Methodology

How Kinora Signal is made

Kinora Signal is a two-page report on how AI assistants describe one hotel. This is how every report is prepared.

  1. The assistants

    We ask the assistants travelers use most: ChatGPT, Google Gemini, Google AI Overviews, Perplexity and Microsoft Copilot.

    Each question is asked in a fresh session, with no saved history or personal settings.

  2. The questions

    Twelve questions per hotel. Four are the same for all hotels, so results can be compared over time: whether the hotel is recognized, how it is described, whether it can be booked direct, and what booking direct includes.

    Eight are chosen for the property, from what travelers ask about a hotel like it: the occasion, the setting, dining, rooms, and practical details such as pets, parking and check-in.

    Questions are asked the way travelers ask them, without the hotel’s name unless the question needs it.

  3. The evidence

    Every answer is saved with its date, the assistant and its exact wording. We compare it with the hotel’s own published facts: its website, and its Kinora page if it has one.

    Every report carries that record, so any finding in it can be traced back to what was asked, what was answered, and where the answer came from. For each hotel we record:

    • Which AI assistants were tested
    • The exact questions asked
    • The date of each test
    • The answers and source links found
    • Which findings were reviewed by a person

    The report shows a sample of the questions; the full record is kept, and we send it to any hotel that asks.

  4. The findings

    A finding is marked only when the saved answers support it. Findings are never added up into a score.

    Verified
    Correct wherever it appears.
    Incomplete
    Present, but partial or wrong.
    Not found
    No reliable answer anywhere.
    Opportunity
    A clear gain within reach.

  5. What is excluded

    Sponsored placements are excluded from every test, and so are answers shaped by a signed-in history.

    The report records what an assistant recommends on its own.

  6. Review and delivery

    A person reviews every finding before the report is sent, within 48 hours of the request.

We do not guarantee AI rankings, recommendations, or bookings.

Back to the homepage