Hey everyone! I’ve been trying to get my company’s brand voice really dialed in on Copy.ai, and I think I’m finally getting somewhere. Our marketing lead gave me a bunch of our past high-performing emails and ad copy and asked if I could “train” the AI on it. Sounded like a data pipeline problem to me! 😅
I figured out a process that seems to work okay, but I wanted to run it by you all to see if I'm missing any best practices. Basically, I gathered all the winning copy into a single text file, making sure to include a variety of formats (subject lines, body text, CTAs). Then I went into the Brand Voice section and used the “Custom” option. Instead of just pasting a generic brand guideline, I fed it several of our top-performing pieces as examples.
My main question is about *volume* and *structure*. How much past copy is ideal? Is there a point of diminishing returns? Also, should I be labeling the examples in the input somehow, like “This is a successful LinkedIn ad” or is the AI smart enough to figure out the context on its own? I noticed the generated output got a lot closer to our tone after about 10 solid examples.
Also, has anyone tried a more programmatic approach? I was thinking of using Python to scrape and clean all our past copy into a consistent format before uploading. Maybe even using some basic NLP to find common phrases first? Or is that overkill for this tool? I’d love to hear how others have set this up.
-- rookie
rookie