L'auditeur de cohérence de voix de marque : évaluez tout brouillon face à votre guide de style
Pourquoi ce prompt est important
Inconsistent brand voice is invisible until a customer reads three pieces of content in one sitting and notices the company sounds like four different people. Teams that skip a structured voice check burn two to three extra review rounds per piece in back-and-forth Slack threads, and the inconsistency that slips through anyway erodes the trust a distinct voice is supposed to build in the first place.
À quoi nous l'utilisons
Your startup just brought on three freelance copywriters for a product launch sprint, and marketing ops has fifteen drafts to clear before anything ships — but the only brand reference anyone has is a scattered, two-year-old Google Doc that none of the freelancers have fully read.
Prompt
Act as a senior brand editor who has fully internalized [COMPANY NAME]'s voice guidelines and reviews content the way a copy chief would before it ships. Context: Here are our brand voice principles and guidelines: [PASTE YOUR STYLE GUIDE OR VOICE PRINCIPLES HERE — tone words, banned phrases, sentence-length preferences, and 1-2 examples of on-brand vs. off-brand copy]. Here is the draft content to review: [PASTE DRAFT CONTENT HERE]. Task: Score this draft against our voice guidelines across five dimensions: tone match, vocabulary choices, sentence structure and rhythm, use of banned phrases or words, and consistency with the provided on-brand examples. For each dimension, give a score from 1-5 and cite the specific sentence or phrase that drove the score. Constraints: - Do not fix grammar or typos — focus only on voice and tone issues - Do not suggest a full rewrite; propose the smallest possible edit that resolves each flagged issue - If a section is already on-brand, say so explicitly rather than inventing a problem to flag - Flag no more than [MAX NUMBER, e.g. 3] issues per dimension, ranked by severity Output format: 1. A scorecard table: Dimension | Score (1-5) | Example quote | Why 2. A prioritized list of the 3-5 highest-impact fixes, each with the original phrase, the suggested replacement, and a one-line reason 3. An overall verdict: Ship as-is / Ship with minor edits / Needs another pass
Résultat
Scorecard:
Dimension | Score | Example Quote | Why Tone match | 2/5 | "Our revolutionary platform leverages cutting-edge AI to seamlessly transform your workflow." | Reads like generic SaaS marketing copy, not the warm, plain-spoken voice described in the guide — four buzzwords stacked into one sentence. Vocabulary | 2/5 | "leverages", "seamlessly", "cutting-edge" | All three terms are explicitly listed as banned words in the style guide's vocabulary section. Sentence structure | 3/5 | Opening sentence runs 28 words | The guide specifies intro sentences under 20 words; this one runs long and buries the actual customer benefit under qualifiers. Banned phrases | 1/5 | "revolutionary platform" | Directly listed as a banned phrase in section 2 of the guide, with a note that it appears in nearly every competitor's homepage copy. Consistency with examples | 4/5 | Paragraph 2: "Most teams waste an hour a day on manual data entry. We built this to get that hour back." | This closely matches the provided on-brand example — plain, concrete, benefit-first, no adjectives doing the work a specific claim should do.
Priority fixes: 1. Replace "Our revolutionary platform leverages cutting-edge AI to seamlessly transform your workflow" with "Our AI handles the repetitive parts of your workflow so you don't have to." Removes three banned words and cuts the sentence from 28 to 17 words. 2. Replace "seamlessly transform" with "handles." "Seamlessly" is on the banned list; "handles" is plainer and matches the guide's preferred verb choices elsewhere in approved copy. 3. Cut "cutting-edge" entirely rather than replacing it — the guide notes this phrase describes the technology, not the customer's actual outcome, which is the opposite of what the brand voice principles ask for. 4. Keep paragraph 2 unchanged. It's the strongest on-brand section in the draft and should be used as the reference example for the freelance writer going forward.
Overall verdict: Ship with minor edits. The opening sentence needs the fix above before this goes out; everything from paragraph 2 onward is already on-brand and ready to publish as-is.
La dérive de la voix de marque est invisible jusqu'à ce qu'un lecteur rencontre trois contenus en une seule session et remarque que l'entreprise semble parler avec quatre voix différentes. Ce prompt détecte cette dérive avant publication, en transformant un jugement vague « est-ce que ça nous ressemble » en une grille de notation structurée et défendable.
Pourquoi ce prompt est construit ainsi
La section Rôle demande au modèle d'agir comme un éditeur de marque senior plutôt qu'un assistant d'écriture générique, car la tâche est évaluative, non générative — le modèle doit adopter la posture sceptique et analytique de quelqu'un qui révise des textes professionnellement.
La section Contexte exige à la fois le guide de style et le brouillon dans le même prompt, car la cohérence de voix ne peut être jugée que par rapport à une norme énoncée. Sans le guide, le modèle se rabattrait sur des heuristiques génériques de « bonne écriture » sans rapport avec votre marque spécifique.
La section Contraintes fait deux choses importantes. D'abord, « ne corrigez pas la grammaire » garde la sortie concentrée sur la voix, pas sur une correction générale. Ensuite, « si une section est déjà conforme à la marque, dites-le explicitement » existe parce que les modèles non contraints inventent souvent des problèmes mineurs pour paraître exhaustifs.
Le format de sortie — grille de notation, puis corrections hiérarchisées, puis un verdict unique — reflète la façon dont une révision éditoriale réelle est utilisée.
Où cet outil fait ses preuves
Le cas d'usage le plus précieux n'est pas des rédacteurs solo revoyant leurs propres brouillons, mais l'intégration de rédacteurs freelances ou nouveaux face à un guide de style qui existe mais que personne n'a pleinement intériorisé. Ce prompt fonctionne aussi bien comme vérification avant soumission.