How do you translate a Docebo course for several languages without duplicating the learner journey, changing the grading rules or losing track of completion? Begin with the structure of a multilingual course, not with a list of sentences. Docebo training materials may sit inside language-specific folders, while other materials can be shared across learners. What appears to be a simple translation therefore involves navigation, evidence, learner choice and course governance.
This practical Docebo localization guide explains language folders, fallback language, course titles and descriptions, training materials, shared assets, tests, navigation policies, reporting and the difference between an interface language and the language a person actually studies. It is intended for global learning teams, educators, course owners and specialist translators. The examples are fictional, but they reflect decisions that real organizations must test in their own configured portals.
Docebo introduced multi-language eLearning course support in 2025, organizing training materials within language folders and retaining a common course structure. Official discussion at that time explicitly distinguished multilingual course content from automatic translation of course metadata and from any built-in machine translation engine. Product capabilities may change, so verify the currently enabled version before promising a localized course shell or automatic content creation.
The short answer: translate the learner’s obligations, not merely the learner’s sentences
A successful multilingual Docebo course is not just one set of attractive lessons per language. Each learner must be able to find the right edition, understand the same expectations, perform the same essential tasks, receive equivalent assessment feedback and leave with a trustworthy completion record. Language selection and fallback should be explicit rather than accidental. The source-language course remains an auditable reference, but it should not become an invisible barrier to the person studying in another language.
Think in three connected layers. First comes meaning: instructional explanations, warnings, examples and assessment questions. Second comes behavior: folder visibility, prerequisites, ordering, retries, completion rules and links. Third comes evidence: which course and materials were delivered, what the learner did and what a manager may truthfully infer. Treat a defect in any one layer as a localization defect even if the translated prose is grammatically excellent.
How the Docebo multilingual course model changes the translation problem
In a course that uses language-specific training-material folders, one shared course identity can serve several localized learner experiences. This is conceptually different from publishing three entirely separate courses with different enrolments and progress histories. It may reduce administrative duplication, yet it also makes it important to document which parts are language-specific and which are universal. A single shared item might be beneficial for a language-neutral demonstration but unsafe if it contains a warning only one audience can read.
Official Docebo guidance explains that navigation policies operate on the training materials associated with a learner’s selected language and, where relevant, common materials outside language folders. It also distinguishes free and sequential navigation, mandatory first material and a locked final item. Those are teaching and operational decisions. Translators should not alter them as a side effect of changing text. They must make the visible instructions describe the actual configuration.
A separate caution concerns metadata. Product discussions about the initial multilingual release noted that course names and descriptions were not automatically localized along with materials. This distinction is not a reason to avoid multilingual courses; it is a reason to audit the course card, search results, catalogue text and learner opening page individually. Do not infer a feature exists today solely because an older roadmap mentioned it, and do not claim that a language folder performs AI translation.
Build an inventory before any translation
List every source item with a stable identifier, type, source language, target language, owner, latest approved revision, required learning outcome, content location and external dependencies. Distinguish instructional text from configuration, third-party package data and records of previous learning. A quiz stem can be translated; a historical score must not be rewritten. A localized filename may be useful; an opaque integration identifier must remain intact. The inventory makes these decisions visible before a translator encounters them by accident.
Then mark the level of consequence. A friendly introduction needs an editorial review. A safety warning needs an expert review. A graded question needs checking of every answer option and feedback state. A privacy notice needs authorized approval. A course completion claim may require a separate credential owner. This graduated approach is efficient because it concentrates effort where a misunderstanding changes the learner’s rights, duties or record.
Worked example: a multilingual laboratory induction course
Imagine a fictional laboratory induction delivered to employees in English, French and Spanish. Every version teaches the same sequence: read the hazard map, identify protective equipment, practise responding to a spill, answer a ten-question assessment and acknowledge the emergency contact route. The organization chooses one shared course identity, with language-specific lessons and one common diagram whose labels are supplied accessibly in each language. The source course records an 80 percent passing threshold and a maximum of two attempts.
The translation review discovers four discrepancies. The French materials place the final test before spill-response training. The Spanish PDF still contains English emergency telephone instructions. The course card is titled only in English. And the Spanish completion message claims the worker is licensed to operate machinery, although the course is an induction rather than a professional licence. None is fixed by simply improving sentence fluency. The team must repair sequencing, supply an equivalent PDF, clarify discovery and correct the credential claim.
The acceptance test gives a fresh learner account each language, runs through all items in the actual player, deliberately fails a practice assessment, then passes it. Reviewers compare instructions with unlock rules, inspect attachments and confirm that the completion record is consistent with what the learner achieved. The organization retains the original grade and event records and notes which localized revision was tested. That is a defensible definition of a translation release.
Thirty-two practical localization problems and the repairs that make a difference
1. Course profile language
Consider what happens in a real learning journey. A new employee registered with English as a provisional language, then selected French after enrollment; the assigned training is not necessarily the same as the account preference. Confusing interface language with the language chosen for an individual course can leave a learner in the wrong lessons. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Check the learner’s actual course-language selection, enrollment state and available variants before editing source material; explain the choice in the learner’s own words. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Enroll a fabricated user, change the profile preference and reopen the course; record precisely what language appears and whether an explicit switch is required. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
2. Language-folder ownership
This problem usually becomes visible only when a learner rather than an editor opens the course. One safety course contains English, Spanish and Arabic training materials in dedicated folders, each maintained by a different instructor. A folder label alone does not prove that every equivalent lesson, test and supporting attachment exists in that language. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Create a source-to-target inventory using a stable lesson identifier, a language owner and a completeness status for every folder. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Walk each folder with a learner account and compare actual displayed items against the approved inventory. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
3. Default and fallback choices
The important distinction is between what the learner reads and what the platform actually does. A bilingual course supports German and French, but a learner who prefers Dutch opens it from a catalogue. An unannounced fallback can make the content look broken even when the software is behaving as configured. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Choose and document an approved fallback language; include a human-readable explanation that a complete Dutch edition is not available. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Test an unsupported preference and a supported preference, checking exactly which variant appears in the course player. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
4. Shared safety lesson
A fluent sentence can still describe the wrong action. A globally shared emergency procedure sits outside language folders to avoid maintaining multiple versions. A shared material may remain unreadable to people unable to understand the language of the common resource. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Use a genuinely language-neutral asset or provide an accessible multilingual version with equal duties, contact instructions and emergency sequence. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Ask target-language reviewers to finish the shared lesson without relying on the source language; verify the same emergency action is understood. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
5. Course title and metadata
Consider what happens in a real learning journey. The training body is French but the course card, descriptive excerpt and administrative title remain English. Multi-language course support originally addressed training materials, not automatic translation of the entire course shell or its metadata. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Inspect the current release before promising multilingual titles; provide an agreed bilingual title convention and supplementary catalogue guidance where permitted. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Review card, search results, course pre-page and opening lesson separately, marking what has and has not actually been localized. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
6. Folder ordering
This problem usually becomes visible only when a learner rather than an editor opens the course. In the English folder the introduction precedes laboratory access instructions, while the Spanish folder lists the test first. Translation can preserve every sentence yet change the pedagogical prerequisites if lesson order diverges. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Map the intended learning sequence and reorder the target-language materials to preserve requirements, especially before graded work. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Compare a first-time English learner with a first-time Spanish learner and note each unlock point. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
7. Sequential navigation
The important distinction is between what the learner reads and what the platform actually does. A course is configured for sequential navigation, and a translated lesson has been placed after the final assessment. Docebo applies navigation rules to the selected language’s learning-material sequence, not merely to a document outline. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Check course navigation policy together with the order inside each language folder; protect dependencies before publication. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Test next, previous and locked steps from a fresh learner account in both languages, not only the admin preview. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
8. First mandatory item
A fluent sentence can still describe the wrong action. The opening compliance confirmation is translated as an optional introduction, although the course requires it before other materials. The target wording understates an operational requirement and may cause unnecessary support tickets or invalid completion claims. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Preserve explicit mandatory language and ensure the first material is the correct requirement in each version. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Try bypassing the material, then completing it, and compare learner-facing instructions with the actual unlock behavior. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
9. Final material lock
Consider what happens in a real learning journey. An end-of-course test is locked until earlier materials are finished; the French closing page merely says the test is available now. Language that describes premature availability contradicts the configured final-material rule. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Rewrite the instruction to state the precise prerequisite and keep the locked test in the intended final position. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. In an incomplete account confirm the lock and message; in a completed account confirm the test becomes available. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
10. Maximum attempts
This problem usually becomes visible only when a learner rather than an editor opens the course. Learners are allowed two graded attempts, but the translated help text casually promises unlimited retries. An inaccurate promise can influence learner decisions, time spent and complaints about a failed assessment. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Verify attempt restrictions for each scorable training material before translating retry language; keep policy terms precise. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Exhaust attempts in a sandbox test account and check the status, explanation and support route. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
11. Assessment distractors
The important distinction is between what the learner reads and what the platform actually does. A multiple-choice safety question contains two technically close answers whose translated wording becomes identical. Fluency checks miss the serious failure: the target-language assessment can no longer discriminate understanding. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Have a subject specialist review stems, distractors, correct choices, scoring and rationales in the same pass. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Perform an answer-key comparison and a small bilingual pilot, documenting ambiguity rather than accepting matching answer letters alone. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
12. Passing score descriptions
A fluent sentence can still describe the wrong action. An 80-percent passing rule appears in the platform, while one translated instruction says seventy points are sufficient. Numbers and grading thresholds are business logic, not stylistic choices for the translator. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Treat the numeric threshold and its unit as protected and ensure all lessons, feedback and learner emails refer to the same rule. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Submit boundary-case test attempts and observe the actual pass, fail and retry messages. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
13. Video narration and subtitles
Consider what happens in a real learning journey. The German course links to a French voiceover and an Italian caption track despite having a correctly translated text handout. Language labels and media versions must agree; subtitles do not magically make unsafe narration equivalent. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Catalogue audio language, subtitle language, time synchronization, non-speech cues and attachment ownership for every video. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Play the whole clip with sound off and then audio only, asking separate reviewers to confirm the same instruction sequence. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
14. PDF attachments
This problem usually becomes visible only when a learner rather than an editor opens the course. A target-language training folder includes a translated page but still links to the source-language laboratory checklist. Learners may act on an inaccessible document after completing a fluent online lesson. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Translate the checklist as a controlled artefact, preserve revision numbers and check text embedded in diagrams as well as visible paragraphs. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Download each resource as a learner and compare its filename, version, contents and accessibility with the intended locale. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
15. Image-only procedures
The important distinction is between what the learner reads and what the platform actually does. A machine-maintenance PDF uses screenshots where the important warning text is burned into the picture. Translation memory cannot reliably extract text that was never stored as editable prose. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Rebuild source graphics or place equivalent accessible text beside them, taking care not to alter component names or control positions. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Check zoom, screen-reader descriptions and native-language comprehension of each warning image before sign-off. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
16. Central repository reuse
A fluent sentence can still describe the wrong action. A shared course material from the central repository appears in several courses, and one team requests a localized replacement. Replacing a reused asset without checking its consumers can introduce an unwanted language change elsewhere. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Identify the shared asset, owners and dependencies; determine whether a separate localized resource or approved update is appropriate. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Open every affected course’s player in representative learner contexts before accepting a shared-material change. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
17. Content marketplace variants
Consider what happens in a real learning journey. A licensed Docebo Content item is offered as one multilingual asset, while another collection presents separate language-set items. Those are different purchasing and import structures, and their language choices should not be treated as interchangeable. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Inspect the item’s language availability and import behavior; distinguish multiple languages inside one item from separate catalogued versions. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Preview each candidate and confirm language switching, course count and learner visibility after import. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
18. Content search labels
This problem usually becomes visible only when a learner rather than an editor opens the course. Imported training material has been renamed for a regional team while the library displays the original title. Mismatched naming across course search and vendor catalogues can make the correct content difficult to locate. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Keep a source title, approved display title, asset identifier and locale alias, and document who owns the naming decision. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Search by original and localized terms and ask a target-language learner to locate the correct resource without insider knowledge. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
19. Certificates and achievements
The important distinction is between what the learner reads and what the platform actually does. A learner completes the Spanish materials and receives a certificate whose achievement title is still English. A translated learning experience does not guarantee certificate fields, credential claims or regulatory wording are localized. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Inspect the certificate’s issuing rules and wording separately; preserve award date, identifier and attested learning outcome. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Complete the course in each target language and inspect the generated credential and learner transcript. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
20. Completion reporting
A fluent sentence can still describe the wrong action. A manager sees a completion recorded in Italian but assumes the employee trained in English because the shared course title is English. Language information and course identity answer different questions; reports can collapse them if not interpreted carefully. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Define a reporting column for actual selected learning language, where supported, alongside the single course identity and completion state. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Compare the report against known test learners who chose different variants, without changing their historical attempts. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
21. Regulated training equivalence
Consider what happens in a real learning journey. A compliance department needs proof that French and English learners received the same legally approved safety duties. Word-for-word translation is not enough when the two editions omit different exceptions or caution statements. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Create an obligation-by-obligation equivalence checklist reviewed by a qualified domain owner, keeping the effective source version visible. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Ask bilingual reviewers to explain the required action from each variant and verify that no condition was added or lost. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
22. Instructor-led sessions
This problem usually becomes visible only when a learner rather than an editor opens the course. An administrator assumes the eLearning multi-language folder feature also translates instructor-led session details and participant messages. Platform functions for eLearning materials must not be projected onto separate event or session features. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Scope the article’s advice to the course types actually supported; prepare independent multilingual notices and instructions for instructor-led events. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Register test learners for the event and inspect date, venue, timezone and communications in each language. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
23. Date and timezone
The important distinction is between what the learner reads and what the platform actually does. A live session reminder is written for a global audience without making the event timezone explicit. Translating a friendly time phrase can move the perceived appointment across timezones or even calendar dates. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Store the event instant separately from localized display text; include the timezone and convert only through verified portal behavior. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Invite users in two zones and have them restate the start time in their local calendars. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
24. Notifications and emails
A fluent sentence can still describe the wrong action. A welcome notification uses an English-only subject, while the French learner receives a localized course invitation body. A fragmented message may look suspicious and can obscure the correct action or link. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Inventory subject, preview text, body, personalization variables, URLs, sender name and opt-out language; check supported template localization settings. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Send test messages and inspect a real email client rather than trusting the notification editor. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
25. Completion feedback
Consider what happens in a real learning journey. A course passes the learner, but the target-language congratulation message suggests a professional licence has been granted. Course completion and legal certification are different outcomes, even when both appear near the same button. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Define precisely whether the record is attendance, training completion, competence, accreditation or an externally issued licence. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Compare the feedback text with the certificate and administrator report, and obtain legal review for regulated claims. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
26. Mobile player views
This problem usually becomes visible only when a learner rather than an editor opens the course. Desktop French instructions fit cleanly, but mobile navigation truncates a critical step label after translation. Language expansion can hide necessary decisions in narrower player layouts. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Test actual mobile breakpoints and rewrite only the display label where the underlying action remains unambiguous. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Complete a representative task on a phone and tablet with screen magnification, keyboard navigation where supported and normal touch. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
27. Right-to-left languages
The important distinction is between what the learner reads and what the platform actually does. Arabic content is placed in a language folder but borrowed English identifiers and numerals appear in the wrong visual order. Directionality affects punctuation, inline codes, tables and accessibility, not just alignment. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Keep identifiers atomic, choose bidirectional formatting carefully and review the real course player rather than translating the HTML alone. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Have a fluent Arabic reviewer test instructions, code strings and form navigation on real devices. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
28. External SCORM packages
A fluent sentence can still describe the wrong action. A translated SCORM learning object displays target-language slides but reports completion to Docebo differently from the source package. The package’s runtime data and completion conventions are separate from visible translation. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Preserve manifest structure and tracking logic; validate the translated package in the intended course and navigation policy. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Run both passing and failing sample journeys and inspect the platform’s recorded results. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
29. Privacy notices and consent
Consider what happens in a real learning journey. A French course asks for voluntary feedback, but the adjacent English form collects identifiable information without an understandable notice. A translated course may still send users to privacy-sensitive external material that has not been localized. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Map the data collected, purpose, recipient and retention notice; use an approved legal-language edition before requesting information. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Open all linked forms as a target-language learner and verify the notice is visible before submission. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
30. Release changes
This problem usually becomes visible only when a learner rather than an editor opens the course. An administrator updates the source course after Spanish and Portuguese versions were previously approved. Course language folders can drift apart silently when the original content changes without a version-tracking rule. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
A useful repair protects both instructional meaning and the existing course record. Keep source revision, target revision and review owner in a change ledger; flag changes to obligations, answer keys or resources for priority review. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Make the test reproducible: note the language, learner state, settings and actual result. Compare release inventories and run regression checks on the changed materials before marking the language editions current. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
31. Accessibility of exercises
The important distinction is between what the learner reads and what the platform actually does. A translated knowledge check uses color alone to distinguish correct from incorrect answers and a screen reader cannot identify the feedback. Translation quality is not complete when the learner cannot perceive the instructional result. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Do not assume that an available configuration is correct merely because it is accepted by the editor. Supply explicit textual feedback, meaningful labels and keyboard-operable elements, then have accessibility reviewers examine all variants. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Ask whether a learner who never saw the English course would still know what to do next. Test screen-reader announcement, focus order and magnification using the actual published lesson. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
32. Proctoring and exam rules
A fluent sentence can still describe the wrong action. A high-stakes final assessment advertises proctoring in one language but not in the other. The learner must understand surveillance, equipment, privacy and attempt conditions before consenting to the assessment process. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Keep a source-to-target comparison that another teacher or administrator can understand. Use approved equivalent rules and refer to the institution’s proctoring policy rather than inventing permissions in translation. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
Record any unsupported feature or unresolved discrepancy as a release blocker rather than concealing it. Inspect the exam entry screen and any consent or system-check workflow in both languages. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
33. Help and escalation route
Consider what happens in a real learning journey. The target-language course points to an English-only support mailbox with no explanation of expected response language. The learner may abandon an otherwise correct course after a simple access or assessment issue. The question is not whether isolated terms can be substituted, but whether the same person can understand the same responsibility, prerequisite or result without needing to know the original language. In a shared-course environment, a small mismatch can propagate into a surprising learner state.
Start with the smallest controlled change that restores equivalence. Provide a truthful support-language statement, translated contact instructions and relevant course identifiers for the support team. Review the immediate teaching context as well as any dependent resource, setting, support instruction or assessment outcome. Where the platform keeps an identifier, enrolment relationship, grade or completion event, protect that data from text-level editing; a wording correction does not authorize a history rewrite.
An admin screenshot is useful evidence, but it cannot replace the learner’s journey. Submit a harmless test enquiry and verify the learner can identify whom to contact and what information to share. A complete check includes the target-language wording, the relevant behavior and the reported outcome. If one of these is unavailable to the reviewer, label that particular part unverified and do not turn it into a confident promise to learners.
A four-week Docebo localization and testing plan
Week 1 — Scope and inventory
Identify course ownership, source revisions, target audiences, course types, source and target folders, shared materials, assets, assessments and official configuration boundaries. Capture current learner-language behavior with safe test accounts before making assumptions about fallback or profile preferences. Separate material that can be rewritten from records or settings that must be preserved.
Week 2 — Translation and subject review
Translate and edit lessons, subtitles, attachments, quiz text, feedback and navigation guidance. Use domain experts for safety and compliance content and independent reviewers for assessment equivalence. Check the course shell separately, especially metadata, catalogue labels, certificates and notifications, so that a visible English-only element is never wrongly claimed to be translated.
Week 3 — Functional learner testing
Run the intended path in every supported language using new learners and a small matrix of unsupported preferences, failed attempts, completed attempts, mobile screens and accessibility settings. Validate sequencing, mandatory first items, final locks, score thresholds, course-language choice and evidence in reports. Do not use a passing administrator preview as a substitute for this test.
Week 4 — Controlled release and maintenance
Obtain sign-off from language, pedagogy, platform and compliance owners. Publish only verified language variants, keep reversible source versions and record a compact release note with known limitations. Establish a change-trigger rule so future edits to the source lessons, attachments, exams or legal wording cause a targeted recheck of the corresponding translations.
Frequently asked questions
Does Docebo automatically translate a course when I add another language?
Not simply by creating a language folder. The multilingual course feature organizes learner-facing materials by language. The organization must provide suitable translated assets or use separately evaluated translation services. Check the actual feature set in the current tenant and distinguish delivery configuration from translation generation.
Do the same course title and description appear in every language?
The initial multi-language feature focused on training materials; official vendor discussion said automatic course metadata localization was not part of that release. Verify the present product and plan independently and audit the course card and catalogue. A multilingual lesson does not prove multilingual discovery.
Can shared materials be used across language folders?
Yes, the documented model permits common training materials, but shared is not synonymous with language-neutral. If the item contains instructions, required statements or assessment information, it must be understood by every intended language group or replaced with equivalent accessible variants.
Should I create a different Docebo course for every language?
Not by default. Dedicated folders can support language variants within one eLearning course, while some products, integration constraints or business processes may still justify separate courses. Decide by learner identity, reporting requirements, enrolment rules, certification and maintenance burden rather than by a universal rule.
Will a language change alter a learner’s previous grades?
A display or teaching-language change should not be used to rewrite historical results. Test what the enabled course model actually does with current enrolment and language selection; preserve previous attempts and distinguish reported completion from the translated wording shown later.
What about imported marketplace content?
Inspect whether the licensed item contains multiple selectable languages inside one content object or is supplied as separate language-set versions. Verify title, actual media language, subtitles, available assets and import behavior before relying on a catalogue language badge.
Is translating quiz answers enough for a fair assessment?
No. The stem, all distractors, hints, feedback, vocabulary complexity and scoring must be checked together. Target-language candidates should face equivalent reasoning difficulty, not merely see correctly spelled translations of the answer letters.
When should publication be blocked?
Block an edition when a required lesson or attachment is missing, a safety obligation changes meaning, assessment scoring has not been verified, a learner cannot reach required content, privacy notices are defective or completion records are being described inaccurately. A visible limitation is better than a false claim of readiness.
Official references and connected translation guides
Official product documentation and vendor discussions are useful starting points, but features, plans and interfaces may change. Check the currently enabled Docebo environment before applying any instruction. Docebo: Managing course navigation policy; Docebo: Managing Docebo Content; Docebo Community: multi-language course support Q&A; Docebo Community: multilingual reporting update.
For related but distinct technical tasks, read SCORM course runtime and sequencing, QTI assessments and scoring, TalentLMS course translation, and the Translate Like a Pro series index. This guide covers the Docebo platform decisions rather than replacing those separate standards.
The practical goal is simple: a learner who does not know the source language should still find the right course, follow the right sequence, understand their responsibilities, answer a fair assessment and receive an accurate record. Everything else is a supporting technique.