How it works

How we investigate your hotel

Most audit tools do not show their work. This page does.

A Mystery Shop is a five-stage investigation of your website and your AI presence. Every finding arrives with evidence: a screenshot, a line of code, or a recorded answer from an AI assistant. The same scan also surfaces the demand you are missing, handed back as a brief on what to win. See Win the demand you are missing.

43

checks live in your report

69

in calibration until we trust them

The pipeline

Five stages. Every finding has to earn its place.

From the first page we capture to the report in your inbox.

01

Crawl

Up to 30 mobile pages, screenshots and HTML captured

02

Deterministic

Rules that return the same answer every time

03

AI vision

Each screenshot read the way a guest reads it

04

Cross-property

Your hotels compared to each other

05

AI investigator

An agent that follows hunches and files evidence

Stages 01 + 02 / The groundwork

Capture first. Then the rules we trust completely.

We crawl up to 30 pages per property on a mobile viewport, capturing screenshots and raw HTML. We simulate the public digital journey: we never log in, never make bookings, never touch guest or booking data. The crawler identifies itself, follows standard rate limits, and has no effect on your rankings.

Then the deterministic pass: checks that return the same answer every time. A missing sitemap is missing. An absent security header is absent. No judgement calls. This is the layer search engines and AI assistants actually read when deciding whether to surface your hotel.

AI crawler accessStructured dataSitemapsCore Web VitalsSecurity headersBroken links and imagesSocial previews+ more

Stage 03 / AI vision

What the camera sees, the model reads.

Example. A Christmas promotion from 2024 still live on a homepage in May. No rule catches that. The vision pass does, because it reads the screenshot the way a guest would: prices, dates, booking buttons, broken layouts, placeholder text.

A visual finding never reaches you on its own. It has to be corroborated by the page code, and it has to survive a set of bright-line rules that dismiss anything the site itself contradicts.

The HarboursideAmsterdam
Example
Critical

Christmas Getaway 2024

From €189 per night. Book by 24 December 2024.

Captured in May · Expired offer still selling. Filed with the screenshot as evidence.

Stage 04 / Cross-property

Defects only a portfolio scan can find.

Example. On a recent scan of a European hotel group, the Milan property’s page carried the Lisbon hotel’s full identity block in its code, verbatim. Both pages looked perfect to guests. To search engines and AI assistants, Milan was Lisbon.

What guests saw

Two beautiful, correct hotel pages. Right photos, right address, right phone number.

What machines read

"name": "Hotel Lisbon"
"addressLocality": "Lisbon"

On the Milan page. One hotel, as far as AI is concerned.

This stage runs when you have more than one property. It catches identity drift, copy-pasted content, missing local-language pages, and coordinates that put your hotel in the wrong country. A group’s biggest invisible risk is its properties blurring into each other.

Stage 05 / AI investigator

The newest part, and the hardest to do well.

After the audit, an AI agent investigates what the rest of the pipeline cannot resolve on its own. It reads your site the way a consultant would: it follows hunches, fetches evidence, and only files a finding it can pin to a specific piece of proof. It is how multiple symptoms collapse into one upstream cause, and how a small bug inside a critical booking flow gets promoted to the top of your report.

It runs on hard time and spend caps, and its findings face the same evidence bar as everything else. Today it informs our team’s review of your report; its findings will appear directly in your dashboard once calibration proves their accuracy.

The other audience

We do not stop at your website. Every scan also asks the AI assistants guests actually use how they describe and recommend your hotel, and records the answers as evidence.

Claude
ChatGPT
Gemini
Perplexity

The trust ledger

We hold back more checks than we ship.

A check goes live the day we trust it, not the day we build it. Right now that means:

Live in your report

43

Accuracy proven on real hotel websites

In calibration

69

Running silently until they earn their place

Our team keeps an eye on every scan and fine-tunes the system behind it. When a finding does not hold up, we dismiss it and tighten the check that produced it. That feedback loop is why the live column stays trustworthy as it grows.

Drawn lines

What we never do.

  • ·We never access guest or booking data. No booking engine, no PMS, no payment systems, no logins.
  • ·We never affect your rankings. The crawler identifies itself and follows standard rate limits.
  • ·We never make real bookings, and we never submit forms with invented guest details.
  • ·We never ship a check until we trust its accuracy.

See what a guest sees before they book

We use essential cookies to keep the site working. No tracking or advertising cookies are used. See our Cookie Policy for details.