Quick answer
The fastest way to ask for feedback by email using AI is to paste your actual work into ChatGPT or Claude and ask it for two specific questions rather than a finished email. Narrow questions about one named section get real answers; wide ones get "looks good."
You sent the deck on Tuesday with “let me know what you think.” On Thursday you got two words back: “Looks good.” That is the whole problem with the way most of us ask for feedback by email — the question was too wide to answer, so you got the cheapest reply available.
The fix is not a politer email or a better subject line. It is a narrower question. AI is genuinely useful at this part — not at sounding warm, which you can already do, but at taking a messy situation and turning it into two questions a busy person can answer in ninety seconds.
✕ The ask that gets 'looks good'
Hi Priya — attached is the Q3 deck for Thursday's review. Let me know what you think!"
✓ The ask that gets an answer
Hi Priya — attached is the Q3 deck. Two things only: does slide 4 make the budget case clearly enough for Finance, and is there anything on slide 7 you'd cut? A line each is plenty — no need to read the rest.
Same person, same deck, same ten seconds of their attention. The second one comes back with something usable because it told the reader exactly where to look and gave them permission to stop reading.
Will this work for you?
Why “Any Feedback?” Gets You Nothing
Wide requests produce vague replies, and that is a documented pattern rather than ordinary politeness. In a Harvard Business Review write-up of four experiments, reviewers asked for “feedback” on a document sent back general praise. Reviewers asked for “advice” on the identical document sent back specific, critical suggestions. One word changed the quality of the reply.
That finding is the most useful thing you can carry into your inbox, because it means the problem was never the other person’s generosity. Three things make a request unanswerable, and most of us do all three at once:
- It has no edges. “What do you think?” covers everything from a typo on page two to whether the project should exist.
- It looks backward. “Was that okay?” asks for a verdict on finished work, which is uncomfortable to answer honestly.
- It has no ceiling. The reader cannot tell whether you want two minutes or two hours, so they give you two words.
There is a timing problem sitting underneath all of it. HBR’s guidance on getting the feedback you need makes the point that waiting for a scheduled review means hearing about a habit months after it mattered. Small, specific, frequent asks beat one annual one, and email is where most of those small asks live.
The Answerable Ask, in Three Parts
An Answerable Ask is a feedback request built so the reply takes two minutes and still tells you something you can act on: one named thing, a forward-looking question, and a capped effort. It is the approach we teach at PromptPeakAI for any email where you need a real answer rather than a courtesy.
It is not a template. Every time you ask for feedback by email the wording changes; the shape underneath it does not.
The three parts of an Answerable Ask, and what each one removes from the reader's way.
Pin the thing. Name one piece of the work: slide four, the pricing section, the way you handled the objection in the second half of the call. Never “the deck.”
Ask forward. “What would you change next time?” is far easier to answer honestly than “what did you think?”, because it is about the next version rather than a judgment on the person who made this one.
- Cap at two questions. A third one makes people defer the whole email to a quieter day that never arrives.
- Give the deadline a reason. “Before I send it Thursday” lands better than “whenever you get a chance.”
- Name the exit. “A line each is plenty” tells them how much you actually want.
How to Ask for Feedback by Email
To ask for feedback by email using AI, paste the real work and the real situation into ChatGPT or Claude, then ask it for two specific questions rather than a finished email. You write the greeting. The AI does the narrowing, which is the part you are bad at when you are close to your own work.
Give the AI the real context
The reason a first draft comes out generic is that the prompt was generic. “Write an email asking for feedback” gives the model nothing to narrow. Paste the thing itself — the draft, the deck outline, a summary of what happened — plus who you are asking and what you are privately worried about.
That last part matters more than people expect. Our guide to giving AI enough context covers the general habit; here the minimum is the work, the reader, and your actual worry.
What is safe to paste
Most feedback requests involve a real deliverable, which means real company information. Check it against this before pasting anything:
Your own drafts, public material, outlines with names stripped out.
Client names, internal figures, anything that identifies a colleague.
Unreleased financials, staff personal data, anything covered by an NDA.
When in doubt, describe the deliverable instead of pasting it. “A four-slide budget case written for a Finance audience” is enough for the model to write good questions.
Ask for questions, not an email
This is the step almost everyone skips. Ask for the email and you get a polite paragraph with “any thoughts welcome” buried at the bottom. Ask for the questions first, judge them yourself, then wrap the two best ones in your own greeting.
You are helping me write a short feedback request email.
The work: [paste the draft, outline, or a description of what you made]
The reader: [their role and how they relate to the work — my manager, the client who signed off, a peer who has done this before]
My real worry: [the thing you are actually unsure about]
The deadline: [when you need the reply, and why that date exists]
Do NOT write the email yet.
First, give me five candidate questions I could ask this reader about this work. For each one, tell me in a single line what kind of answer it would produce and whether a busy person could answer it in under two minutes.
Rules for the questions:
- Each question must point at ONE named part of the work, not the whole thing.
- Phrase them forward ("what would you change") rather than as a verdict ("was it good").
- No yes/no questions.
- Do not invent details about the work or the reader that I did not give you.Cut it down and send it
Pick the two questions that would actually change what you do next, then have the AI assemble the email around them. Two is the ceiling, not a target — one sharp question beats two mediocre ones.
Use question [N] and question [N] from your list. Drop the rest. Write the email around them: - Under 120 words total. - Plain, warm, no flattery, no "I hope this finds you well". - State the deadline and the reason for it in one clause. - End with an explicit exit line telling them a short reply is fine. - Match this tone: [paste two or three sentences of an email you actually sent recently] Give me the subject line separately. The subject should name the work and the ask, not just say "quick question".
You should now have an email under 120 words containing exactly two questions, a dated reason, and a line that lets the reader off the hook.
Read it once before sending and change at least one phrase to something you would actually say out loud. If the result still reads a little stiff, fixing email tone with AI is a separate ten-second pass.
Four steps, and the one most people skip is the second.
Draft the Same Ask in Claude
Claude runs the identical framework when you ask for feedback by email, with one practical difference worth knowing about. If you have connected the Google Workspace connector, Claude can put the finished request straight into your Gmail drafts folder. Anthropic’s support documentation is explicit that Claude creates the draft but cannot send it on your behalf — which is precisely what you want for an email like this one.
I need to ask [name, their role] for feedback on [the work]. Here is the work: [paste it, or describe it] Here is what I am unsure about: [your real worry] I need a reply by [date] because [reason]. Give me two questions that each point at ONE named part of this work, phrased forward rather than as a verdict. Show me the two questions on their own first. Once I confirm them, create a draft in my Gmail to [email address] with those two questions, under 120 words, ending with a line that makes clear a short reply is fine. Use a subject line that names the work.
Confirming the questions before the draft is created is the whole point of splitting it into two turns. You get to reject a question that misreads the situation before it is sitting in your drafts folder looking finished.
The Gmail Drafts folder open, showing a Claude-created feedback request draft with its subject line visible and the two questions in the body.
Crop to the drafts list and the open draft only. Hide the sidebar account name, the email address, and any other subject lines in the list.
File: claude-gmail-feedback-request-draft.png · Alt: "A Claude-created feedback request sitting in the Gmail drafts folder before sending"If you do not use Gmail, or have not connected anything, nothing is lost. Copy the text out of the chat window and paste it into whatever you send mail from — the framework is doing the work, not the integration.
Four Situations, Four Question Shapes
The framework stays fixed and the question changes, so the situation you ask for feedback by email about decides the shape. These four cover most of what lands in a working week, and each one fails in a different direction if you get the shape wrong.
| Situation | Ask this instead | Why this shape |
|---|---|---|
| Deliverable review | Which section would you rewrite before this goes out? | Forces one named section, not a verdict |
| Client project wrap-up | What would make the next project run more smoothly? | Forward-looking, so criticism feels safe |
| Before a performance review | How did I handle the Latham escalation, and what would you change? | Names one event, avoids a career-wide judgment |
| After a presentation | Did the middle section land, or did I lose the room there? | Offers two answers, so silence is not the default |
Each situation fails in its own direction — the question shape is what prevents it.
The review case is the one worth practising, because it is the highest stakes and the easiest to fumble. Most managers will tell you the same thing: they would give better feedback if anyone asked them about something specific enough to remember. Raise it in your next 1:1 rather than in a cold email, and use the written ask only for the part you want them to think about first.
I have a performance review with [manager's name and role] on [date]. I want to ask them for input beforehand, by email. What I want to get better at: [one skill or behaviour, not three] A specific recent moment involving it: [name the project, meeting, or escalation] Write me two questions about that ONE moment, phrased so they are easy to answer honestly and do not put my manager in the position of delivering a verdict on my whole year. Then write the email around them, under 120 words. Do not include praise-fishing questions. Do not ask "how am I doing".
Asking an interviewer works the same way, though the window is narrower — see writing the thank-you email after an interview for where that ask naturally fits.
When the Request Falls Flat
Sometimes you do everything right and still get nothing. Each failure has a different cause, and only one of them is about the other person.
| What happened | Why | The fix |
|---|---|---|
| Still got "looks good" | Your question still had no edges | Name a section and ask what you would change |
| No reply after a week | Three questions, or no deadline reason | Resend with one question and a dated reason |
| They answered the wrong part | The email buried the ask at the bottom | Put both questions in the first three lines |
| Reply was harsh and unusable | You asked at the wrong moment | Wait a day, then ask about the next version |
✕ The rescue email that fails
Thanks! Any other thoughts on the deck? Really appreciate any feedback you can give — no rush at all."
✓ The rescue email that works
Thanks — that helps. One follow-up: on slide 7, would you cut the vendor comparison or keep it? I'm sending this Thursday morning, so even a one-word answer is useful."
The silence case is the common one, and it is usually a scheduling problem dressed up as a rudeness problem. A single follow-up with one question works far more often than people expect — writing a follow-up email with ChatGPT covers the timing, and chasing a stalled approval uses the same pattern when the thing you need is a decision rather than an opinion.
I asked [name] for feedback on [the work] and got this back: [paste their reply] This is too general to act on. Write me a two-sentence follow-up that: - Thanks them without being effusive - Asks ONE specific question about the part of the work I still need judged: [name it] - Gives them a one-line way to answer Do not ask them to elaborate on what they already said. Ask something new and narrower.
Three ways a feedback request fails, and where each one actually went wrong.
What AI Cannot Do Here
The narrowing is the part AI is good at. Everything else that makes it work when you ask for feedback by email sits outside the model’s view, and pretending otherwise is how people end up sending something technically well-written and socially wrong.
Four things stay yours, and no prompt gets you out of them:
- Choosing who to ask. The model does not know that your manager has been underwater since the reorg, or that the colleague who would give you the most honest answer is the one you have never asked.
- Reading the history. It cannot see that you asked this person for feedback twice last month, which is the actual reason they stopped replying.
- Judging the timing. Asking for advice on a presentation two hours after it went badly gets you a defensive reply from both sides.
- Deciding what to keep. Feedback is data, not instructions. Some of what comes back will be wrong, and sorting that is judgment, not processing.
There is also a limit worth naming plainly: a written request cannot do the work of a conversation. If what you actually need is to understand why someone is unhappy with your work, preparing for that conversation is the right tool, and writing genuinely difficult emails is a different skill with different rules.
Turn the Answers Into Next Steps
Most people treat a good reply as the finish line. It is the halfway point, and the second half is where the relationship is either built or quietly spent.
- Reply within a day, briefly. Name the one thing you are changing because of what they said.
- Close the loop later. “Cut the vendor slide like you suggested — Finance approved it on Thursday” costs you ten seconds and buys you every future reply.
- Do not defend the work. The moment you explain why they misread it, you have taught them not to bother next time.
When feedback arrives from several people at once, the sorting becomes its own job — summarising multi-source feedback with AI picks up exactly where this leaves off.
Start with the next email you were about to end with “let me know what you think.” Replace that line with one named section and one forward-looking question, and see what comes back. That single substitution does most of the work of learning to ask for feedback by email properly; the prompts above just make it faster to do well every time.
Rewrite your next feedback request
- Pick one named section of the work instead of the whole deliverable. Paste that work, the reader's role, and your real worry into ChatGPT or Claude. Ask for five candidate questions, not a finished email. Keep the two that would change what you do next, and cut the rest. Add a deadline with a reason and one line letting them reply briefly.