Editorial standards

Editorial policy

Last reviewed: 7 August 2026 · 8 sections

TravelShifu publishes desk assessments and decision guides. Most hotel and attraction pages are not first-person trip reports. When we have not stayed, flown or visited, the page says so in the reader's language — usually as a labelled “Desk assessment.” This policy explains exactly how a page earns its verdict, what can change that verdict, and the things we refuse to publish even when refusing costs us money.

1. Who decides what we publish

Named editors own page verdicts. The byline on a guide or hotel page is the accountable person or the editorial team label when the work is collaborative desk research. Commercial partners do not write, rank or approve editorial conclusions, and no partner sees a page before publication. See Kaiser Khan for the editor of record and About for how the site is funded.

The practical test is simple: if a paragraph would change because a partner asked for the change, the paragraph does not ship. Verdicts change because facts change — a closure, a policy change, a partner going out of business — and for no other reason.

2. What “desk assessment” means

A desk assessment is a verdict built from named, dated sources we can show: supplier guest analyses and amenity facts for hotels; operator pages and published face prices for attractions; official rules for visas, compensation and airline policies. It is not a stay we invent, a queue time we guess, or a star rating we quietly re-issue as our own.

  • We cite supplier scores as attributed, dated text with the scale and the count — never as aggregateRating structured data.
  • We do not invent hotel rows, image URLs, kilometre claims, opening hours or nightly rates.
  • When evidence is missing, the page says what we do not know instead of filling the gap.
  • Every specific number carries its primary source and the date we checked it. A number we cannot source today does not publish.

3. How a verdict is reached, step by step

  1. State the decision. Every page names the choice the reader is actually making — book this hotel, buy this pass, take the train or fly — before it says anything else.
  2. Gather dated facts. Primary or supplier-attributed facts with a capture date. Nothing enters a page from memory, from a competitor's phrasing, or from a language model's confidence.
  3. Reach the verdict. The page answers in its first 60 words, then argues. A page that hedges gives the reader nothing to decide on, and it is sent back.
  4. Name who it is not for. Every verdict carries a “skip it if”. A recommendation with no downside is marketing, and our build contract treats a missing one as a defect.
  5. Price the decision honestly. Where the free or official option wins — the official ticket site, the walk, the free viewpoint — the page says so, even where that costs us a commission.
  6. Label the money. Every monetised link is labelled before the click, and inactive partners are identified as non-commissioned.

Method notes by content type live at /method.

4. How suppliers can — and cannot — influence coverage

Suppliers can influence what exists, never what we say about it. A booking partner's inventory decides which hotels, tours or fares are available to link to; our hotel data provider's records are the factual layer our hotel pages join against. That is the limit of it.

  • A supplier may correct a fact — an address, an amenity, a closure — with evidence, and we re-check it against our own sources.
  • A supplier may not see a page before publication, approve a verdict, buy a ranking, or ask for a “not for” section to be softened.
  • We decline partnerships that require guaranteed placement or praise, and we say when a partner is weak at something. That is what makes the rest credible.
  • Hotels and attractions cannot pay to appear. The curated lists are short on purpose; a paid listing would end the only thing that makes them worth reading.

5. What we will not publish

  • Invented first-person experiences of stays, flights, lounges or attractions we did not take.
  • Award prices, APRs, live hotel rates or queue times we cannot source today.
  • Years in evergreen slugs, supplier marketing copy re-issued as our review, or scraped user reviews re-written in our voice.
  • Scaled template pages — thousands of near-identical combinations generated to catch search traffic. Every page must justify its own existence.
  • Credit-card recommendations until affiliate approval and regulatory compliance are genuinely in place. Existing card guides are educational and earn nothing from card issuers.

6. Corrections — what triggers one, and who decides

A correction is triggered by a factual error that could change a decision: a wrong price, a closed venue described as open, a policy described as current when it has changed, a partner named as live when it has shut down. Typos that change nothing are simply fixed.

The editor of record decides what is material. Material errors are fixed promptly, the review date is updated because the content changed, and the change is recorded on /corrections with a date and a plain description of what we got wrong. Page-level notes appear near the byline when the history matters on that URL. Anyone can report an error via contact.

7. Dates and re-verification

Every page carries a published date and a last-updated date, and the second changes only when the content does. A date that moves without a content change is a lie told to crawlers, and our own rules forbid it. Pages are re-verified on a rolling cycle — oldest first — checking the facts that decay: partner names, closures, policy changes, whether a linked page still exists. Where nothing has changed, nothing is touched.

8. Commercial independence

Affiliate links may earn commission; the price to the reader is identical to booking direct. They do not buy a positive verdict, a ranking, or a place on a curated list. Full legal wording: affiliate disclosure.

9. AI assistance vs. human responsibility policy

We use AI systems to assist with data indexing, formatting, passage retrieval, and code validation. However, no AI system generates unverified facts or makes editorial decisions on TravelShifu:

  • Zero hallucination policy: Every price, policy, room feature, or queue guideline must originate from a verified, primary human-checked record.
  • Human editor accountability: A named human editor reviews and takes sole accountability for every published verdict, advice piece, and recommendation.
  • Machine legibility: We publish machine-readable summaries (llms.txt and Markdown mirrors) so AI search engines can cite our human-verified travel research accurately.

10. Stated provenance standards for ratings

Rating manipulation is rife across online travel publishing. To protect our readers and respect structured data guidelines, TravelShifu enforces strict provenance rules:

  • We never emit third-party supplier review scores (e.g. LiteAPI, Booking.com, Agoda ratings) as structured JSON-LD AggregateRating or Review schema.
  • All supplier scores appear as visible, attributed text with explicit capture dates, original scale (/10 or /5), and review sample counts.
  • First-party stay reviews (marked with stated provenance) are reserved exclusively for properties personally audited by our team.

11. Non-fabrication guarantee & data integrity

In travel publishing, an incorrect or invented detail—such as an outdated visa fee, wrong train station, or inaccurate terminal connection—can derail a reader's journey. We adhere to an absolute non-fabrication guarantee:

  • We do not guess dynamic pricing, peak room rates, or exact queue durations.
  • When data is missing or unverified, we explicitly state what is unknown rather than outputting a plausible placeholder.
  • All numerical claims (luggage allowances, lounge access fees, connection times) carry their primary source and capture timestamp.

12. Reader feedback & dispute resolution

Our readers are our primary partners. If you discover an error, find a closed venue listed as open, or disagree with an editorial verdict, we invite your feedback. Material corrections are logged publicly on our Corrections log.

TravelShifu is owned and operated by Elite Digital AI Solutions Ltd, registered in England & Wales, Company No. 17250017. Related: about · method · corrections · contact