When the job description looks AI-generated: how to respond with specifics

More job posts read like they were drafted in thirty seconds by a model: perfect headings, hollow requirements, no numbers, no names, no mess.

That does not always mean the client is fake. Often it means they were busy, copied a template, or let AI “write the post” without adding facts. Your proposal still has one job: prove you can deliver this outcome, not that you can spot ChatGPT from orbit.

Do not mock the post. Do not list “AI tells” like a detective thread. Win with specifics.

Related reads: short job post proposals, clarifying questions before you spend a connect, and why “I used ChatGPT” hurts in your own proposal.

Signs the post might be AI-shaped (use privately, not in the proposal)

Common patterns:

• Long lists of skills with no priority.
• Buzzwords without metrics (“cutting-edge,” “world-class”).
• Perfect grammar, zero constraints (budget, deadline, tools).
• Same post structure you saw on ten other jobs today.
• Deliverables that could mean anything (“full social media management”).

Treat these as low information, not as proof of bad intent.

Your strategy in three moves

1. Extract one concrete hook (tool, industry, audience, deadline, platform).
2. State a reasonable assumption if data is missing.
3. Offer milestone 1 small enough to quote without guessing the universe.

If you cannot do step 1 even after reading their profile or site, consider skipping (red flags vs bid tighter).

Move 1: steal a hook from their world

Even a fluffy post usually hides one clue:

• A brand name or URL.
• A platform (Shopify, WordPress, Upwork itself).
• A role (fractional CMO, VA, React dev).
• A number buried in noise (5 pages, 3 videos, 10k emails).

Proposal line example:

You mentioned [exact phrase from post]. I would start by [first step] and deliver [narrow artifact] before we expand to [bigger word from post].

That sentence cannot be sent to every client unchanged. That is the point.

Move 2: assumptions beat interrogation

AI posts invite twenty questions. Asking twenty makes you look like free consulting.

Pick two questions max in public, or zero in the proposal if you can assume safely.

Inside the proposal:

I am assuming [assumption]. If [alternative] is true instead, timeline shifts by [X]. Milestone 1 below is based on [assumption].

Buyers respect stated assumptions. They ignore vague “let me know your thoughts.”

Move 3: milestone 1 is your truth serum

If the client is real, they react to a clear first step. If they are fishing, they argue about everything for free.

Milestone 1 ([price range]): [3-5 bullets of tangible outputs], [revision limit], [timeline], pending [access or asset].

Link to fixed-price proposal pricing and scope creep / out of scope.

Opening templates (paragraphs, not code blocks)

Template: buzzword soup, e-commerce hint

You want to improve online sales and brand presence. Without a fixed SKU count I will assume a single Shopify store under 200 products. Milestone 1 would be a 90-minute audit plus a prioritized fix list (checkout, mobile speed, core collections), not a full rebrand.

Template: “marketing guru” post, no channels named

The post focuses on growth but does not name channels. I will assume organic content plus one paid channel you already use. Milestone 1: content calendar for 4 weeks plus 2 ad creative sets, with reporting you can read in ten minutes.

Template: dev post listing twelve frameworks

You listed several stacks. For milestone 1 I propose we lock [one stack that matches your profile] and ship [one feature slice] on staging. That de-risks architecture debates before a larger build.

Swap details. Keep the structure: assume, bound, offer milestone 1.

When to ask questions instead of assuming

Ask (briefly) when wrong assumptions would waste days:

• Access you cannot guess (production DB, ad spend, legal review).
• Regulated work (health, finance) where compliance changes scope.
• Fixed deadlines under 72 hours (48-hour deadline jobs).

Use the question list in clarifying questions before you spend a connect.

What your proposal should avoid

• Calling the post “AI slop” or “obviously ChatGPT.”
• Quoting generic advice back at them with zero new detail.
• Competing on “I am human” vibes without proof.
• Saying you used AI to write their deliverables unless they asked and you frame it honestly (mention AI tools).

Your edge is judgment and specificity, not moral superiority about how they wrote the post.

If the post is duplicated across platforms

Same text on Upwork, Fiverr, and a Google Form usually means spray hiring.

Tighten your proposal, raise your price floor, or skip. See reuse proposals without copy-paste tells for your side of the arms race.

After you send: read their reply quality

Good sign: They correct an assumption with facts.

Bad sign: They want full scope at sample-task prices. Route to unpaid test task scripts.

Checklist

• I cited at least one phrase or fact from their post or profile.
• I stated assumptions where data was missing.
• Milestone 1 is small, priced, and measurable.
• I asked at most two questions, or none with clear assumptions.
• I did not insult how the post was written.
• I ran the proposal checklist before send.

AI-shaped posts are the new normal. Freelancers who sound specific still win, because buyers hire people who reduce confusion, not people who win grammar contests.