TL;DR
On this page
You are applying to 15 jobs a day. You paste each posting into ChatGPT, type "write me a proposal for this," copy the result, and send it. The drafts read fine. Professional, even. And you are getting nothing back.
This walkthrough is part of the complete guide to generating client documents with AI.
One freelancer documented exactly this and put the result bluntly:
Crickets. No views. No invites. No responses. Just pure silence.
Source: a freelancer's account, Freelancers Hub (Medium)
They had sent 15 to 20 proposals a day and hit 30-plus submissions with zero replies before realizing what was wrong. The problem was not that they used ChatGPT. It was that they used it the lazy way, the same way every other applicant does, so every client received the same polished, generic draft and skipped all of them.
The model is not the problem. The input is. This post gives you a prompt that fixes the input, then shows you the three things that make an AI draft win instead of lose.
The prompt that writes a proposal clients answer
Paste this into ChatGPT, Claude, or Gemini. The job post goes in verbatim, which is the part most people skip.
You are an expert freelance proposal writer. Write a proposal for the
job below.
JOB POST:
[paste the full client job post here, word for word]
ABOUT ME:
- Role: [e.g. web developer]
- Most relevant proof: [1-2 specific results, with a number]
- Rate or package: [your price or range]
Rules for the proposal:
1. Open with one sentence that proves I read THIS posting: reference a
specific detail, problem, or goal the client named. Never open with
"I am excited to apply."
2. Restate the client's problem in my own words, then state the OUTCOME
I will deliver, not just the tasks I will do.
3. List 3-4 scoped deliverables as bullets, so the work is concrete and
bounded.
4. Add one sentence of proof with a number.
5. End with one clear next step (a question or a short call offer).
6. Under 150 words. Plain English. No buzzwords, no "I am passionate,"
no filler.
Then give me two alternative opening lines I can swap in.
The difference between this and "write me a proposal" is the job post and the rules. Without them, the model has nothing specific to work with, so it produces the fluent nothing that clients have trained themselves to ignore. With them, the draft is forced to be about this client.
Why generic AI proposals lose
A proposal does not lose because a client can tell a machine wrote it. It loses because it says nothing only you could have said. The opening could be pasted onto any posting, the body lists capabilities instead of answering a problem, and the whole thing reads as effort-free. Sameness is the signal clients react to.
What wins is the opposite, and the data backs the specifics. The average proposal close rate across industries sits at about 36% per Proposify's analysis, and the separation between winners and losers is not talent. As the Plutio proposal guide puts it:
The gap between winning proposals and losing ones usually comes down to structure, specificity, and speed.
Source: Plutio, "How to Write a Winning Proposal"
Specificity shows up even in formatting. Proposify's own data found that proposals containing images were 72% more likely to close, which is really a proxy for effort: the freelancer who added a relevant visual is the one who tailored the rest too. You do not need to bolt images onto an Upwork message, but the lesson transfers directly. The proposal that is visibly built for this client beats the one blasted to a hundred.
The three inputs that turn a draft into a winner
The prompt encodes three things. They are worth understanding on their own, because they are what you check before you send.
1. Signal extraction. Pasting the full job post is what lets the draft reference a specific detail: the deadline they mentioned, the tool they are migrating off, the frustration in the third paragraph. That single referenced detail is what produces the "they actually read my posting" reaction. The Upwork-specific tactics go deeper on which signals to pull.
2. Outcome over tasks. A generic proposal says "I will build your website." A winning one says "you will have a site that loads in under two seconds and lets you add products without a developer." Same work, different sale. The first describes your time; the second describes the client's result. Force the model to state the outcome, because left alone it defaults to listing tasks.
3. Scoped deliverables. Three or four bounded bullets make the work concrete, and they do something the model will not tell you about: they become the skeleton of your contract. Vague scope is not a small risk. Of the freelancers who get stiffed by a client, 37% blame vague or poorly written contracts, and that vagueness almost always starts in the proposal. A proposal that scopes the work cleanly protects you twice. The scope-of-work structure shows what a tight deliverables list looks like.
Here is what those three inputs change, side by side:
| Element | Generic AI draft | After the prompt |
|---|---|---|
| Opening | "I am excited to apply for this opportunity." | One line referencing a specific detail from the job post |
| Body | A list of your skills and services | The client's problem restated, then the outcome you will deliver |
| Deliverables | Vague ("full website build") | 3-4 scoped bullets that bound the work |
| Proof | "I have years of experience" | One result with a number |
| Close | "Looking forward to hearing from you." | One clear next step: a question or a call offer |
Read it before you send
The prompt gets you a strong draft, not a finished send. Read it once and check three things: the opening references something real from the posting (not a detail the model invented), the deliverables match what you can actually do at your quoted price, and the proof number is true. AI will occasionally invent a specific to sound confident. Your name is on the proposal, so the facts have to be yours.
pro tip
Save the prompt with your "ABOUT ME" block already filled in. The job post is the only part that changes per application, so a reusable prompt with your role, proof, and rate baked in turns each proposal into a single paste. That is also how you keep the speed advantage without dropping back to generic output.
Or skip the prompt and keep the structure
The prompt works, and if you send a few proposals a week it is a clean way to do it. The friction is the same as every AI-document workflow: you tune the prompt, fix the opening, paste it somewhere, and reformat, every single time. And the winning structure (a specific opening, an outcome, scoped deliverables, a number, one next step) is the same on every proposal you will ever send.
If you would rather start from that structure instead of rebuilding it each time, FreelanceDesk gives you a proposal with the scoped-deliverables and outcome sections already in place. You fill in the project and export a clean PDF, and it is free. Once the proposal wins, the same details flow into the contract without retyping.