I briefed AI the same wrong way for a month
I briefed AI the same wrong way for a month. Three things I changed after a client sent the whole lot back.
Every piece landed back with the same comment. It does not sound like us. I kept rewriting at the sentence level and it kept not working. The problem was deeper than the wording.
1. I was describing tone in adjectives.
Friendly but professional. Warm but authoritative. These mean nothing concrete. Everybody pictures something different and so does the model. Now I paste in two pages the client has already published and say match this. The difference was immediate and honestly a little humbling.
2. I was briefing the piece and not the reader.
I would spell out what the article was about and never say who it was for. The reader changes length, vocabulary, how much background I give and what can be left out. The audience has more influence on the draft than the subject does.
3. I was treating the first draft as the blueprint.
The model returned a structure and I edited inside it, so every piece kept the shape the model chose. Now I ask for an outline first, fix that, and only then ask for prose. Reshaping an outline takes two minutes. Shifting a finished draft means rewriting the entire thing.
The notes stopped after that. None of it was fancier sentences. It was deciding things before the draft existed instead of arguing with a draft that already had a mind of its own.
The copy still arrives sounding like AI. That is not a briefing problem and there is only so much context can fix.
#clientwork #creativebrief #copywriting #freelance #humantone
Briefing AI models effectively can be a real game-changer when it comes to producing content that feels authentic and fits the brand’s identity. From my experience, the biggest breakthrough wasn’t in tweaking individual sentences but in refining the entire briefing process upfront. One of the most important lessons I learned relates to how tone is communicated. Instead of using vague adjectives like "friendly" or "professional," I started providing actual examples by pasting two full pages of the client’s existing writing. This creates a direct style reference for the AI, significantly reducing interpretation errors. It’s amazing how much difference this concrete context makes. Another overlooked aspect that dramatically improved results was shifting the focus from the article's topic to the intended audience. Understanding who will read the content changes whether the language should be simple or technical, detailed or concise. Before this, I’d only describe what the piece was about, ignoring the reader’s personality or knowledge level. Once I clarified the target audience in the brief, the drafts became more relevant and engaging. Additionally, treating the first AI-generated draft as a final blueprint limited flexibility. Instead, I adopted a two-step approach: first, request an outline and review it thoroughly; then ask for the full draft. Fixing an outline is quicker and allows guiding the AI better from a structural perspective before it commits to detailed text. This approach saved hours of rewriting and improved the flow of the content. Despite these improvements, some AI-generated content can still sound artificial. This is less about briefing and more about the inherent limitations of AI language models today. Supplementing AI drafts with human editing or tools like humantone.io can add that final human touch. Ultimately, effective briefing combined with thoughtful editing ensures content genuinely represents the brand voice and connects with readers.
