{"@context":"https://schema.org","@type":"BlogPosting","headline":"AI Hotel Direct Links Study 2026: Do Different Models Link Differently?","description":"API-level comparison of direct links across ChatGPT (5.1–5.4) and Gemini with web search enabled.","datePublished":"2026-03-10","dateModified":"2026-03-10","url":"https://nicolassitter.com/research/ai-hotel-direct-links-study-2026","category":"research","keywords":["AI direct links","ChatGPT vs Gemini","model linking behavior","hotel AI visibility"],"articleSection":"Research","wordCount":4700,"readTime":"19 min","articleBody":"Nicolas Sitter ResearchMarch 2026\n\n# Do AI Models Link Differently?\n\nAPI-level comparison of direct links in hotel recommendations across ChatGPT (GPT-5.1 → 5.4) and Gemini models. With web search enabled, where do the links actually point?\n\n6\n\nModels Tested\n\n—\n\nQueries\n\n2\n\nProviders (OpenAI + Google)\n\n[Summary](#executive-summary)[Hypothesis](#hypothesis)[Models](#models-tested)[Results](#link-distribution)[ChatGPT vs Gemini](#chatgpt-vs-gemini)[FAQ](#faq)[Methodology](#methodology)\n\n## TL;DR\n\nWe called ChatGPT (GPT-5.1, 5.2, 5.3, 5.4) and Gemini (2.0 Flash, 2.5 Pro) via their APIs with web search enabled, using identical hotel queries. **The hypothesis: direct links should be model-agnostic** — because links come from the search tool, not the language model. The model decides what to say; the search infrastructure decides where to link.\n\n## Executive Summary\n\nWhen ChatGPT recommends \"Hotel Le Marais\" and includes a link, that link comes from the web search tool (SerpAPI, Google Places) — not from the language model's training data. This means switching from GPT-5.1 to GPT-5.4 should not change _where_ links point, only _which hotels_ get mentioned and _how_ they're described.\n\nThe same logic applies across providers: ChatGPT and Gemini use different search backends, so their link distributions may differ — but within each provider, model version shouldn't matter.\n\n**Why this matters for hotels:** If links are model-agnostic, then optimizing for AI visibility is about being present in the right search sources (Google, TripAdvisor, Booking) — not about gaming specific model versions.\n\n## Hypothesis\n\n### H1: Within-provider consistency\n\nDifferent ChatGPT model versions (5.1 → 5.4) should produce the same link distribution when web search is enabled. The model generates text; the search tool provides links.\n\n### H2: Cross-provider differences\n\nChatGPT vs Gemini may show different link distributions because they use different search backends (SerpAPI vs native Google Search). The search infrastructure matters more than the model.\n\n### The Architecture Argument\n\nUser Query → \\[Language Model\\] → decides to search\n\n→ \\[Search Tool\\] → fetches results + links\n\n→ \\[Language Model\\] → writes answer using those links\n\nThe language model controls which hotels to mention and how to describe them. But the URLs come from the search tool. Changing the model changes the text — not the links.\n\n## Models Tested\n\n#### Models included in the study\n\nModel\n\nProvider\n\nAPI Access\n\nWeb Search\n\nGPT-5.1\n\nOpenAI\n\nGPT-5.2\n\nOpenAI\n\nGPT-5.3\n\nOpenAI\n\nGPT-5.4\n\nOpenAI\n\nGemini 2.0 Flash\n\nGoogle\n\nGemini 2.5 Pro\n\nGoogle\n\nAll models were tested via their official APIs with web search explicitly enabled. For ChatGPT, this means using the `web_search` tool in the Responses API. For Gemini, this means enabling Google Search grounding. This ensures all responses use fresh data, not training-data knowledge.\n\n## Link Distribution by Model\n\nWhere do the links in hotel recommendations actually point?\n\n#### Link destination distribution by model (% of all links)\n\n#### Link destination breakdown by model\n\nModel\n\nHotel Direct Website\n\nOTA (Booking, Expedia, etc.)\n\nReview Site (TripAdvisor, Yelp)\n\nEditorial / Media\n\nGoogle Maps / GBP\n\nOther\n\nGPT-5.1\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\nGPT-5.2\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\nGPT-5.3\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\nGPT-5.4\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\nGemini 2.0 Flash\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\nGemini 2.5 Pro\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n0%\n\n**Data pending.** Once the API tests are complete, this section will show whether link destinations are consistent across model versions — confirming that optimizing for search sources matters more than targeting specific models.\n\n## ChatGPT vs Gemini\n\nDifferent search backends, different link patterns?\n\n### ChatGPT Search Stack\n\n-   • SerpAPI → Google Search results\n-   • Google Places API → entity data\n-   • Bing Images → secondary images\n-   • Yelp → US cities + Berlin\n-   • OpenStreetMap → map tiles\n\n### Gemini Search Stack\n\n-   • Native Google Search integration\n-   • Google Maps / Places (direct access)\n-   • YouTube (Gemini-native)\n-   • Google Hotels / Travel\n-   • No intermediary (SerpAPI) needed\n\n**The key difference:** ChatGPT accesses Google through intermediaries (SerpAPI), while Gemini has native access. This may give Gemini more Google-specific link types (Google Hotels, Maps direct links) while ChatGPT may surface more third-party editorial links from web search results.\n\n## Frequently Asked Questions\n\n## Methodology\n\n### Setup\n\n-   • All queries run via official APIs\n-   • Web search enabled on every request\n-   • Identical prompts across all models\n-   • Same set of hotel queries per model\n-   • Links extracted and categorized from responses\n\n### Why Web Search Matters\n\n-   • Ensures fresh, real-time data\n-   • Links come from search results, not training data\n-   • Without web search, models may hallucinate URLs\n-   • Isolates link behavior to search infrastructure\n-   • Reflects real-world ChatGPT/Gemini usage\n\n### Link Classification\n\nEach link in a hotel recommendation is classified into one of these categories:\n\nHotel Direct WebsiteOTA (Booking, Expedia, etc.)Review Site (TripAdvisor, Yelp)Editorial / MediaGoogle Maps / GBPOther\n\n## Continue Reading\n\nExplore more Nicolas Sitter research on AI hotel search.\n\n[AI Hotel Landscape 2026](/research/ai-hotel-landscape-2026)\n\n[Anatomy of ChatGPT Search](/research/anatomy-chatgpt-hotel-search-2026)[GPT Model Changes Study](/research/gpt-5-1-to-5-2-model-changes-study-2026)[All Research](/research)","author":{"@type":"Person","name":"Nicolas Sitter","url":"https://nicolassitter.com/about","sameAs":["https://www.linkedin.com/in/nicolassitternolleau/","https://github.com/Nicositter88","https://hotelrank.ai"]},"publisher":{"@type":"Person","name":"Nicolas Sitter","url":"https://nicolassitter.com"},"image":"https://nicolassitter.com/api/og/ai-hotel-direct-links-study-2026","mainEntityOfPage":{"@type":"WebPage","@id":"https://nicolassitter.com/research/ai-hotel-direct-links-study-2026"},"tags":["ChatGPT","Gemini","Direct Links","Model Comparison"],"sameAs":["https://hotelrank.ai/research/ai-hotel-direct-links-study-2026"],"alternateFormat":{"html":"https://nicolassitter.com/research/ai-hotel-direct-links-study-2026","json":"https://nicolassitter.com/api/post/ai-hotel-direct-links-study-2026","rss":"https://nicolassitter.com/rss.xml"},"datasets":[{"name":"summary","contentUrl":"https://nicolassitter.com/data/ai-hotel-direct-links-study-2026/summary.csv","encodingFormat":"text/csv"}]}