Brand voice and grounding

Where the voice comes from

An email can only sound like you if the model knows something about you. Running the generator inside a project versus running it cold is the difference between a writer who has read your whole brand binder and one working from a single sentence.

Pick a project in the header dropdown and leave Use this project's details on, which is the default. Every run then loads that project's brand, research, strategy, funnel, and active reference pack into the prompt before a word is written. That is what lets the email carry your voice and stay off your banned-phrase list, instead of reading like generic marketing copy.

Leave the header on Quick tools and there is nothing for the model to read, so it works from your brief alone. That is fine for a one-off or a fast test, but the result will be more generic. If voice matters, run inside a project.

What grounding does and does not do

Grounding means the email is written from your real brand context. It does not mean the model invents facts. The proof block in an email names something real from the grounding, and when the grounding offers nothing specific, that block gets shorter rather than padded with a made-up number or testimonial.

So when an email keeps landing vague, the fix is almost always richer brand and research context in the project, not a longer brief in the generator. For how to build that context, see the Brand Intelligence and Reference Packs chapters.