The example is made up. Northgate Creative, a made-up five-person agency, needs a brief for its made-up client Tidewell Coffee Roasters, which is launching a coffee subscription. Every number in the lesson is made up.
What are the most common ChatGPT creative brief mistakes?
| # | Mistake | Fix |
|---|---|---|
| 1 | A one-line prompt | Paste the interview prompt instead |
| 2 | Trusting a brief that looks finished | Look for invented audiences, channels and goals |
| 3 | A goal with no number or date | Give a target and a deadline |
| 4 | Leaving out the audience's doubt | Say what they think today and the answer |
| 5 | Only saying which channels are in | Say what is out too |
| 6 | Forgetting must-haves and banned claims | List them, with who signs off |
| 7 | Skipping the summary | Check every line before "go ahead" |
| 8 | Not saving the prompt | Save it as a template with two blanks |
Why is the one-line prompt the biggest mistake?
1. A one-line prompt. The lesson's quick prompt is Write a creative brief for a client's spring campaign. It sounds reasonable, but ChatGPT has to fill every gap itself.
2. Trusting a brief that looks finished. The result has headings, bullets and a key message. It also has blanks like [Client Name] and [Budget], an audience of 18 to 34 year olds, TikTok, Instagram, Facebook, LinkedIn, Google Ads and influencers, and goals like "increase brand awareness by 50%" and "drive a 25% increase in sales". Blanks are easy to spot and fill. The invented details look plausible, so they slip through and the creative team builds on them. Hand that to designers and you pay for it in revisions.
The fix is the interview prompt: ask up to 5 questions, one at a time, wait for each answer, make no assumptions, sum up and wait for an OK, then write. In the lesson you click New chat in the sidebar and paste it there.
Which answers make the brief weak?
3. A goal with no number or date. ChatGPT's first question asks who the client is, what they sell and what the campaign should achieve. "More subscriptions" is hard to judge. The lesson answers with 300 new subscriptions between March 2 and April 26. A real number and a date make the whole brief easier to judge, and it becomes the line in How we'll measure it.
4. Leaving out the audience's doubt. Question 2 asks who you are trying to reach and what they think today. The lesson's answer: regulars aged 30 to 55 who like the coffee but think subscriptions lock you in and cost more. Then the answer to that doubt: pause or skip any month, free local delivery. That doubt is exactly what the first brief invented around, and it's what the creative has to answer.
Try it free, step by step
Write a Creative Brief Your Client Signs Off On: Make ChatGPT Ask You Questions First. Ask ChatGPT for a creative brief the quick way and watch it invent the audience, budget and goals. Then use the interview prompt: it asks for the client’s real goal, audience, budget and must-haves… You do every step yourself in a practice copy of ChatGPT, it checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, and nothing touches your real accounts.
Start the free lesson5. Only saying which channels are in. Question 3 asks for budget, channels and the due date. The lesson's answer gives $6,000 for paid social on Instagram and Facebook, an email to 4,200 loyalty members and table cards in all 3 cafés, then says: no TikTok, no influencers. Saying what is out matters as much as what is in. The brief carries it as Not in scope.
6. Forgetting must-haves and banned claims. Question 4 asks how it should sound, what every piece must include or avoid, and who signs off. The lesson's answer: warm and a little nerdy about coffee, never pushy; every piece shows the logo and says "Pause or skip any time."; the only offer is the first bag free; never say "best coffee in Richmond" because they can't prove it; Lena Ortiz signs off by February 6. Claims you can't prove belong in the brief as a rule, not in a revision note.
What goes wrong at the end?
7. Skipping the summary. After four questions, ChatGPT sums up what it understood and waits. This is your chance to catch a mistake before it is in the brief. Check every line, and say so if something is wrong. Only then send "Looks right. Go ahead." This is the same habit as asking a model to ask clarifying questions rather than assume.
Then read the brief before it goes anywhere. Click the Copy icon under it, have the account lead read it, and send it to the client for sign-off before the creative team starts.
8. Not saving the prompt. The prompt is the part you'll reuse. In the lesson you click the Google Docs tab, open the Prompt Templates doc, add the heading Interview prompt and paste the template with [WHAT YOU NEED] and [MY BUSINESS] as blanks. Use it for a campaign recap, a social calendar, a case study or a proposal. A shared prompt library like this saves the whole team from writing it again.
Key takeaways
- One line in, guesses out. Paste the interview prompt.
- Give a number and a date for the goal.
- Name the audience's doubt and the answer to it.
- Say which channels are out, and which claims are banned.
- Check the summary, have a person read the brief, then save the prompt.
The ChatGPT creative brief FAQ answers the how-to questions, and ChatGPT interview prompt mistakes covers the same prompt for other jobs. For brand voice that sticks across chats, try the ChatGPT Projects brand voice lesson.
Frequently asked questions
- What if real ChatGPT asks different questions?
That's fine. It may word them differently, ask them in another order or ask a different number. Answer what it asks, and check the summary before you say go ahead.
- Should I fill the blanks in the first brief instead?
No. The blanks are the easy part. The invented audience, channels and goals are the problem, and filling blanks doesn't remove them.
- Can I include client sales numbers?
Follow your agency's policy. Keep anything said in confidence, like unreleased prices or sales numbers, out of chats unless your policy allows it.
- Does memory affect the quick prompt?
If ChatGPT's memory is on, it may fill in some details from your past chats. It still guesses at the rest.
