Translation is not localisation. Demand has a different shape in every market.
Ten points on the scale, and the layer where one misconfigured attribute makes four markets compete with each other instead of with the competition.
What actually breaks
7 defects · observed, not hypotheticalA translated site assumes every market asks the same question in a different language. They do not. Volume, intent and vocabulary all move — and in some markets the demand you are translating for is simply absent, which is a finding worth having before you pay for the translation.
Then there is the machinery. Language targeting is one of the few areas of search where a technically valid configuration can be completely ignored, and where the failure is invisible from inside any single market.
| Defect | What it looks like | What it costs |
|---|---|---|
| hreflang without return tags | Page A points at B; B does not point back at A. | The whole cluster is disregarded and the wrong language surfaces. |
| Translated pages, untranslated metadata | Body copy localised; titles and descriptions still in the source language. | The market is shown a result written for somebody else. |
| Keywords translated word for word | The literal translation is not the phrase that market types. | You answer a question nobody in that language asks. |
| Country targeting by currency alone | The market switcher changes the price and nothing else. | Nothing distinguishes the pages, so they compete as duplicates. |
| Legal copy contradicting the live configuration | Terms restrict sales to one country while checkout accepts twelve. | Trust and conversion both take the loss, in every market but one. |
| x-default missing or pointing nowhere | No page claims responsibility for the unmatched visitor. | The market with the most traffic gets served to everyone. |
| One market ranking in another market’s results | Two locales, the same intent, no distinguishing signals. | You out-compete yourself and pay twice for it. |
What it moves on the scale
Weights, not scoresMoves 25 of 100 points. The muted categories are not this service’s job, and saying so is the point: a proposal that claims to move all eight is describing a retainer, not a diagnosis.
What you get
Same format as every Atlas documentDelivered as a document, not a dashboard login: dark cover, verdict columns, evidence IDs, a provenance line stating where every number came from. Shipped in two-week loops — the smallest change that moves the number, then the next one.
Proof
What exists, and what does notNo fabricated clients, statistics, testimonials or rankings appear on this site. Where a figure is not evidenced it is not printed, and the absence is stated instead.
Common questions
Answer first, then the reasoningWhat breaks hreflang most often?
Reciprocity. If page A declares page B, page B must declare page A, or the whole cluster is disregarded — so the failure is rarely one wrong page. It is a market that quietly stops being served the version written for it.
Is a translated keyword list a localised one?
No. Translation preserves words and loses purpose. The same product is searched for by material in one market, by occasion in another and by brand in a third — and in some markets the demand a plan assumes is simply absent.
What does x-default do?
It names the page to serve when no declared language or region matches the visitor. Without it, the market with the most signals becomes the fallback by accident.
Where it bites hardest
By sectorThree ways to start
None of them is a discovery callScan your own site
A visibility scorecard against the published 100-point scale, an hreflang validator, an answer-readiness reader and a SERP preview measured in pixels — all running in your browser. Nothing is uploaded and nothing is stored.
Run a scanThe AEO & GEO field guide
What changes when a result is assembled rather than listed — and the eight things to fix first. Written for engineers and founders, not for a newsletter funnel.
Read the guideBook a working call
Bring the site and the constraint you already suspect. You leave with the diagnosis whether or not there is an engagement in it.
Book on Cal.comAtlas reads the site before the first conversation. Seven questions; what comes back is a reading against the 100-point scale, the bottleneck named, and the smallest change that moves it.