Untargeted outreach does more than get ignored, it damages the sender. AI Central's 11 ChatGPT Prompts To Target the Right Prospects answers that with one move: stop asking ChatGPT to write better messages and start asking it to define who should receive them. Eleven prompts, each built on the same two variables, your industry and your product, walk an ideal customer profile from definition through decision makers, buying behavior, messaging, and the signals that a company is turning into a buyer.
Reviewed August 2026.
Targeting is a reputation problem, not a volume problem
Most outbound advice treats a weak reply rate as a copy problem. Rewrite the subject line, shorten the opener, add a case study. AI Central starts somewhere less comfortable, with the argument that sending to the wrong people carries a cost that outlives the campaign.
The opening line of 11 ChatGPT Prompts To Target the Right Prospects puts it plainly:
Not only are untargeted messages annoying, but they also hurt the sender's reputation.
That reframes the job. If a bad list is a liability rather than merely a waste, the highest-leverage use of ChatGPT in sales is not drafting, it is deciding who gets drafted to. Every prompt in the set sits upstream of the first sentence you write.
Eleven prompts, five jobs
The set is numbered one through eleven, but it is not a strict sequence. Read across it and the prompts cluster into five distinct jobs, and knowing which job you are in is what stops the outputs blurring into one long generic profile.
- Define it. Prompt one builds an ideal customer profile from scratch for your industry and product. Prompt six narrows a broad industry into niche segments and specific company characteristics.
- Find the humans. Prompt two asks who the decision-makers, influencers and champions are. Prompt ten asks what motivates those personas and what actually drives them.
- Understand how they buy. Prompt three covers purchasing behaviors, decision-making hierarchies and vendor evaluation processes. Prompt seven asks for the early signals that a company is becoming a candidate.
- Shape the message. Prompt four is the only one aimed at outreach copy, and it deliberately comes after the profile exists rather than before.
- Keep it alive. Prompts five, eight, nine and eleven handle feedback loops, expansion inside accounts that already fit, how the profile shifts over five years, and the external economic, political and technological forces acting on it.
Two variables do all the heavy lifting
Every one of the eleven prompts is built on the same two placeholders, an industry name and a product or service description. Nothing else varies. That is the elegance of the set, and it is also its single failure point.
Type in SaaS and our software, and you will get a profile that reads like a textbook chapter, because that is the only thing a model can build from an input with no edges. Type in mid-market logistics companies running their own fleets, and a route optimization tool that plugs into existing telematics, and the same prompt returns something you can argue with. The prompts do not fail quietly on vague inputs, they fail generically, which is harder to notice.
The wording is also demanding on purpose. The prompts ask for "advanced characteristics", for a "nuanced" profile, for the "complex" purchasing behaviors rather than the obvious ones. Those adjectives are load-bearing. They push the model past the first and blandest layer of its answer, and outputs get noticeably thinner when they are stripped out.
Why splitting the question beats asking it once
The obvious shortcut is to ask ChatGPT to define your ideal customer profile in one go. It will happily do it, and the answer will be a template. Firmographics, two or three pain points, a list of job titles.
The eleven-prompt approach works because each question is narrow enough that you can check the answer against reality. You know whether the decision makers it named exist at the accounts you sell to. You know whether the buying process it described matches the last three deals you closed. A single monolithic profile gives you nothing to test. An eleven-part one gives you eleven places to catch the model being wrong.
Prompt seven shows what that buys you, because it asks for something you can go and monitor:
What predictive indicators or early signals might suggest that a company within [industry_name] is becoming a strong candidate for [product/service_description]?
An answer to that is not a description of a customer, it is a list of triggers. Hiring patterns, funding events, tooling changes, whatever the model surfaces for your market. You take that list and you go and watch for it, which is the difference between a profile you file and a profile you run.
The four prompts most people will skip
Four of the eleven point at the future, and they are the ones that get ignored. Feedback loops, expansion, evolution, external factors. None of them help you send an email this afternoon.
They are also why the set is worth more than an afternoon. Prompt nine asks the question almost nobody puts to a model:
How might the Ideal Customer Profile for a product like [product/service_description] evolve over a span of 5 years, especially considering the dynamics of [industry_name]?
AI Central labels the expected output there as projections and speculations on the changing nature of the profile, and the honesty of that label matters. You are not being handed a forecast. You are being handed hypotheses about which parts of your profile are load-bearing and which are temporary. Prompt five then closes the loop by asking for the feedback mechanisms that would tell you when the profile has drifted, and that is the step that turns an ICP from a document into a process.
What to do with this
Run the prompts you need in one conversation rather than eleven separate ones. They share context, so each answer builds on the last instead of starting cold.
- If you have never written an ideal customer profile, run prompts one, six and two in that order and stop. Definition, niche, decision makers. That is a usable first version and it takes about twenty minutes.
- If you already have a profile, skip prompt one entirely and start at three and seven. Buying behavior and early signals are where an existing profile is almost always thin, because they are the parts nobody writes down.
- If you run a team, prompts two, three and four are the ones to standardize. Everyone selling into the same market should work from the same map of who decides, how they buy, and what language lands.
- If you sell to an installed base, prompt eight is the highest-return question here, because expansion revenue needs no new profile, only a sharper read on the one you have.
Then treat every output as a draft. AI Central's own closing instruction is a loop rather than a one-off: drop a prompt, polish the output, repeat. The model gives you a structured starting position on a market. Your last three closed deals tell you where it is wrong, and that correction is the part no prompt can do for you.
Do I need to run all eleven prompts?
No. They cover different jobs and most people need three or four of them. Definition, decision makers and buying behavior get you a working profile. The future-facing ones are worth running once a quarter rather than once per campaign.
What exactly do I put in the brackets?
Two things, and both should be narrower than feels comfortable. The industry field wants a segment rather than a sector, so mid-market dental practices rather than healthcare. The product field wants the mechanism and the buyer's outcome in one line, not your tagline. The prompts are only as specific as what you hand them.
Can ChatGPT really know my industry well enough for this?
It knows the published shape of your industry, which is a genuinely useful starting point and not the same thing as knowing your market. Treat every output as a hypothesis list to check against your own closed-won data. Where the model and your deal history disagree, your deal history wins.
How is this different from just asking for an ICP?
Asking once returns a template you cannot argue with. Splitting the question into eleven narrower ones gives you eleven separate claims, each small enough to verify or reject on its own. The value sits in the checkability, not the word count.
How often should I redo this?
Prompt five exists precisely because the answer is not never. Set up the feedback mechanism it produces, then revisit the profile when that mechanism trips, or on a fixed schedule if you would rather not think about it. Prompt eleven is the one to rerun whenever something in your market visibly shifts.