Methodology 1.1 · 26 October 2026
How the Mirror Score works
The MAD MIRROR Awards measure 55 criteria in seven disciplines and turn them into a Mirror Score from 0 to 100 by fixed rules. From it follow one to three Mirrors and the test winners of each edition. Every rule is here.
Weights
Seven disciplines, 100 percent
Per discipline, the share of points earned counts. The Mirror Score is the weighted mean, rounded to whole points. A discipline that couldn't be measured drops out; the score is then provisional.
- Google (SEO)18%
- AI visibility (GEO)18%
- Mobile speed22%
- Booking and contact14%
- Tech and trust10%
- Languages9%
- Video and social9%
Criteria
All 55 criteria and their points
Google (SEO)
May Google index the page? Plus title, description, main heading, sitemap, canonical, alt texts and real 404 pages.
18% · 25 points
Google may index the homepage
With noindex or a block, the page doesn't show up on Google at all, however good it is.
4 pts
Page title with 10 to 60 characters
The title is the blue line in Google and the first thing AI assistants read about the page.
1 point if there is a title but it's under 10 or over 60 characters.
3 pts
Meta description with 50 to 160 characters
Without one, Google cuts some random text out of the page.
1 point if there is a description but it's under 50 or over 160 characters.
3 pts
Exactly one main heading (h1)
The h1 is the page's core message for Google and AI; several or none water it down.
1 point with more than one h1.
2 pts
Canonical URL pointing to this site
It tells Google which address counts when the same page is reachable under several.
2 pts
XML sitemap present
The sitemap lists every page for search engines; without it Google finds new pages later.
2 pts
Mobile viewport declared
Google ranks the mobile version of a page first.
2 pts
Served over HTTPS
Browsers warn against the site otherwise, and Google ranks it lower.
2 pts
Page language declared (html lang)
Search engines, translators and screen readers then know which language the page speaks.
1 pts
Open Graph tags for link previews
Guests share links in chats; without a preview a link looks like spam.
1 pts
Images with alt text
Google and AI understand images through their alt text, and screen readers read it aloud.
1 point if at most one image in ten lacks alt text.
2 pts
Unknown addresses answer with 404
Otherwise Google takes empty pages for content, and old links land on the homepage without a word.
1 pts
AI visibility (GEO)
May ChatGPT, Perplexity and Gemini read the site, do they find text without JavaScript, and do they learn what the business is, where and when it's open?
18% · 20 points
AI crawlers allowed in robots.txt
What ChatGPT and the others may not read, they can't recommend.
4 pts
robots.txt present
It's the first file every crawler asks for.
1 pts
Readable without JavaScript: at least 300 words in the HTML
The crawlers of ChatGPT, Claude and Perplexity don't run JavaScript: whatever only appears afterwards, they don't see.
1 point from 120 words.
3 pts
Valid structured data (JSON-LD)
Structured data is the language machines read facts in.
2 pts
Structured data names the business type
These exact fields feed Google Maps, AI Overviews and travel assistants.
4 pts
Address and phone in the structured data
The same address and number everywhere make the business unmistakable for Google and AI.
1 point for address or phone alone.
2 pts
Opening hours or check-in in the structured data
Guests ask Google and AI assistants all the time whether a place is open today.
1 pts
Profiles linked in the structured data (sameAs)
That's how search engines and AI see that website, profiles and reviews belong to the same business.
1 pts
Questions and answers marked up (FAQPage)
AI assistants like to answer with lines they find as question and answer on the page.
1 pts
llms.txt present
A short version for AI assistants that prevents misunderstandings about your business.
1 pts
Mobile speed
Lighthouse in mobile emulation: largest element load time, layout shifts, responsiveness, server, page weight, image formats, lazy loading.
22% · 25 points
Lighthouse performance on mobile
Every second of load time costs bookings, above all on mobile data at the beach.
Proportional: the Lighthouse score divided by 10, so 5.7 points at 57.
10 pts
Largest image visible within 2.5 seconds (LCP)
The largest element decides when the page feels ready to guests.
2 points up to 4 seconds.
4 pts
Layout stays still while loading (CLS up to 0.1)
When the page jumps, guests tap the wrong thing, often right when booking.
1 point up to 0.25.
2 pts
Reacts quickly (blocking time up to 200 ms)
While scripts are busy, the page doesn't react to taps.
1 point up to 600 ms.
2 pts
Server answers within 0.6 seconds
Until the server answers, the guest sees only a white screen.
1 point up to 1.8 seconds.
2 pts
Homepage under 3 MB
Every megabyte costs time and data on mobile networks.
1 point up to 6 MB.
2 pts
Images in WebP or AVIF
Modern formats are much smaller than JPEG and PNG at the same quality.
Full points from half of the images, 1 point if any image is modern.
2 pts
Lazy loading for images further down
The browser then loads first what guests see right away.
Pages with three images or fewer always meet it.
1 pts
Booking and contact
Reservation system, call to book, WhatsApp, tap-to-call phone, email or form, Google Maps: the path from looking to booking.
14% · 14 points
Reservation or booking system built in
Every detour between interest and table costs guests.
4 pts
Clear call to book or reserve
Guests tap what they see; a hidden path is no path.
2 pts
WhatsApp contact
In Ibiza and across Spain people book over WhatsApp, especially at short notice.
3 pts
Phone number as a tap-to-call link
On a phone, a number you have to type is a hurdle.
2 pts
Email link or contact form
Groups, events and enquiries from abroad prefer to write.
1 pts
Link to Google Maps
Anyone who wants to come needs directions, and the link strengthens your Maps listing.
2 pts
Tech and trust
Legal notice, privacy policy, working links, a secure connection, current content, accessibility and reviews: whether guests can trust the site.
10% · 16 points
http:// redirects to https://
Anyone who types the address without https or follows an old link otherwise lands on an insecure page.
2 pts
Reachable with and without www
Guests type it with or without www; neither should lead nowhere.
1 pts
Browsers stay on https (HSTS)
It stops anyone from diverting guests to an unencrypted connection on the way, in hotel Wi-Fi for example.
1 pts
No insecure http content
Browsers block such content or show a warning instead of the padlock.
1 pts
Homepage links work
A dead link right before booking costs trust, and Google reads it as a sign of a neglected site.
We follow up to 12 internal links of the homepage. 1 point with exactly one broken link.
3 pts
Legal notice linked
Germany and Spain require it for business websites; without it, legal warnings and fines can follow.
2 pts
Privacy policy linked
The GDPR requires it, and guests look for it before they send a form.
2 pts
Current year in the footer
"© 2019" in the footer makes guests wonder whether the rest still holds.
Only counts when the page shows a year; last year is enough.
1 pts
Accessibility: Lighthouse 90 or more
An accessible page is easier to read for every guest, in the sun at the beach too; in the EU, booking pages increasingly have to be accessible.
1 point from 75.
2 pts
Guest reviews on the page
Other people's words convince more than your own, especially before a first booking.
1 pts
Languages
How many languages the site speaks, whether they're linked with hreflang and whether there's an English version.
9% · 11 points
At least two languages
Guests prefer to book in their own language.
4 pts
Three or more languages
Every extra language opens its own Google search and its own AI answers.
2 pts
Language versions linked with hreflang
Otherwise Google may show guests from Germany the Spanish page.
3 pts
English version
Almost every international guest understands English.
2 pts
Video and social
Video on the homepage, a picture for shared links and linked channels: Instagram, TikTok or YouTube, Facebook.
9% · 10 points
Video on the homepage
Video shows in seconds what a place feels like.
3 pts
Picture for shared links (og:image)
Sharing happens in WhatsApp and Instagram; a link with a picture stands out in a chat and gets opened more.
2 pts
Instagram linked
Guests check the profile before they come; the link shows Google and AI that both belong together.
2 pts
TikTok or YouTube linked
Short video is how many guests discover new places.
2 pts
Facebook linked
Many guests and directories still look there for opening hours and reviews.
1 pts
Mirrors
When it's how many Mirrors
A Mirror is the unit of the rating, like a star in a restaurant guide. The three bars come from the MAD MIRROR STUDIOS logo: the more of them light up, the better the rating.
1 Mirror · Very good
From 70 points in the Mirror Score.
2 Mirrors · Excellent
From 80 points, no discipline below 50.
3 Mirrors · Exceptional
From 90 points, no discipline below 70, every discipline measured.
Test Winner
Who becomes Test Winner
The Test Winner of a category is the venue with the highest Mirror Score, provided it reaches at least one Mirror. In a tie, the venues share the title. Websites of international groups count, because guests find them the same way; they're marked. Clients of MAD MIRROR STUDIOS from the 24 months before the measurement are out of competition.
What we don't measure
- Design and taste
- Quality of food, rooms or service
- How good reviews and stars on other platforms are
- Prices
- Follower counts
- The content of subpages and booking flows (we only check that their links work)
Process
From selection to correction
01
Selection
Candidates from public lists and our knowledge of the market, fixed per edition before the measurement.
02
Verification
A website only counts if it's reachable and demonstrably belongs to the venue.
03
Measurement
Homepage, robots.txt, llms.txt and sitemap fetched once each, plus the http address, the spelling with or without www, a made-up address and up to 12 links of the homepage, and Lighthouse in mobile emulation. The same day for everyone.
04
Rating
The code of this methodology calculates Mirror Score, Mirrors and Test Winners. Nobody hands out points by hand.
05
Publication
Only those who earn something are named. Clients are out of competition and marked.
06
Correction
If a measurement went wrong, we measure again on request and correct the edition publicly.
Principles
What is never up for negotiation
- Measured, not opinion
- Every point comes from the same automated measurement, on the same day, under the same rules. Taste doesn't count: design stays out of the score.
- We only name those who earn something
- A venue is named only when it earns a Mirror, a Test Winner title or a top-five place. Nobody is put on display for a weak website.
- Not for sale
- A rating can't be bought, and work for us earns no points. Venues we worked for in the 24 months before a measurement are measured the same way, but out of competition: no Test Winner title, no Mirrors, with a disclosure.
- Open method
- Every criterion, point and threshold is public, with a version number (currently 1.1). We announce changes before an edition.
- Corrections at any time
- If a website was down during the measurement or we checked the wrong address, write to us. We measure again and correct it publicly.
- Valid for one year
- Mirrors and Test Winner titles apply to one edition. Next year we measure everyone again, winners included.
FAQ
Questions about the method
How is the Mirror Score calculated?
Each of the 55 criteria has fixed points. Per discipline, the share of points earned gives a value from 0 to 100, and the Mirror Score is the weighted mean of the seven disciplines: Google 18%, AI visibility 18%, mobile speed 22%, booking and contact 14%, tech and trust 10%, languages 9%, video and social 9%. A criterion that couldn't be measured counts neither way.
What if speed couldn't be measured?
Then the discipline drops out of the score, and the score counts as provisional. Three Mirrors are only possible with all seven disciplines measured.
Why Lighthouse in mobile emulation?
Because most guests search on a phone first, and Lighthouse is the common, repeatable yardstick for it, the one Google's PageSpeed Insights uses too. Values vary by a few points per run, which is why everyone is measured under the same conditions.
My website was down during the measurement. What now?
Write to us with the address and the date. We measure again, and if the error was in the measurement, we correct the edition publicly.
