Method
How we review hotels
Last reviewed: 2026-08-07
TravelShifu hotel method: desk assessments from supplier facts, POI distances, guest analyses and dated scores — never invented stays or aggregateRating schema.
What we use
Hotel pages join editorial prose to ingested property rows: name, address, star class, amenity list, check-in/out times, supplier rating with count and capture date, sentiment category scores with summaries, pro/con highlights, and POI kilometre rows when present.
Images come only from the supplier CDN URLs already on the row. We never invent a photo URL, a liteAPI id, a nightly rate or a kilometre claim.
What “desk assessment” means on a hotel page
Unless a page carries a first-party stay review with a named reviewer and stay date, the verdict is a desk assessment. The byline says we have not stayed. Supplier scores appear as attributed, dated text only — never as schema.org aggregateRating.
How a city shortlist is built
Each seeded city publishes only when a CityEditorial and HotelEditorial exist for every property on the page. Matchups are hand-curated either/or pairs, not combinatorial generators. CTAs go to live partners (Booking, Agoda, Trip.com, Vrbo and related hotel partners) and are labelled before the click.
Related: editorial policy · all methods · corrections