Design Feedback Email Prompt

What this prompt does

A design feedback email prompt turns your reaction to a design round into the email that gets the next round right — the decision stated up front, every requested change tied to what the design is supposed to do, and personal taste labeled as personal taste. Its input is your honest notes on the work; its output is a message the designer can act on without a follow-up call to decode it.

Most feedback emails fail in one of two directions. The vague ones ("love where this is going, just a few tweaks, can we make it feel more premium?") send the designer into another round of guessing. The prescriptive ones ("move the logo left, make the headline blue") turn a designer into a pair of hands and usually make the work worse, because the person writing them can see the problem but is guessing at the solution. The Nielsen Norman Group's guidance on design critique points at the fix: feedback works when it's anchored to the design's objectives rather than the reviewer's preferences. That's a discipline, and it's exactly the kind of discipline a prompt can enforce.

"Make it pop" has started more bad second rounds than any other three words in client work. The prompt below bans its whole family.

The prompt

Write a design feedback email from the inputs below. Follow every rule.

**What you're reviewing:** [THE ARTIFACT AND THE ROUND — "homepage mockup, round 2", "logo concepts, first pass"]
**The goal from the brief:** [WHAT THIS DESIGN IS SUPPOSED TO DO, in one sentence. If there's no written brief, write the sentence that should have been in one.]
**Who the designer is to you:** [AGENCY YOU'RE PAYING / FREELANCER / TEAMMATE / JUNIOR YOU MANAGE]
**What's working:** [THE SPECIFIC THINGS THAT MUST SURVIVE THE NEXT ROUND]
**What isn't working:** [EACH PROBLEM, AND THE GOAL IT GETS IN THE WAY OF]
**Pure taste:** [ANYTHING YOU JUST DON'T LIKE — be honest about which notes belong here]
**The decision:** [APPROVED WITH CHANGES / ANOTHER ROUND NEEDED / WE NEED A CALL]
**Deadline and rounds left:** [WHEN YOU NEED THE REVISION, AND HOW MANY ROUNDS REMAIN IN SCOPE]

Rules:
- Open with the decision and the overall read in two sentences. The designer shouldn't have to reach the last paragraph to learn whether this is an approval or another round.
- Name what's working before what isn't, and be as specific about it. "Keep the photography treatment" is an instruction that protects good work, not a courtesy.
- Every requested change states the problem and the goal it blocks, never the solution. "The booking button is easy to miss on a phone" — where it moves is the designer's call.
- Number the changes, in two clearly named lists: MUST CHANGE and CONSIDER.
- Taste is labeled as taste, capped at one note, and marked optional.
- No new goals. If a note doesn't trace back to the goal input, it becomes one line under "for a future round" or it gets cut.
- No compliment padding around the hard notes. Plain statements, neutral tone, no apology for having feedback.
- Close with the deadline, the number of rounds left, and where to ask questions.

How to use it

The "goal from the brief" line does the heavy lifting. Every note you write either traces back to it or gets exposed as taste, and both outcomes are useful — the first makes the note land, the second keeps it from masquerading as a requirement.

Sort your own notes honestly before you paste them in. The prompt can structure feedback; it can't tell whether "the palette feels dated" is blocking the goal or just isn't to your taste. You know. Put it in the right input and the email writes itself around that honesty.

The relationship line changes the register more than you'd expect. The same three must-changes read differently going to an agency on a paid round versus a junior on their second week, and the model adjusts the framing — but only if you tell it which one it's writing.

One thing this prompt won't do: soften a rejection into feedback. If most of your notes trace to the same misreading of the brief, you don't have a revision list, you have a conversation — see the variation below.

The three kinds of note

An email where every note is implicitly a must-change burns the designer's trust on the notes that don't deserve it. The labels are what make the real must-changes land at full weight.

Example output (partial)

Inputs said: homepage mockup, round 2, for a 12-person accounting firm; goal is getting visitors to book a consultation; the designer is a freelancer with one round left in scope; deadline Friday the 29th.

Round 2 is close — this is approved pending the three numbered changes below, no full round needed. The direction you took with the client photos is right and I want to protect it through these edits.

Keep: the photography treatment, the shortened nav, and the testimonial placement — all three survive as-is.

Must change:

1. On a phone screen the "Book a consultation" button doesn't appear until the third scroll. Booking is the page's one job, so it has to be visible without scrolling — where it goes is your call.
2. The headline reads as 26 words in the mock. Visitors deciding whether to stay won't read past the first line, so it needs to make its point faster.

Consider: 3. The gray body text over the office photo may be hard to read for some visitors — check the contrast and use your judgment.

Personally I'd try the serif from round 1 for the headline. Optional — ignore freely.

The model put the mobile problem first because the goal input said booking was the page's one job — the ordering came from the brief, not from the order the notes were typed in.

Variations

For an agency round you're paying for

Add the round count to the opening line, not just the close. "This is round 2 of 3" at the top changes how an agency reads every note below it, and it keeps scope creep visible to both sides. Money makes the numbered list format matter more, because an unnumbered note that gets missed costs a change order.

For a junior designer you manage

Add a rule: each must-change also names the principle behind it, in one sentence. "Booking is the page's one job" teaches; "move the button up" doesn't. This is the one context where explaining the why is worth the extra length, because the goal of the email is partly the next round and partly the next project.

For the round you want to reject

Count how many of your notes trace to the same misreading of the brief. Past two or three, stop writing feedback — the email becomes a short note naming the misread and proposing a call, because a numbered list of fifteen changes is a rejection wearing a revision's clothes, and it wastes a scoped round on patching the wrong direction.

For creative work that isn't design

The structure carries to copy drafts, video edits, and photography selects unchanged: decision first, keeps named, problems tied to the goal, taste labeled. Only the vocabulary of the notes changes.

Common pitfalls

• Don't prescribe solutions. "Make the logo bigger" answers a question nobody asked; the problem you actually saw — the brand isn't registering — might have five better fixes.
• Don't sandwich the hard note between two soft ones. In my experience the padding is what gets remembered, and the designer walks away unsure whether the middle sentence was a requirement.
• Don't deliver feedback in prose paragraphs. A designer revising against a paragraph has to reconstruct your list; some of it won't survive the reconstruction.
• Try instead: before sending, read only your must-change list and ask whether each item names a goal. Any note that doesn't is either a consider, a taste, or a cut.

Who uses this prompt

• Designers: design leads reviewing work asynchronously, and freelancers who quietly rewrite their clients' vague feedback into this format before acting on it. More in prompts for designers.
• Marketers: reviewing agency creative where the brief, the budget, and the round count all live in the same inbox thread. See prompts for marketers.
• Small business owners: hiring a designer for the first time and unsure how much direction is too much. See prompts for small business owners.

Both ChatGPT and Claude handle this well, and the reason is in Anthropic's prompt engineering guidance: explicit rules that ban the easy outputs — the compliment sandwich, the prescription, the unlabeled taste — force the model into the harder, more useful structure. The input people skip most is "what's working," and it's the one designers say they miss most: an empty keeps list reads as "nothing survived," and round 3 quietly loses the best thing round 2 had.