A prompt is the written instruction given to a generative AI so it produces text, an image or a video. For brand content, a good prompt specifies the role, the task, the context and the expected format, and includes the brand DNA (tone, vocabulary, limits). According to Google Workspace's 2024 guide, the most effective prompts run about 21 words with relevant context.
What is a prompt, exactly?
A prompt is the message a person writes to a generative AI tool to describe the expected result: "write a 120-word LinkedIn post announcing our new autumn menu, warm tone, no emoji." The AI knows nothing about the business, its customers or its usual way of communicating: the prompt is the only information it has, beyond whatever the tool may have already stored. A brand prompt is therefore an instruction that passes the business's own editorial rules to the AI, either with every request or once and for all.
Why is the prompt often an SME's weak point?
The prompt is often the weak point because it's too short. The Google Workspace with Gemini: Prompting 101 guide, published by Google in September 2024, states that the most effective prompts run about 21 words on average with relevant context, while prompts people actually try are generally under nine words. A request like "write a post for my restaurant" leaves the AI to invent a tone, a style and details that match nothing real: the result is generic because the instruction was.
This now matters for a significant share of small businesses. According to the France Num 2025 barometer presented by France's Directorate General for Enterprises, 26% of French micro and small businesses use AI solutions, twice as many as in 2024. The more SMEs adopt generative AI, the more the quality of their instructions becomes a differentiator between generic content and on-brand content.
What should a good brand prompt include?
A good brand prompt includes four elements, which Google's 2024 guide calls "Persona," "Task," "Context" and "Format," plus the business's own rules. The method is to fill in each element in this order:
- The role (persona): "You are the communications manager of an artisan bakery in Lyon."
- The task: "Write an Instagram post announcing the king cake (galette des rois)."
- The context: the target audience, the post's goal, the information to convey (real dates, prices, hours).
- The format: length, social network, structure, call to action.
- The brand rules: tone, words to favor, forbidden phrasing, topics to never raise (unkept promises, comparing to a competitor, unverifiable figures).
A prompt that includes these five blocks naturally runs past the 21 words Google mentions, which is not a problem: what matters is that every word gives the AI useful information.
What's the difference between a generic prompt and a brand prompt?
A generic prompt only describes the task, while a brand prompt also describes the business's identity and its limits. The table below compares three levels of instruction for the same need.
| Prompt level | Example | Likely result |
|---|---|---|
| Generic | "Write a post for my restaurant." | Generic text, invented tone, imaginary details |
| Precise | "80-word Instagram post announcing our €16 lunch menu, friendly tone." | Usable text, but style varies from post to post |
| Brand prompt | Precise prompt + tone charter, in-house vocabulary, limits, examples of approved posts | Text consistent with the brand identity, fewer corrections before publishing |
The difference mainly shows up in consistency. A precise prompt can produce one good post; only a brand prompt, reused for every request, produces a series of content that looks alike and that the audience recognizes.
Do you need to rewrite the prompt for every post?
Rewriting the full prompt for every post is neither realistic nor necessary for an SME posting several times a week. The solution is to formalize the brand rules once in a brand AI charter saved in the tool, then only write the variable part day to day (topic, format, date). On a platform like Isalis, the brand DNA is entered once and reused for every image, short video or text generated for social media, which avoids restating the business's context with every request.
What I often see while building Isalis is that SME owners underestimate the "limits" part of the prompt: specifying what the brand never says often improves the result more than adding adjectives about the desired tone.
Why does generative AI concern marketing first?
Marketing is the top use of AI in European businesses. According to Eurostat's "Use of artificial intelligence in enterprises" (2025 data), 19.95% of EU businesses with 10 or more employees used at least one AI technology in 2025, and 34.70% of those businesses used it for marketing or sales, making it the most common purpose. For content published under a brand's name, prompt quality therefore directly shapes the image customers perceive.
Does the prompt replace human approval?
The prompt does not replace human approval: it reduces correction work, it does not remove the review step. Even with a precise instruction, a generative AI can produce an awkward phrasing, inaccurate information, or a visual that doesn't match the real product. A configurable validation workflow, where someone reviews and approves each piece of content before publishing, remains the guarantee that what gets published matches both the real offer and the brand's tone.
In brief, 3 questions
What is a prompt? A prompt is the written instruction given to a generative AI to get a text, an image or a video.
What length should you aim for? Google's 2024 Workspace guide states that the most effective prompts run about 21 words with context, versus under nine words in everyday use.
How do you avoid rewriting everything? By saving the brand's DNA and limits once in the tool, then only specifying the topic and format day to day.
What brand rule do you systematically add to your prompts so the AI writes "like you"?
To go further on formalizing brand rules, read our article Glossary: brand AI charter.
Sources
- Google Workspace, Google Workspace with Gemini: Prompting 101, September 2024.
- Direction générale des Entreprises, France Num 2025 Barometer: SMEs confirm their commitment, September 2025.
- Eurostat, Use of artificial intelligence in enterprises, 2025 data extracted December 2025.
