# Using ChatGPT for sales prospecting
> A practical workflow for using ChatGPT to turn LinkedIn profiles into specific outreach, with prompts, short examples and a human editing pass.
- HTML: https://getkinlead.com/chatgpt-sales-prospecting
- Markdown: https://getkinlead.com/chatgpt-sales-prospecting.md

**Short answer:** To use ChatGPT for sales prospecting, stop asking it to write a message from a blank page. Give it a specific sales playbook, a target persona, and raw profile text, then ask for one observation you can check. The output should be a conversation starter that names a real problem, not a polished paragraph about the lead's impressive background.

## The mistake of generic prompting

Most people treat ChatGPT like a copywriter. They paste a profile and ask for a message. The result reads like a brochure: fake enthusiasm, generic praise, and a request that could have been sent to anyone.

The model does not know your offer, your market, or why this person should care. It fills the gap with polite filler. To get a useful result, stop asking it to write and make it read the profile first.

The fix is to separate research from writing. Ask the model to find the hook first. Write the message after you agree with the hook. That one change removes most of the generic praise because the model has a specific job.

**The generic prompt**

- Write a personalized LinkedIn message to this person based on their profile.
- Draft a friendly outreach note to a CEO of a marketing agency.
- Use this profile to write a connection request that gets a high response rate.

**The research prompt**

- Using the playbook below, name one problem this person likely has from their job title and recent posts. Quote the line that made you think it.
- Read the About section. Find one project they mention and explain in under 15 words how it connects to our offer.
- They posted about hiring a sales team. Ask one question about the onboarding bottleneck that follows a hire like that.

## Building your AI sales playbook

ChatGPT needs a framework so it stops guessing. A sales playbook is a short document that tells the model how you think about your market, what makes a lead relevant, and what a good message looks like. Without it, the model falls back on its training data, which is usually too formal for LinkedIn.

Keep the playbook in one place and paste it into each new chat, or upload it to a custom GPT. It is the guardrail for every message you generate.

The playbook should be specific enough that two people could read it and agree on whether a lead fits. If it says 'we help companies grow', it is useless. If it says 'we help 20 to 50 person agencies replace manual reporting for clients', the model can match that to a profile.

1. **Define the core offer.** Write one sentence stating what you do and the single biggest problem it solves. Example: We help 20-person agencies automate reporting so founders stop spending Sundays in spreadsheets.
2. **Map the trigger events.** List 3 to 5 events that make a lead need you now. Examples: hiring a new VP of Sales, launching a product, or posting about a specific industry problem.
3. **Provide gold-standard examples.** Paste 3 opening lines you wrote by hand that got a reply. Tell the model to mimic the length and tone of those examples, not to invent a new style.
4. **Set negative constraints.** List words and phrases the model cannot use. Ban 'I hope this email finds you well', 'excited to connect', 'quick question', and 'I'd love to pick your brain'.

```
You are a sales researcher, not a copywriter. Use the playbook below. Read the profile text and return: 1) the one trigger event that fits the playbook, 2) the exact line from the profile that supports it, 3) a question I can ask in a connection note. Do not write the note yet.

Playbook: [paste]
Profile: [paste]
```

## The prospecting workflow

Once the playbook is in place, use the same process for each lead. Doing one at a time is slow, but asking for 100 at once kills quality. Work in small batches and keep the profile text close to the prompt.

You are not looking for a full message yet. You want the specific reason to reach out today. That reason becomes the hook, and the hook is what makes the message different from the last one you sent.

The table is not a message generator. It is a filter. If the model cannot name a trigger event and quote the line that supports it, you do not have a reason to reach out. Move to the next lead or use a company-level observation instead.

| | Step | Action | Input needed |
| --- | --- | --- | --- |
| 1 | Paste the playbook into a new chat. | Playbook document |
| 2 | Paste the lead's About section and last two posts. | LinkedIn profile text |
| 3 | Ask the model to name the trigger event and quote the line that supports it. | Playbook and profile |
| 4 | Ask for a connection note under 200 characters that opens with the hook. | The identified hook |

## Refining the output

AI output needs a human edit. Even with a good playbook, the model will slip back into assistant mode and add adjectives that sound like a performance review. If the message feels too polished, it is.

People on LinkedIn write in fragments. They use lower-case letters sometimes. They do not use three adjectives to describe a company. Strip the shine until the message looks like it was typed on a phone in 30 seconds.

Read the draft out loud. If you would not say it to someone standing next to you, rewrite it. The test is not whether it sounds professional. The test is whether it sounds like a person who looked at the profile.

- **draft:** Hi Sarah, I was impressed by your distinguished career in SaaS. I would love to connect and share how our tool can help your workflow.
- **you:** Hi Sarah, saw your post about churn in Q3. Have you tried automating the offboarding survey yet?
- **them:** We haven't. The survey is still manual and it takes a week to get answers. Is that what your tool does?

> Delete every adjective that praises the lead. Words like 'distinguished', 'impressive', or 'incredible' tell the prospect you generated the message before reading the profile.

## How to handle profiles with little to say

Some profiles are thin. The person has a headline, a company name, and no posts. Do not force a personal hook from nothing. Use the company and role to make a relevant guess, and say what made you look.

Read the company page, recent job posts, and the person's headline. A job post that repeats the same role three times tells you the team is growing or struggling to hire. That is a real observation, even if the profile is quiet.

Use a short question that is easy to answer. Do not ask for a meeting in the first line. The goal is a reply, not a calendar link.

- Company page: recent posts, new locations, product launches.
- Job posts: roles that repeat, roles that stay open, roles with unusual requirements.
- Headline: a title change, a new company, or a phrase that hints at a project.
- Mutual connections: someone you can mention without making the message about you.

## Batching leads without losing quality

Once the prompt works on one lead, save it as a template. Then process 10 leads at a time. Paste each profile in a separate block and number them. Ask for a table with the lead number, the trigger event, the exact line from the profile, and a draft question. Do not ask for full messages in the same pass.

Read the table before you generate anything. If a row has no exact line from the profile, the hook is invented. Delete that row and skip the lead, or send a message that asks about the company instead of pretending to know the person.

After the table looks right, ask for connection notes one by one or in a batch with the same playbook. Keep the notes under 200 characters. If a note needs a comma splice to fit, the idea is too complex for a connection request. Save it for the first message.

## Doing this with Kinlead

If copy-pasting between LinkedIn and ChatGPT is too slow, Kinlead handles the lead finding and scoring. It finds people on LinkedIn who match the customer you describe or your website and scores each lead from 0 to 100 against that profile. Leads below the fit line are not kept.

Leads live in lead groups. Any campaign can point at any group, and leads can be exported to CSV or imported from a CSV of LinkedIn profile URLs. The Steal Customers agent watches competitor company pages and optional founder or employee profiles, reads comments on recent posts, and turns commenters who look like buyers into leads. Each lead keeps the post URL, what the post was about, and what they commented.

Connection requests can carry an AI-written note under 200 characters that opens with why that person became a lead. Messages and follow-ups are written from the campaign goal, tone, and optional sales playbook. Queued messages can be read and edited on the Actions page before they send, and any campaign can be paused. Outreach runs from your own LinkedIn account through Unipile, and Kinlead never stores the LinkedIn password. It sends at a human pace, at most one action every 5 minutes, inside the daily invite and message limits you set and inside a send window in the lead's time zone. See [AI LinkedIn outreach](/features/ai-linkedin-outreach).

## Related

- [AI sales playbook guide](https://getkinlead.com/ai-sales-playbook-for-linkedin-outreach.md)
- [Connection request templates](https://getkinlead.com/ai-linkedin-connection-request-templates.md)
- [AI LinkedIn outreach](https://getkinlead.com/features/ai-linkedin-outreach.md)

## Common questions

**Can ChatGPT find LinkedIn profiles for me?**

No. ChatGPT cannot browse LinkedIn in real time to find new people for you. You have to provide the profile text or use a separate lead search tool.

**How do I avoid getting my account restricted when using AI messages?**

Do not send the same note to everyone. Use the model to vary the hook for each person, keep your volume inside whatever limit your account has, and stop when LinkedIn pushes back. LinkedIn does not publish exact invite or message limits, so there is no safe number to copy.

**Should I use a custom GPT or a standard chat for prospecting?**

A custom GPT is easier because you can keep your sales playbook in the knowledge file and avoid pasting it every time. A standard chat works if you paste the playbook at the start of each session and check the output.

**What if the profile has no posts?**

Use the company page, job posts, and headline to find a real observation. Ask a short question about the work, not about the person's background.

[Create a Kinlead account](https://getkinlead.com/signup) or [book a call](https://getkinlead.com/demo) to talk through pricing and setup.
