Our digital marketing team is trying to scale content production. We’ve noticed that generic prompts for Generative AI (ChatGPT, Gemini) result in very repetitive "AI-sounding" copy. What are your best tips for creating "Chain-of-Thought" prompts that maintain a specific brand voice and avoid common clichés?
3 answers
To get high-quality output from Generative AI (ChatGPT, Gemini), you have to provide "Few-Shot" examples. Feed the model three paragraphs of your best existing content and tell it to analyze the tone, sentence structure, and vocabulary before asking it to write something new. Also, use a "Persona" prompt—tell it it’s a veteran copywriter with a focus on psychological triggers. This forces the model to move past the standard "In the fast-paced world of..." clichés that plague lower-effort AI generations.
Have you tried using a "critic" agent where one AI generates the copy and a second prompt tells it to find flaws in that copy?
Consistency is the hardest part. We maintain a "Prompt Library" for different channels like LinkedIn vs. Instagram to keep the output distinct.
A shared prompt library is a great idea. It ensures the whole team is hitting the same quality benchmarks regardless of their individual AI experience.
That "Agentic" workflow is a game changer, Jeffrey. Having the AI review itself based on a checklist of brand guidelines significantly cuts down on the manual editing time for our editors.