Content Writing
Strategy & Planning
Research & Briefing
PDF
PDF Icon
PDF
Excel
Excel Icon
Excel
Web analysis
Web Icon
Web analysis
PDF
PDF Icon
Excel
Excel Icon
Web analysis
Web Icon

Localize a campaign for a new market with AI: a claim and culture audit, transcreated copy with back-translations & a per-asset sign-off register

Name the client, their site and the market they are entering, and get the claims and jokes that will not travel flagged before anyone rewrites them, local copy with a literal English back-translation beside every line, and a register that routes each asset to the person who has to clear it.

Name the client, their site, and the market they are moving into. Juma reads the brand's live pages and public campaign material, builds the source set itself, and runs the market audit before a single line gets rewritten. What comes back is a branded PDF review plus an editable CSV register, with one row per asset carrying the original wording, the verdict, the local copy, a literal English back-translation and the sign-off each line needs.

The failure this prevents is quiet. A campaign gets translated, approved by people who cannot read it, and ships a joke that lands as a threat or a recycling claim the market will not allow. 400+ marketing teams use Juma, and the localization work that survives is the work where the reviewer could see what changed. Strategy, taste, and judgment stay human. What changes is that the risky lines arrive flagged instead of discovered.

1

Take a campaign into a new market without losing the joke or the claim

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Try This Flow

Example Flow result

Hide details
  • Name the market, not the language. Germany and Austria share a language and share neither a deposit system nor a promotions rulebook. "Translate into German" produces copy. "Launch in Germany" produces copy plus the rules that decide whether it can run.
  • Ask for the audit before the rewrite. Order the verdicts first and the rewrite inherits them. Order the rewrite first and you get fluent local copy with the original problem still inside it, now harder to see.
  • Read the back-translation, not the local copy. A literal back-translation is a comprehension check, not replacement copy. If it reads smoothly and says something different from the original, that is the finding.
  • Say where every asset runs. The same line can be fine in an age-gated email and a problem in broad-reach outdoor. Placement, not wording, decides which platform and youth-protection rules apply, so put it in the first message.
  • Ask Juma to parse its own register. Request a field count per row against the header before delivery. A single unescaped comma in a sign-off column shifts every column after it and silently routes the wrong owner.
  • Keep claims and jokes in separate lanes. A joke needs a brand decision. A nutrition, environmental or comparative claim needs evidence and a named comparator. Mixing them into one approval queue is how the claim ships unchecked.
2

How do you know which campaign lines will not survive translation?

Run the audit as its own pass, before anyone writes a word of the new language. Ask for a verdict on every line and the review comes back with a fixed vocabulary: keep when the line can travel closely, adapt when it needs reworking, rebuild when a literal version breaks it, and hold when the line cannot run until evidence or mechanics are locked. Each verdict carries the reason in plain English, so a reviewer who does not speak the target language can still see why a joke about violence reads as cartoonish in one market and as a threat in another. The audit is what makes the rewrite reviewable.

Prompt
Copy

Before any rewriting, audit the source lines for this market. Give every asset one verdict from keep, adapt, rebuild or hold, state in English what fails and why, and name whether the problem is the claim, the cultural reference, the register of the humour or the placement. Keep the original wording verbatim next to each verdict.

Try This Flow
3

How do you review copy in a language you do not speak?

Approving work you cannot read is the standard way a localized campaign goes wrong. Ask for a literal English back-translation beside every local line and the review becomes possible again. Add a meaning-change column and it becomes honest: the register states what shifted between the original and the new version, or says none. That column is where a pun quietly disappears, where a joke's logic inverts, and where a softened line stops saying what the brand meant. Teams read the back-translation first and the local copy second, which is the right order when a native speaker is not in the approval chain.

Prompt
Copy

Give me a literal English back-translation of every localized line, not a polished English version. Then add a meaning-change column saying what shifted between the original and the transcreation, or none. Flag any line where a pun, a joke's logic or the strength of a claim changed, so I can approve copy in a language I do not read.

Try This Flow
4
Lightbulb Icon
Tips and tricks
This step flags what needs clearing and who clears it. It is not legal advice, and the register says so on every row that carries a claim.

Which claims need legal clearance before a campaign runs in a new market?

Nutrition wording, environmental wording, comparative claims and promotion mechanics are the four that move fastest from copy decision to legal problem. Ask for them to be separated out and each one comes back with the rule family it sits under, the evidence it needs, and the qualifying line it cannot ship without. A comparative claim gets its comparator named or gets held until the client supplies one. An environmental claim gets checked against how packaging actually works in that market, which is where a well-meant rewrite can introduce a worse claim than the line it replaced. Local counsel still decides. The register makes sure nothing reaches them unlabelled.

Prompt
Copy

Separate the claim risk from the creative risk. For every nutrition, environmental, comparative or promotional line, name the rule family it falls under, whether it is a national or a market-bloc rule, the evidence required, and the exact qualifying wording it cannot run without. Where a comparator is missing, hold the line rather than inventing one.

Try This Flow
5

How do you turn a localization review into something the team can track?

A review document gets read once. A register gets worked. Ask for the CSV to carry the fields that make it a live record: the channel and placement each asset runs in, the rule family it sits under, the owner role who signs it off, a status, a decision date and a link to the evidence. Then ask Juma to parse its own file and confirm every row has the same field count as the header, which catches the unescaped comma that silently shifts a column and assigns the wrong owner. The register becomes the thing the account team runs the clearance process from, rather than a table someone retypes into a spreadsheet.

Prompt
Copy

Turn the review into a working register. Give the CSV one row per asset with channel and placement, the rule family, the owner role, a status of open, approved or rejected, a decision date and an evidence link. Quote every field containing a comma, then parse your own file and confirm each row has exactly as many fields as the header before you deliver it.

Try This Flow

Set up your client project: market rules, approved claims, and the do-not-translate list

Run one project per client and the second market costs a fraction of the first. The things that decide what can run are stable per brand, so add them once and every market review inherits them. The brand lexicon in particular only gets written down once, and it is the item that stops a trademarked line being helpfully translated into something nobody owns.

What to add

Market rules and prior clearances

What local counsel has already ruled on, which packaging and deposit rules apply, and any platform or broadcaster policy the brand has run into before. Past clearances stop the same question going back to legal twice.

Approved claims and their evidence

The claims the brand is allowed to make, the comparator each one rests on, and where the substantiation lives. A claim without its evidence is a hold, and the register will say so.

Brand lexicon and do-not-translate list

Product names, trademarked lines, loyalty terms and the words that stay in English on purpose. Without it, a reward currency or a tagline gets translated and the equity goes with it.

Local reviewers and sign-off owners

Who clears copy, who clears claims, and who clears promotion mechanics in each market. The register routes to roles, and the roles need names behind them.

Guide Juma with project info

Each knowledge item takes a short description. One line is enough, and it tells Juma when to reach for that item instead of guessing.

  • Market rules: "Local rules and past clearances per market, including packaging, promotions and platform policy."
  • Approved claims: "Claims we are allowed to make, the comparator behind each one, and where the evidence sits."
  • Brand lexicon: "Names, taglines and terms that never get translated, with the reason each one is protected."
  • Sign-off owners: "Who clears copy, claims and promotion mechanics in each market, by role and by name."
Juma Logo
Find out what breaks before you translate it

Frequently Asked Questions

What does the localization review include?

Two files. A branded PDF covering the market take, the lines that cannot travel literally, the counts behind each verdict with the asset IDs listed so they can be checked, and the routing for everything that needs clearing. Alongside it, an editable CSV that is the authoritative register.

Each register row carries the asset, its channel and placement, the original wording verbatim, the source URL and the date it was observed, the verdict, what fails and why, the local copy, a literal English back-translation, what the meaning change was, the rule family, the owner role, a status, a decision date and an evidence link.

How is this different from translating the campaign?

Translation moves the words. This moves the intent and tells you where that was not possible. A line can be translated perfectly and still fail, because the joke's register shifts, the cultural reference is local to the first market, or the claim is allowed in one place and not another.

The audit runs before the rewrite for that reason. Each line gets a verdict and a stated reason first, so the version that comes back is traceable to a decision rather than to a translator's instinct.

Can we review copy in a language nobody on the team speaks?

That is what the back-translation column is for. Every local line comes with a literal English rendering, and a separate column stating what changed in meaning, or none.

Read the back-translation and you can see a pun collapse, a joke invert, or a claim get stronger than the original allowed. Human review on every output, and the back-translation is what makes that review real rather than a formality.

Does this handle the legal side?

It prepares the legal side; it does not replace it. Nutrition, environmental, comparative and promotional lines are pulled out of the creative queue, labelled with the rule family they fall under, and held when the evidence or the comparator is missing.

Local counsel still makes the call. What the register changes is that nothing arrives on their desk unlabelled, and no claim ships because it was mistaken for a joke. Where a rewrite would introduce a new claim, the review says so rather than presenting it as a safe fix.

Does it work for markets outside Europe, and for B2B?

Yes. The structure holds for any market move: the verdict vocabulary, the back-translation, the claim separation and the register are market-agnostic, and what changes is the rule family each row is routed against.

B2B works the same way with a different risk profile. The jokes matter less, and the regulated claims, the role and seniority conventions, and the contractual or procurement language matter more. Name the market and the audience in the first message and the audit weights itself accordingly.

Don't take our word for it

Ask AI about us

400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.

Ready when you are

What's on your plate today?

Start typing a task and Juma will figure out the rest.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.