If you search for AI prompts for marketing online, you'll find a lot of lists, generated by...marketers, looking to attract visitors! But most of the prompts are rudimentary and generate generic garbage. Use the tips and frameworks, from Trust Insights AI, to generate better quality results in less time.
Since AI models are trained on large datasets to learn from examples and find patterns, short, general queries are going to give you generic, most-common-denominator copy! You won't get something particularly creative, unusual, or unique with simple prompts. The copy is almost always going to sound like an average marketer, unless you get specific and targeted.
Starting with short prompts (like these 230 ChatGPT Prompts Marketers Should Use), then going back and forth a dozen time to narrow down your task and hone results is going to be time consuming.
If you intend to get facts, data, analysis, or strategies from your AI model, the likelihood of hallucinations is high if you don't steer it to established knowledge in your domain.
A few months ago as part of the MarketingProfs AI for Digital Marketers series, Christopher S. Penn from Trust Insights AI, presented their free R.A.C.E. framework for AI prompts.
It works great for me, and helps me remember what to include so even 'Round 1' results are useful.
You are an expert in ____________________. You know about _______________ and use ______ framework/model, etc.
Your first task is ________________.
References to use are ________________. Content to analyze is here: _____________________. Writing style to use: ___________. Target Audience: _______________.
Generate _______________________ and provide results in a [bullet point] format using a maximum of ____ words.
This is a part two from Trust Advisors AI, which you'll find here on their site. PARE stands for:
I love this question and often start with it for research and analysis. Once you've reviewed the answer, keep querying, until you're satisfied that the AI chatbot has a good foundational understanding of your domain, without mistakes. Then continue in the same thread to build your query, having primed the model with key context.
Once you're ready for R.A.C.E., you won't need to provide so much context or hope that your context was detailed enough. You'll already have primed it.
Let's say you've started with a detailed R.A.C.E. prompt and need a pretty complex response. This question should help uncover if anything is unclear or missing.
I haven't used this one yet!
Have you ever received an AI model response that didn't fully align with your instructions, after a long back and forth?This questions will help the AI model go back and review, then adjust the answer to match your specifications.
One last tip for prompting - please try Perplexity AI! It's great. 😍
I've switched my paid ChatGPT-4 subscription over to Perplexity Pro. And heading over to Poe.com to take a look...