Multilingual website restoration starts with recovering the right page in each language, not duplicating one recovered page into several language folders. Preserve surviving localized URLs, identify which translations actually exist, and rebuild language relationships only between verified equivalents. A missing translation remains a documented gap until the owner supplies or approves new content.
There are two separate acceptance questions: does each restored page accurately represent its source, and does the website send visitors and search engines to the corresponding language version? A working language selector cannot compensate for an incorrect translation. Correct text cannot compensate for a selector that sends everyone to a homepage.
Build language families before rebuilding navigation
Inventory the original paths, subdomains and domains. Record the page’s actual language as well as its apparent URL pattern. A language folder is evidence about structure, not proof that its contents were translated. Include old sitemaps, navigation links, native CMS exports and captures in the investigation.
Create one family for each equivalent subject: a particular service, product or article. Do not group pages merely because they occupy similar positions in the menu. A regional delivery page may describe a different service from a similarly named international page. Its relationship needs review rather than an automatic match.
- Family identifier
- A stable project label for the same subject across languages, independent of a new CMS record number.
- Original localized URL
- The full address, including domain, capitalization and path.
- Evidence
- The capture or owner-held source and its date. Different translations can have different surviving dates.
- Content status
- Recovered, owner-supplied, newly commissioned or unresolved. Keep these categories separate.
- Reviewer decision
- Whether the pages are equivalent and accurate enough to be connected.
This language-family record is useful even when only two versions survive. It exposes a missing page without turning it into an invented translation and provides a basis for testing each relationship later.
Resolve translation gaps and conflicting dates
Ask someone competent in the relevant language to review the visible content, not just the menu labels. Check names, quantities, dates, contact routes and service claims. A restored translation should not repeat a withdrawn offer simply because its capture is newer than another language’s capture.
If the source is incomplete, label the gap in the private recovery record. The owner may supply an existing translation, commission a new translation, or decide not to reinstate that page yet. Newly written material should be reviewed as new content, not described as recovered historical text.
A partially captured page is also different from a page with no evidence. Recover readable material without guessing the missing sentences. Image-only wording and small screenshot text require legible evidence and language review before transcription. Check reuse rights for photographs and other localized assets separately.
Connect verified equivalents with hreflang
Google’s localized-version guidance says each language version should list itself and its alternate versions using full URLs. The relationships must work in both directions. HTML tags, HTTP headers and sitemap annotations are equivalent supported methods; maintaining all three is not necessary.
Use supported language and optional region codes. A regional code alone does not identify the language. An x-default destination can support a genuine fallback or selector page; it is not a replacement for the language pages. Google determines page language from its content rather than relying on hreflang or the HTML language attribute.
Implement annotations against the approved restored address, not a Wayback replay URL or an unpublished preview. Avoid automatically assigning every English page as the alternate of a foreign-language homepage. Match equivalent content first, then produce the annotation set.
Review canonicals separately from language relationships
A canonical identifies the preferred representative of duplicate content. Hreflang describes localized alternatives. Using a canonical to send every translated page to the English homepage mixes those jobs and may undermine the pages being restored.
Review each page’s actual content and existing canonical configuration before changing anything. Google’s canonical documentation recommends a canonical in the same language when using hreflang, or the best available substitute if that is not possible. Do not replace established settings merely to impose a generic pattern.
Test alternate destinations directly: the address should return the intended public page, contain the intended language content and be available for crawling as planned. A redirect to a language homepage is not the equivalent product or article destination.
Test the visitor’s language-switching journey
From each recovered page, use the selector to visit its verified counterpart and switch back. Repeat on mobile, where selectors can be hidden in navigation. Check that the selected language is understandable, keyboard access works and the current-page relationship is retained.
Where an equivalent does not exist, the interface should make the available choice clear. Do not disguise a fallback homepage as a direct translation. Record this exception so it is not later mistaken for a broken link or filled with unrelated content.
Finish with a review of localized titles, descriptions, page language, navigation and key customer contact instructions. Treat untranslated strings as review items, including error messages and form labels. Do not test private submissions without the owner’s approval.
Accept the recovered language set, not a promised page count
The handoff should identify which language families are complete, which have approved exceptions and which remain unresolved. Keep original URLs wherever feasible and document genuine platform constraints before approving any replacement. The website restoration guide explains the broader recovery and URL-preservation process.
For a multilingual restoration scope, share the language list, domains, surviving sources and available reviewers with Wayback Restore. The useful deliverable is accurate content connected to accurate equivalents, with missing translations disclosed, rather than an apparently complete site assembled from guesses.
Questions about multilingual recovery
Can one captured language version recreate every translation?
No. It can supply source material for an owner-approved new translation, but it does not establish the original wording of missing localized pages.
Should missing translations point to any available page?
Only use a fallback deliberately and label it clearly. An unrelated page should not be presented as an equivalent translation or included in an automatic alternate-page family.
