[Nov 25] OTA Bias in Google Gemini Models

Methodology

This study is based on HotelRank.ai’s proprietary analysis of over 10,000 prompts designed to reflect real-world hotel search scenarios. They are the same prompts as our former analysis on ChatGPT models.

To capture a wide range of user behaviors, we defined a consistent panel of representative traveler personas:

• Couples
• Elderly travellers
• Families with children
• Group Business travelers
• Leisure groups
• Luxury travelers
• Solo Business travelers
• Solo leisure travelers
  

Each persona was used to query AI systems across a variety of destinations and contexts (romantic weekend, family stay with a pool, boutique hotel near a convention center, etc.).

We focused on 3‑star, 4‑star, and 5‑star hotels (even though hotel standards are not always consistent across countries).

Here is the list of cities:

• New York
• Los Angeles
• Miami
• Paris
• London
• Rome
• Barcelona
• Amsterdam
• Berlin
• Dubai
• Bangkok
• Tokyo
• Shanghai
• Hong Kong
• Singapore
• Sydney
• Melbourne
• Cape Town
• Cairo
• Istanbul
• Athens
• Mexico City
• Rio de Janeiro
• Buenos Aires
• Toronto

Model Comparison Protocol

We submitted the same prompts to a selected set of three Google Gemini model variants, including:

• Gemini 2.5 Flash
• Gemini 2.5 Flash Lite
• Gemini 2.5 Flash Preview 09-2025
  

Gemini 2.5 Flash corresponds to the version deployed in the main Gemini interface.

The other models are frequently used through API calls by third-party services, such as AI travel agents.

For each AI-generated response, we extracted and classified:

• All hotel-related links, broken down into:
  • OTA (Online Travel Agencies) — e.g., Booking.com, Expedia, Hotels.com
  • Direct hotel websites
• Model (name)
• Prompt context (persona, city)

Below are our results.

No major differences across models — Google tends to favor OTAs more than ChatGPT

Whereas GPT-5 almost exclusively returns direct hotel links (>95%), as shown in our previous analysis, this is not the case for Gemini.

All Gemini models include a significant share of OTA links — between 37% and 47%.

It would be interesting to test with Gemini Flash Pro, to see whether a more advanced “reasoning” capability would lead to a much lower OTA rate.

The higher the star rating, the fewer OTA links appear in the AI’s responses

Like ChatGPT, Gemini models show a strong bias depending on the hotel category.

• 3-star hotels: heavy reliance on OTAs — from 58% to 76% of the links.
  • In comparison, GPT-5 lists only about 12% OTA links for the same segment.
• 4-star hotels: the effect remains, though less pronounced — 35% to 55% OTA links.
• 5-star hotels: the bias drops — under 10% OTA links
  • partly because some luxury properties are not listed (or less present) on OTAs.
    

In France, as of 2024, 3-star hotels represented 48.6% of all classified hotels, and 4-star hotels made up 19.33% (source: INSEE).

This means the observed bias disproportionately impacts the majority of hotels, especially those in the mid-range — and likely extends to 2-star properties as well.

The type of hotel — and by extension, its budget category — also has a clear impact on the rate of OTA links in AI responses

Similarly, the most affordable properties — hostels and budget hotels — show the highest share of OTA links, often exceeding 65%.

In contrast, luxury hotels, resorts, boutique hotels, and spa hotels are much less frequently cited through OTAs.

This also highlights a broader issue of "algorithmic fairness": the lower-budget properties — often the ones most reliant on disintermediation — are also the most confined to OTA monopolies in AI-generated responses.

Luxury and group-oriented business hotels show a much lower presence of OTA links

Leisure travel shows the highest OTA rates — but so do solo business trips, which is a key difference compared to ChatGPT.

This suggests there’s also a bias based on traveler segments, which could represent a strategic challenge for hotels depending on their target clientele.

Booking — still the undisputed reference point.

Even though Booking.com is arguably the most well-known OTA, the numbers are striking: 80% of all OTA links point to Booking.

However, other OTAs — especially Expedia — perform better than in ChatGPT.

With ChatGPT Apps on one side and Google’s ad ecosystem on the other,

the question is: what comes next for distribution visibility in AI?

Conclusions

• Gemini consistently includes OTA links — between 37% and 47%.
• The higher the hotel category, the fewer OTA links appear.
• Similarly, the higher the traveler’s budget, the fewer OTAs are cited
  • although this pattern is less pronounced than in ChatGPT.
    

Beyond the Findings

How can we explain these biases?

Here are a few hypotheses:

• Google Hotels, Google Ads, and Google Maps all play a role in what Gemini surfaces.
  • In short: if you pay or have strong visibility in Google’s ecosystem, you appear more.
  • This helps explain why Gemini features more OTA links than ChatGPT — OTAs invest heavily in SEA, distribution, and structured data.
• It would be worth testing whether a more advanced reasoning model (e.g., Gemini Flash Pro or a more constrained prompt) would, like GPT-5, reduce OTA prevalence by better understanding travel context.
• The same pattern holds across segments
  • Hotels with weaker digital presence or brand image — often in lower categories — are more easily replaced by OTAs in results.
  • Meanwhile, premium and well-branded hotels perform significantly better, even within Gemini.