How to Give AI Brand Rules for Translating Product Listings
Summary
- Start by extracting brand rules into a translation-ready “Brand Translation Brief” with do/don’t language and examples.
- Give the AI a fixed output structure (title, bullets, description, attributes, compliance notes) so translations are consistent across listings.
- Provide a term glossary with approved translations, forbidden translations, and formatting rules for units, sizes, and model names.
- Use invented test listings to validate tone, terminology, and marketplace constraints before scaling to your full catalog.
- Run a repeatable review checklist: terminology, claims, measurements, SEO keywords, and “no new facts” compliance.
If you have product listings that need translating, the hard part is not “getting a translation.” The hard part is getting a translation that stays on-brand: the same voice, the same approved terms, the same claim boundaries, and the same formatting rules across hundreds (or thousands) of SKUs.
This guide shows how to give an AI clear brand rules for translating product listings, using a practical workflow: (1) turn your source material into a structured “Brand Translation Brief,” (2) feed the AI a listing input package, (3) require a specific deliverable format, and (4) apply review checks that catch the common failure modes. All examples below are clearly labeled as invented.
What “brand rules” mean in product listing translation
For product listings, “brand rules” are not just tone notes like “friendly” or “premium.” They are operational constraints the AI must follow while translating:
- Voice and tone: formality level, sentence length, allowed adjectives, and how you address the customer.
- Terminology: approved product names, feature names, and how to translate (or not translate) them.
- Claims boundaries: what you can say, what you cannot imply, and how to handle regulated categories.
- Formatting: units, decimals, capitalization, punctuation, bullet style, and how to handle trademarks.
- Marketplace constraints: character limits, forbidden phrases, and required attribute formats.
- SEO constraints: target keywords to include (without inventing features or stuffing).
Step 1: Turn your source material into a “Brand Translation Brief”
AI performs better when your rules are explicit, testable, and written as do/don’t instructions. The goal is to convert scattered brand assets into one brief the AI can follow every time.
Source material to pull from (use what you already have)
- Brand voice guidelines (tone, style, banned words)
- Product naming conventions (model names, series, variants)
- Existing high-performing listings in the target language (if available)
- Legal/compliance notes (claims, disclaimers, regulated terms)
- Glossary/term base (approved translations, “do not translate” list)
- Marketplace rules (title length, bullet count, attribute formatting)
Brand Translation Brief (usable template)
Copy/paste and fill this in. Keep it short enough to be used repeatedly, but specific enough to constrain output.
| Section | What to include | Example content (invented) |
|---|---|---|
| Target locale | Language + region, spelling variant, currency/units | Invented: Spanish (Spain), metric units, EUR symbol where needed |
| Audience | Who buys, what they care about | Invented: Home cooks; value clarity and durability over hype |
| Voice rules | Formality, sentence length, allowed adjectives | Invented: Neutral-professional; short sentences; avoid “revolutionary,” “miracle,” “best ever” |
| Terminology rules | Do-not-translate list, approved translations, capitalization | Invented: Do not translate “AeroSeal”; translate “leak-proof” as “a prueba de fugas” |
| Claims & compliance | Forbidden claims, required qualifiers, disclaimers | Invented: No medical claims; “BPA-free” allowed only if present in source |
| Formatting | Units, decimals, bullet style, punctuation | Invented: Use “cm” and “ml”; decimal comma; bullets start with a noun phrase |
| SEO constraints | Primary keywords to include, where to place them | Invented: Include “botella térmica” in title; include “acero inoxidable” in bullet 1 |
| Output structure | Exact fields the AI must return | Invented: Title (max 180 chars), 5 bullets, description, attributes table, compliance notes |
Step 2: Provide the AI a “Listing Input Package” (what it may translate)
To prevent the AI from inventing details, give it a clean input package and a strict rule: translate only what is present, and do not add new features, certifications, or performance claims.
Listing Input Package (template)
- Source listing text: title, bullets, description, attributes
- Product facts (structured): dimensions, materials, compatibility, included items
- Allowed claims: only claims explicitly present in source or approved facts
- Prohibited claims: category-specific restrictions
- Target marketplace constraints: character limits, bullet count, forbidden terms
- Glossary: approved translations and “do not translate” terms
Invented example input package
Invented example: (This product and brand are fictional.)
- Source language: English (US)
- Target language: Spanish (Spain)
- Source title: “AeroSeal Stainless Steel Thermal Bottle, 750ml - Leak-Proof Lid”
- Source bullets:
- “Keeps drinks cold up to 24 hours and hot up to 12 hours.”
- “18/8 stainless steel. No metallic taste.”
- “Leak-proof lid with carry loop.”
- “Fits most car cup holders.”
- “Hand wash recommended.”
- Attributes (facts): Capacity: 750 ml; Material: 18/8 stainless steel; Color: Black; Care: Hand wash
- Glossary: “AeroSeal” (do not translate); “leak-proof” => “a prueba de fugas”; “thermal bottle” => “botella térmica”
- Constraints: Title max 180 characters; 5 bullets; do not add certifications (e.g., “BPA-free”) unless present
Step 3: Give the AI a translation prompt that enforces brand rules
Your prompt should do three things:
- Define the rules (Brand Translation Brief)
- Define the allowed source (Listing Input Package)
- Define the deliverable (exact output structure + checks)
Prompt template (copy/paste)
Replace bracketed sections with your content.
Task: Translate the product listing from [SOURCE LANGUAGE] to [TARGET LANGUAGE/LOCALE] while following the Brand Translation Brief and constraints below.
Non-negotiable rules:
- Translate only the provided listing content and structured facts. Do not add new features, certifications, performance claims, or guarantees.
- Follow the glossary exactly: approved translations must be used; “do not translate” terms must remain unchanged.
- Keep measurements accurate; convert units only if instructed. Preserve model numbers and variant names.
- Maintain the brand voice rules and formatting rules.
- If a phrase is ambiguous, choose the most literal translation that preserves meaning and does not increase claims.
Brand Translation Brief:
[PASTE YOUR BRIEF HERE]
Listing Input Package (source of truth):
[PASTE SOURCE TITLE/BULLETS/DESCRIPTION/ATTRIBUTES HERE]
Required output structure (return exactly these sections in this order):
- 1) Translated Title: (max [X] characters)
- 2) Translated Bullets: (exactly [N] bullets)
- 3) Translated Description: (1-2 short paragraphs)
- 4) Attributes: (key: value list; keep keys in [TARGET LANGUAGE] if required)
- 5) Compliance & Fidelity Notes:
- Glossary terms used (list them)
- Any unit conversions performed (or “none”)
- Any phrases kept in source language (and why)
- Any potential claim-risk phrases to review
Self-check before finalizing: Confirm (a) no new facts added, (b) glossary followed, (c) tone matches brief, (d) constraints met.
Step 4: Example output (invented) using the structure above
Invented example output: This demonstrates format and rule application, not a recommended final listing for your brand.
1) Translated Title
AeroSeal Botella térmica de acero inoxidable, 750 ml - Tapa a prueba de fugas
2) Translated Bullets
- Frío hasta 24 horas y caliente hasta 12 horas.
- Acero inoxidable 18/8. Sin sabor metálico.
- Tapa a prueba de fugas con asa de transporte.
- Compatible con la mayoría de portavasos de coche.
- Se recomienda lavar a mano.
3) Translated Description
La botella térmica AeroSeal de 750 ml está fabricada en acero inoxidable 18/8 para un sabor neutro. Su tapa a prueba de fugas incluye un asa para transportarla con comodidad.
Mantiene las bebidas frías hasta 24 horas y calientes hasta 12 horas. Se recomienda lavar a mano.
4) Attributes
- Capacidad: 750 ml
- Material: Acero inoxidable 18/8
- Color: Negro
- Cuidado: Lavado a mano
5) Compliance & Fidelity Notes
- Glossary terms used: “AeroSeal” (sin traducir), “botella térmica”, “a prueba de fugas”.
- Unit conversions performed: none.
- Phrases kept in source language: “AeroSeal” (marca).
- Potential claim-risk phrases to review: time-based performance claims (“hasta 24 horas / 12 horas”) should match source exactly.
Step 5: Review checks (catch the common translation failures)
Use these checks as a repeatable QA pass. They are designed for product listings where small wording changes can create compliance or customer-expectation issues.
1) “No new facts” check
- Did the AI add certifications (e.g., “BPA-free,” “FDA approved”), materials, or compatibility not present in the input?
- Did it upgrade claims (e.g., “leak-proof” becoming “100% leak-proof”)?
- Did it add use cases (e.g., “perfect for hiking”) that were not stated?
2) Glossary adherence check
- Are all “do not translate” terms unchanged (brand names, model numbers)?
- Are approved translations used consistently (same term translated the same way in title, bullets, and description)?
- Are forbidden translations avoided (false friends, misleading equivalents)?
3) Measurements, units, and numerals check
- Are sizes, capacities, and counts identical to the source (or correctly converted if instructed)?
- Are decimal separators and unit abbreviations correct for the locale?
- Are ranges and “up to” claims preserved without strengthening?
4) Tone and brand voice check
- Does the translation match your formality level and sentence length rules?
- Are banned adjectives removed?
- Does it avoid slang or region-specific phrasing you do not want?
5) Marketplace constraints check
- Title length within limit?
- Correct number of bullets and required fields present?
- Any forbidden terms for your category or marketplace?
6) SEO placement check (without inventing)
- Are target keywords included where required (title or first bullets) without awkward stuffing?
- Did the AI insert keywords that imply features you do not have?
Step 6: Scale safely: build a reusable “translation kit” per brand and locale
Once you have one listing translating correctly, freeze the components into a kit you can reuse:
- Brand Translation Brief (stable across the brand)
- Locale rules (units, punctuation, spelling variant)
- Glossary (approved translations + do-not-translate list)
- Output structure (the deliverable format)
- QA checklist (the review checks above)
Then, for each SKU, you only swap the Listing Input Package. This keeps translations consistent while reducing the chance of drift.
Frequently Asked Questions
FAQ 1: What brand rules should I give an AI for translating product listings?
Answer: Give rules the AI can execute: voice (formality, sentence length), terminology (approved translations and do-not-translate terms), claims boundaries (what cannot be implied), formatting (units, capitalization, punctuation), marketplace constraints (title length, bullet count), and SEO constraints (which keywords must appear and where). Put them into a single Brand Translation Brief so the AI is not guessing across multiple documents.
Takeaway: Operational rules beat vague tone adjectives.
FAQ 2: How do I stop the AI from adding features or making stronger claims in translation?
Answer: Make the “source of truth” explicit (your Listing Input Package) and add a non-negotiable rule: translate only what is provided. Then require “Compliance & Fidelity Notes” where the AI lists glossary terms used, unit conversions, and any claim-risk phrases. Finally, run a “no new facts” check that compares each translated claim back to the source text and structured facts.
Takeaway: Constrain inputs, force a structured output, and review for claim inflation.
FAQ 3: Should I translate brand names, model numbers, and feature names?
Answer: Decide this in your glossary. Many teams keep brand names and model numbers unchanged, while translating descriptive feature terms. The key is consistency: if a feature name is a trademarked or official label, list it as “do not translate.” If it is descriptive, provide an approved translation and require the AI to use it everywhere (title, bullets, description, attributes).
Takeaway: Put naming decisions in the glossary so the AI does not improvise.
FAQ 4: How do I handle units, sizes, and measurements across locales?
Answer: Add a formatting section to your brief: whether to convert units (and how), which unit abbreviations to use, and how to format decimals for the locale. If you do allow conversions, require the AI to list them in the “unit conversions performed” note so reviewers can verify accuracy quickly.
Takeaway: Treat measurement handling as a rule, not a translation preference.
FAQ 5: What output format should I require for consistent listing translations?
Answer: Require the exact fields you publish: translated title (with character limit), a fixed number of bullets, a short description, and an attributes list. Add a final “Compliance & Fidelity Notes” section so the AI surfaces glossary usage, conversions, and potential claim-risk phrases. This makes outputs easier to compare across SKUs and easier to QA.
Takeaway: A fixed deliverable structure reduces inconsistency across listings.
FAQ 6: How do I build a glossary the AI will actually follow?
Answer: Keep it explicit and enforceable: for each term, include the source term, approved translation, forbidden translations (if relevant), and a note like “do not translate” for brand/model names. Then instruct the AI that glossary entries override its own preferences, and require it to list which glossary terms it used in the output notes.
Takeaway: A glossary works best when it is both prescriptive and auditable.
FAQ 7: How can I review AI-translated listings quickly without missing important issues?
Answer: Review in passes: (1) “no new facts” and claim strength, (2) glossary adherence and naming consistency, (3) measurements and numerals, (4) marketplace constraints (lengths, bullet count), and (5) tone. The “Compliance & Fidelity Notes” section helps you spot where the AI made choices that deserve a closer look.
Takeaway: A checklist-based QA pass catches the highest-risk errors fast.
FAQ 8: Can I reuse the same brand translation rules across Amazon, Shopify, and other channels?
Answer: You can reuse the core Brand Translation Brief (voice, terminology, claims boundaries, formatting), but you should maintain channel-specific constraints (title length, bullet structure, required attributes, forbidden phrases) as separate add-ons. In practice, keep one brand-and-locale kit, then attach a small “channel constraints” section per marketplace.
Takeaway: Reuse the brand core, swap the channel constraints.