 0.1% |
Ratings and Reviews by Provider
Each provider brings its own rating scale and review volume. Google shows the highest average rating (4.47) and the most reviews (4,378). Yelp's stricter rating culture produces notably lower averages (3.83). TripAdvisor falls between the two at 4.29.
| Provider | Entities | Avg Rating | Avg Reviews |
|---|---|---|---|
| Google Places | 88,105 | 4.47 | 4,378 |
| Yelp | 10,143 | 3.83 | 662 |
| Foursquare | 124 | 4.17 | 563 |
| TripAdvisor | 57 | 4.29 | 501 |
Rating scale differences matter for AI recommendations.
A hotel that's 4.5 on Google but 3.8 on Yelp may be treated differently depending on which data source ChatGPT uses. As multi-source data grows, ChatGPT will need to normalize across rating scales — or hotels with lower Yelp ratings could be disadvantaged in mixed-source results.
What This Means for Hotels
1Multi-platform optimization is now mandatory
The days of "optimize Google Business Profile and you're done" for AI visibility are ending. ChatGPT is actively pulling from Google, Yelp, TripAdvisor, and Foursquare. Build a checklist: GBP + TripAdvisor + Yelp (if US/Berlin) + Foursquare.
2TripAdvisor descriptions are your strongest lever
TripAdvisor descriptions are 8.8x longer than Google's in ChatGPT. A well-written TripAdvisor description will directly influence how ChatGPT presents your hotel. Google only provides descriptions for 1.3% of entities — TripAdvisor provides them for 69.7%.
3Yelp is critical for US city hotels
Las Vegas (40.8%), San Francisco (34.9%), New York (29.9%) — nearly half of ChatGPT's hotel data in these cities comes from Yelp. If your hotel is in a US city, your Yelp profile is as important as your Google listing for AI visibility.
4European hotels: watch TripAdvisor's global rollout
TripAdvisor is appearing across all destinations, not just US cities. As its share grows, European hotels that have neglected TripAdvisor in favor of Google may find themselves without representation in ChatGPT's expanding multi-source results.
5Track the weekly trend, not the aggregate
Google's 88.8% overall share looks comfortable. But 70.3% in the latest week tells a different story. The shift from 100% to 70.3% happened in 90 days. Quarterly AI visibility audits are already outdated — monthly monitoring is needed.
Frequently Asked Questions
Methodology
Data Collection
We scraped ChatGPT Search across 4 proxy countries (US, ES, DE, GB) using 608 hotel-focused queries across ~25 cities. Each query triggers ChatGPT's map widget, which returns entity cards with structured data via SSE stream.
Provider Detection
Each entity's provider is identified from the SSE stream's entity_data object. Google Places uses ChIJ... IDs, Yelp uses 22-char alphanumeric IDs, TripAdvisor uses numeric IDs with tripadvisor-feed provider, and Foursquare uses 24-char hex IDs with images from fastly.4sqi.net.
Coverage & Limitations
99,538 entities across 25,645 captures from December 25, 2025 to March 24, 2026. Not all prompts return results each run due to rate limiting — daily entity counts vary. TripAdvisor's sample is small (66 entities); findings about its content quality are directionally strong but based on limited data. We cannot distinguish between global rollouts and A/B tests.
Related Research
See our Yelp ChatGPT Hotels Study for the original documentation of the first non-Google provider, and our Anatomy of ChatGPT Hotel Search for the full technical architecture behind the provider ecosystem.
Want the Full Picture?
This study is part of our ongoing tracking of how AI is reshaping hotel visibility. Read our flagship study covering all AI hotel data sources and what they mean for your strategy.
Read AI Hotel Landscape 2026