AI prospecting messages: what it does well and what gives it away

Generative AI can write a prospecting message in a few seconds. For a small business owner or a B2B salesperson, the promise is tempting: more messages, faster, without spending your evenings on them. But prospects now receive a lot of messages written by AI, and they recognize them. So the question is no longer whether AI can write your messages. It is which steps it genuinely helps with, and where a human is still essential.

What can AI do when writing prospecting messages?

AI does three things well when you write prospecting messages: it summarizes, it rephrases and it adapts. From a LinkedIn post, a job ad or a profile, it can sum up a person’s situation in one sentence. From a first draft, it can suggest several ways of saying the same thing. And it can adapt one message to the tone of an industry or to another language.

⚠️ It does poorly at anything that requires a choice. On its own, it cannot tell whether this prospect deserves a message today, or which fact about their situation matters most for your offer. Left without instructions, it produces smooth, polite, interchangeable messages that look like every other message in the inbox.

Why are AI-written messages so easy to spot?

AI-written messages give themselves away through a handful of recurring marks. They are not mistakes. Quite the opposite: the sentences are too clean, too long and too general to have been written by someone who actually knows the recipient.

The most common signs:

1. Generic compliments: “your impressive background”, “your inspiring vision”.
2. Stock opening lines: “I hope this message finds you well”, “In today’s fast-paced world”.
3. Lists of three benefits, always with the same rhythm.
4. Long dashes and formal turns of phrase that nobody uses in a quick note, such as “I wanted to reach out” or “delve into”.
5. Excessive length: eight lines where three would have done.

The recipient does not run this analysis consciously. They simply sense the message was not written with them in mind, and they move on to the next one. A practical test: if you swap the prospect’s name for someone else’s and the message still works, it was not personalized, whoever wrote it. We explain this mechanism in why your prospects don’t reply.

How do you use AI well to write your messages?

Using AI well means giving it what it cannot make up, and keeping the decision for yourself. A five-step method works well:

1. 🎯 Pick the prospect and the signal yourself: the reason to write is your job.
2. Give the AI the real context: the exact text of the post or the job ad, the person’s role, and what you offer.
3. Set precise rules: three to five lines, a single question, no compliments, no pitch of your offer in the first message, and a first-name or more formal tone depending on how you usually write.
4. Give it examples of your own messages that got replies, so it picks up your tone.
5. Read every message before sending it, and fix any sentence that rings false.

Step three matters more than it looks. Without a length limit and a ban on compliments, even a good model falls back on the stock phrases listed above, because they are statistically the most common way to open a business message.

If you are unsure what counts as a good reason to write, our list of B2B buying signals with examples is a useful starting point for step one.

Example

Signal: a sales director posts about how hard it is to onboard new reps who work remotely.

Message generated without instructions: “Hi Mark, I hope this message finds you well. I was impressed by your background and your vision. In today’s fast-paced world, we help companies optimize their sales performance.”

Message reviewed and corrected: “Hi Mark, your post on onboarding remote reps raises a real question. On your team, is it learning the offer or getting up to speed on the tools that takes the most time?”

What prompt should you give AI for a prospecting message?

Here is a starting prompt to adapt to your offer. It follows the rules of the method above:

“You are writing the first LinkedIn prospecting message from [your first name], [your job]. Recipient: [first name], [role] at [company]. Real fact that justifies the message: [exact text of the post or announcement]. Rules: three to five lines, a professional but friendly tone, no pitch of our offer, no compliments, no stock opening line, a single simple question at the end, tied to that fact. Match the tone of these two messages that got a reply: [example 1] [example 2]. Do not add any fact that is not given above.”

Ask for two or three versions, keep the best one and edit it by hand. The last rule matters: it reduces the risk of the AI making up a detail about the company.

Which tasks should you hand to AI, and which should you keep?

Task	AI	Human
Spotting the fact that gives a reason to write	Can help sort	Chooses, because they know the offer
Summarizing the prospect’s situation	Yes, very well	Checks the summary is accurate
Writing a first draft	Yes, with rules and examples	Fixes tone and details
Translating or adapting to another language	Yes	Checks local expressions
Deciding whether to send	No	Yes, message by message
Carrying the conversation to a meeting	Can suggest replies	Approves, and handles delicate cases

The dividing line is simple: AI produces, the human chooses. The more a decision affects your relationship with the prospect, the more it stays human. In practice, AI can prepare dozens of drafts in the background while you only spend time on the moments that count: choosing, checking, sending.

How much time does AI really save?

AI saves time on writing, not on thinking. The first draft of a message goes from a few minutes to a few seconds. But choosing the prospect, checking the signal and reviewing the message are still necessary, and they are what separates a message that gets ignored from one that gets a reply.

💡 The real gain comes from combining the two: AI prepares, the human reviews at a glance. A small business owner can keep prospecting going in a few minutes a day this way, without sending messages they would not have written themselves. In practice, reviewing a well-prepared draft takes about as long as reading it twice. We go deeper into this approach in our article on why use AI for prospecting.

What mistakes should you avoid with AI in prospecting?

Four mistakes come up often when people start using AI to prospect:

1. ❌ Sending without reading: a single absurd message, with the wrong first name or a made-up fact, is enough to damage how a prospect sees you.
2. ❌ Letting it invent facts: AI may claim a company has raised funds or opened a new office when it has not. Only give it verified facts, and check whatever it repeats back.
3. Increasing volume instead of quality: writing faster should not be used to write to more people, but to write better to the right people.
4. Forgetting tone: without examples of your messages, AI writes in a style that is not yours, and the prospect notices as soon as you reply in your own voice.

These mistakes have one thing in common: they hand the AI a decision it cannot make. Checking facts and deciding to send remain your job. A simple safeguard helps: before sending, ask yourself whether you could answer calmly if the prospect asked where a given detail came from. If not, cut the sentence.

How do you keep prospecting human while using AI?

📌 Keeping prospecting human means keeping three things in your own hands: who you contact, why you are writing to them, and the approval of each message. Everything else can be assisted.

That is how MeetMagnet works: it spots decision-makers on LinkedIn who show a signal, suggests an opening line built on that signal following your rules and your tone, then lets you read and approve every message before it goes out. AI prepares, you decide, and the prospect receives a message that is about them. To see what happens after that first message, read our guide on going from a buying signal to a meeting.