In my August 2026 test of ChatGPT hotel recommendations across eight Vietnamese destinations, 83% of 555 citation links pointed to hotel-owned or editorial pages. The remaining reported share, 17%, went to the study’s OTA and aggregator group.
That makes a hotel’s own website and its wider editorial presence worth examining together. It does not mean 83% of travellers visited those pages or booked direct: this was a citation test, not a traffic or conversion study.
I ran the test to understand which hotels appeared, which pages received links, and which sources were recorded alongside the recommendations. The findings gave me a more specific starting point for hospitality SEO/GEO work.
| Finding | Result in this test | What was counted |
|---|---|---|
| Hotel-owned and editorial destinations | 83% | 555 citation links |
| Independent and boutique hotels | 427 mentions, approximately 56% | 756 hotel mentions |
| Corporate pages for chain hotels | 89% | 240 mapped hotel-owned chain links |
| Forbes Travel Guide versus Booking.com | 23 versus 19 source records | The report’s separate grounding-source analysis (n = 427) |
How I ran the test
The study used six English-language traveller prompts for each of eight destinations, with two runs per prompt: 48 prompts and 96 query runs in total. The report records ChatGPT / GPT-5.4 with live web search enabled.
The destinations were Hà Nội, Ho Chi Minh City, Đà Nẵng, Hội An, Nha Trang, Phú Quốc, Đà Lạt and Sa Pa. Recommended hotels, citation links and grounding sources were parsed and classified by chain or independent status, source type, and corporate page or property microsite.
A hotel mention is an appearance in an answer, not necessarily a unique property. A citation link is a URL attached to the response. The report also records a separate grounding-source set. Those totals should not be combined, or interpreted as the same denominator.
1. Hotel-owned and editorial pages received most citation links
The largest group was hotel-owned and editorial pages, accounting for 83% of the 555 citation links. That group combines chain corporate pages, property microsites, independent hotel websites and premium editorial pages. It is not an 83% share for hotel websites alone.
The report places Booking.com, Agoda, Expedia, Trip.com and Tripadvisor in its OTA/aggregator link group. Tripadvisor is labelled as a review source in the separate grounding-source analysis. I am retaining that distinction here rather than treating the two classifications as interchangeable.
For a hotel team, I would use this finding to inspect the exact pages cited for relevant guest questions. Is the answer supported by the property website, a corporate profile, a publication or a booking platform? Each points to a different content or information gap.
This test did not compare equivalent Google results. It cannot establish that AI search has reversed Google’s channel mix or reduced OTA bookings.
2. Independent hotels were the majority, with large destination differences
Independent and boutique hotels accounted for 427 of 756 mentions, approximately 56%. Among the named chains shown in the report, Accor recorded 77 mentions, Marriott 74 and IHG 57. These are appearances across the test, not counts of distinct hotels.
The overall figure also hides a substantial difference between destinations. Independents accounted for 92% of mentions in both Đà Lạt and Sa Pa, compared with 37% in Phú Quốc.
| Destination | Independent share | Chain share |
|---|---|---|
| Đà Lạt | 92% | 8% |
| Sa Pa | 92% | 8% |
| Hội An | 65% | 35% |
| Ho Chi Minh City | 53% | 47% |
| Đà Nẵng | 44% | 56% |
| Hà Nội | 40% | 60% |
| Nha Trang | 38% | 62% |
| Phú Quốc | 37% | 63% |
I would not use these figures to claim that independent hotels are inherently favoured. The mix of hotels available and the questions selected may both influence the result. What the test shows is that independents appeared frequently, and that a national average is a poor substitute for checking a property’s actual market.
3. Editorial sources appeared alongside corporate and booking platforms
In the separate grounding-source analysis, Forbes Travel Guide recorded 23 appearances, compared with 19 for Booking.com. Marriott’s corporate domain recorded 39 and Tripadvisor 36.
| Source | Records shown |
|---|---|
| marriott.com | 39 |
| tripadvisor.com | 36 |
| Forbes Travel Guide | 23 |
| ihg.com | 23 |
| booking.com | 19 |
| all.accor.com | 18 |
| Condé Nast and Michelin, combined in the report | 17 |
| hyatt.com | 14 |
The useful observation is that editorial coverage was present in the source mix. The test does not prove that a Forbes rating caused a recommendation, that ChatGPT “trusts” it more than Booking.com, or that securing coverage would outperform every other investment.
My next step would be to examine which publications repeatedly appear for the guest needs a property can serve, and whether there is a relevant, evidence-backed story to contribute. Accurate booking-platform profiles still deserve attention: they also appeared in the recorded sources.
4. Corporate pages received most mapped chain-hotel links
Across 240 mapped hotel-owned chain links, 89% went to corporate pages and 11% to property microsites. The report records 26 microsite links overall.
The split varied by brand. Microsites accounted for 16% of Accor’s mapped links and 33% of IHG’s; Marriott and Hyatt recorded no microsite citations in this sample. InterContinental Danang contributed six microsite citations.
There is an important counting effect: a corporate domain can represent an entire portfolio, while a property domain may represent just one hotel. The report notes that corporate pages still led among properties with both options, but by a smaller margin.
I would start by reviewing the property’s corporate page and the information it provides. A zero in this test is not enough reason to conclude that every Marriott or Hyatt microsite is invisible or should be removed. A separate site needs a clear guest and business purpose, and its value needs to be assessed at property level.
What I would investigate for a hospitality business
Start with the guest question and the page that answers it. Room configurations, named amenities, location, policies and direct-booking inclusions should be specific and easy to find. My guide to hotel website content for AI search explains how I approach that information.
Check the wider information about the property. Compare the website with corporate profiles, booking platforms, review sites and relevant publications. A citation audit should identify missing or conflicting facts as well as absent links.
Evaluate editorial opportunities by relevance. Source appearances can help identify publications to investigate. They do not guarantee coverage, citations or bookings.
Repeat a defined set of guest questions. Record the prompt, model, search setting, date, recommended properties and cited URLs. Look for patterns across runs, then connect any visibility changes with website visits and confirmed bookings where measurement allows.
This follows the GEO Operating System behind my work: establish what the business should be chosen for, improve the evidence available about it, and investigate how discovery connects with the commercial result.
What this test can and cannot tell us
This is a directional snapshot from one reported model, English-language prompts, two runs per prompt and eight Vietnamese destinations in August 2026. It does not represent every traveller question, language, location or AI platform. Two runs also do not establish how consistently a result will persist.
The results describe the recorded recommendations and links. They do not isolate what caused a hotel to appear, prove that the model read every listed page in full, or measure clicks, enquiries, bookings or revenue. The microsite sample is small, and portfolio-level domain counts favour large corporate sites.
For me, the finding worth acting on is that property information and relevant third-party coverage both belong in a hospitality citation audit. The next decision should come from the specific gaps found for a property and its prospective guests.
For context on the interface, OpenAI’s ChatGPT Search documentation explains that search responses may include citations and that the Sources panel can include both cited sources and other relevant links.