The hotels AIrecommends mostin Taipei

We send the same 24 traveler questions to each AI on a fixed schedule: a concert weekend at Taipei Dome, business meetings near Taipei 101, a family weekend with a pool, a romantic weekend, a two-week stay with a kitchen and laundry, and a five-star budget under NT$9,000 a night. Twelve are written as an American visitor flying in, in English. Twelve are written as a Taiwanese traveler coming up from Taichung, in Chinese. Each question starts a fresh chat and asks the AI to ignore saved memories and search the web as it normally would.

We clean the names first. The same property under two names counts once. A name that is not one real property, such as a chain, a platform or a hotel that does not exist, is left out, while its answer stays in the count. A hotel that has closed or has not opened stays, marked as such.

A hotel counts once per answer when the AI recommends it. Hotels it mentions only to rule out are left out, and so are answers where the AI only asked a follow-up question. Share of answers is the share of answers in your selection that recommend the hotel. "#1" counts the answers where AI listed the hotel first, and average position is the hotel's average place in the AI's list when it was recommended.

The period buttons next to Overview set which runs the page counts, and Change and Movers compare that period with the one before it. All time counts every run so far, and its change compares the more recent half of the runs with the earlier half. 7, 14 and 30 days count the runs in that window and compare it with the same number of days before it. Both sides count the same mix of AIs and languages, each weighing its smaller number of answers in the two periods, so an AI that joined partway does not skew the result. The result is in percentage points. "New" marks a hotel recommended in the recent period and not in the one it is compared with.

Muse answers half as many questions as the other platforms and switches between English and Chinese on alternate days. Grok is paused. Names are matched across languages, so 台北君悅酒店 and Grand Hyatt Taipei count as one hotel. Source links are recorded when the AI shows them; Gemini has not so far, and no links were recorded for the Oct 23 run or for ChatGPT on Oct 29. Booking sites are online travel agencies and corporate travel tools; hotel and brand sites are a hotel's own site or its chain's. A recommendation shows what the AI said; it is not a rating of the hotel.

Runs so far: Oct 23, 2026, Oct 24, 2026, Oct 25, 2026, Oct 26, 2026, Oct 27, 2026, Oct 28, 2026, Oct 29, 2026, Nov 3, 2026, Nov 5, 2026. We keep collecting on a fixed schedule, so rankings settle as more answers come in.

Every hotel in every view, by AI and by traveler type: the data as JSON.