The short answer
Sales prompts are reusable instructions you give an AI assistant to research prospects, draft outreach, prepare calls and handle objections. Good ones have five parts: your role and offer, verified facts about the prospect, a single task, constraints such as length and banned words, and the output format. The facts you paste in matter more than the wording.
On this page
Key takeaways
Start with a context block: Paste what you sell and what you know about the prospect first. Every prompt below assumes it's already in the chat.
One task per prompt: Ask for one email, one script or one list at a time. Bundled requests get shallow answers.
Constrain the output: Word limits, banned phrases and a required closing question do more for quality than asking for 'compelling' copy.
Make it check itself: End each session with a review prompt that flags any claim your research doesn't support.
How to Use This Sales Prompt Library
Every prompt here has the same five parts, and the context does most of the work. Fill in the parts once for your product and you'll reuse them all week.
| Part | What to include | Example |
|---|---|---|
| Role and offer | Who you are and what you sell, in one line | "I'm an AE at Northwind; we automate freight quoting for brokers." |
| Context | 2 to 4 verified facts about the prospect, with sources | "Halden Freight is hiring three pricing analysts (careers page)." |
| Task | One deliverable | "Write one cold email." |
| Constraints | Length, tone, banned phrases, required elements | "Under 90 words, end with one question, no 'synergy'." |
| Format | How you want it back | "Subject lines first, then the body." |
Paste this context block at the top of the chat, then run any prompt below.
Context: I'm a [role] at [company]. We help [buyer type] [outcome]
by [how]. Proof: [one customer outcome you can share].
Prospect: [company] ([website]). Contact: [name], [title].
Verified facts: [fact 1, source]. [fact 2, source]. [fact 3, source].
Tone: [plain, direct, no hype]. Never use: [banned words].
These prompts are tool-agnostic. They work in ChatGPT, Claude, Gemini or Copilot. If ChatGPT is your tool, our ChatGPT for sales guide covers ten wider use cases and the limits to watch.
Prospect Research Prompts
Research prompts should extract and rank facts, not invent them. Paste the text yourself (their About page, a press release, job posts) so the model works from sources you can check. For the manual routine, see how to research a prospect.
1. The fact extractor
Use it on raw pages before you write anything. The quotes make every fact easy to check.
From the text below, list every fact that could matter to a seller of
[product category]: changes, hires, launches, tools, goals, numbers.
Quote the sentence each fact comes from. Don't add anything that
isn't in the text.
[paste text]
2. The why-now finder
Use it when you have several facts and need the one to lead with.
Using only the verified facts in my context, pick the one change at
[company] most likely to create a need for [product]. Explain it in
3 lines: the change, the likely pain, why it matters this quarter.
If none of the facts support a why-now, say so.
3. The fit check
Use it before you spend real time on an account that might not qualify.
My ideal customer: [ICP criteria]. My disqualifiers: [list].
Based on the facts provided, call [company] a strong fit, possible fit
or poor fit. Give 2 reasons for, 2 against and the one thing
I should confirm on a first call.
Cold Email Prompts
Ask for angles, not just an email. Three short variants built on different facts teach you more than one long draft. More finished patterns live in our cold email examples.
4. Three angles, three emails
Use it to test angles across a segment, then keep the one that earns replies.
Write 3 cold emails to [name], each on a different angle:
(a) the why-now change, (b) a peer company's outcome, (c) a sharp question.
Each under 90 words, one idea, ending with a low-effort question.
Label each angle and the fact it relies on.
5. Subject lines and preview text
Short, lowercase-style subjects that look like a colleague wrote them tend to suit cold email better than marketing headlines.
For this email [paste], write 8 subject lines of 2 to 5 words that
a [title] would open because they read like internal mail.
No clickbait, no emojis, no 'quick question'.
Add a preview line of under 12 words for the best 3.
6. First lines from one fact
Use it when the rest of the email is solid and only the opener feels generic.
Write 5 opening lines for an email to [name] built on this fact: [fact].
Each under 20 words, stating the fact plainly and connecting it to
[problem]. No flattery, no 'I noticed', no 'I came across'.
LinkedIn Message Prompts
LinkedIn rewards brevity even more than email does. Keep connection notes under 200 characters so they fit a free account's note limit. The LinkedIn connection message guide has formats that get accepted.
7. The connection note
Use it for the first touch on LinkedIn, before any pitch.
Write 3 LinkedIn connection notes to [name], [title] at [company].
Each under 200 characters, built on [fact or shared context],
with no pitch and no meeting ask. Sound like a peer, not a vendor.
8. The first message after they accept
Use it within a day or two of the acceptance, while your name is still familiar.
[name] accepted my connection request. Write a first message under
60 words that references [fact], shares one useful thing
([resource or insight]) and ends with a question they can answer
in one line. No calendar link.
Cold Call and Objection Prompts
Call prompts should produce lines you can say out loud, not paragraphs. Ask for short sentences and read the output aloud before you dial.
9. Opener and reason for calling
Use it to write the first 20 seconds of the call, the part that decides whether they keep listening.
Write a cold call opener for [name], [title] at [company]:
who I am (1 line), why I'm calling them specifically, using [fact] (1 line),
a permission ask (1 line) and one discovery question.
Under 60 spoken words in total. Plain, conversational English.
10. Voicemail and gatekeeper lines
Use it before a calling block so you're never improvising at the front desk.
Write (a) a 20-second voicemail for [name] that references [fact] and
gives one reason to call back, and (b) a gatekeeper line that asks
who owns [problem area] without pretending I know the prospect.
11. Objection replies
Pair this with the full objection handling script for the structure behind each reply.
For a [title] considering [product], write replies to these objections:
[objection 1], [objection 2], [objection 3].
Use acknowledge, respond, ask: agree with something true, add one fact
or reframe, then ask a question. Each reply under 35 words.
Follow-Up Prompts
Every follow-up should add something new. "Just bumping this" teaches the prospect to ignore you. Plan the whole run with a cold email follow-up sequence.
12. The value-add follow-ups
Use it when the first email went unanswered and you have something new to add.
My first email [paste] got no reply. Write two follow-ups:
Day +3 adds [new fact, short insight or relevant example],
Day +7 asks a different, easier question. Each under 60 words.
Don't mention that they haven't replied.
13. The breakup email
Use it as the last step of a sequence, so the thread ends politely instead of fading out.
Write a final email to [name] that closes the loop politely, restates
the one problem in a sentence and offers an easy out: point me to the
right person, or 'not now, try me in [quarter]'. Under 50 words.
Sales Prompts That Check Your Draft
Finish every session by making the AI audit its own work. These two prompts catch the errors that cost you credibility with a buyer.
14. The claim checker
Use it on every AI draft that mentions the prospect. It catches invented details before the buyer does.
List every statement in this draft about [company] or [name].
For each, say whether it appears in the facts I gave you, is a
reasonable inference or is unsupported. Then rewrite the draft
without the unsupported ones.
[paste draft]
15. The skeptical buyer
Use it when a draft feels fine but you can't say why it would get a reply.
You're a busy [title] who gets dozens of sales emails a week.
Read this draft and give me 3 reasons you'd delete it, then a version
you'd actually reply to. Keep my facts.
[paste draft]
Running Sales Prompts for Every Account
Prompts work, but the copy-paste adds up. Gathering facts, pasting the context block and checking the output takes real time per account. Left alone, each rep also builds a private prompt collection and drifts toward a slightly different pitch.
- Share one library. Keep the team's best prompts in one doc so everyone starts from the same context block. Our guide to AI for sales teams covers rollout and governance.
- Version your context block. Update it when your positioning or proof points change.
- Track what gets replies. Note which angle each sent email used, then do more of what works.
When volume grows, a tool that stores the context and does the research step replaces most of these prompts. In Edithly, your product playbook is built once from your website. Paste a prospect's site and you get emails with a personalization basis, a call script or a one-pager, written around that account. See the Edithly platform for how the pieces fit.
How Edithly does it
Get outreach for any prospect without writing a prompt
Edithly already knows your product from your website, so the context block is built in. Paste a prospect's website into the Email Outreach tile and it researches the account and writes the angles, follow-ups and LinkedIn note.
- 1
Click the Email Outreach tile
On Home, pick Email Outreach. Your pitch, proof points and positioning are already in your playbook, read from your own website.
- 2
Enter the prospect's website
Add the company name and website, plus the contact's name, LinkedIn, Notes on the angle and a Tone if you have them.
- 3
Generate, review, send yourself
Click Generate email outreach. Review the angles, subject lines, follow-ups and LinkedIn note, then copy them into your inbox or sequencer.
Website
Ready
- Variant A · Hiring angle
- Subject lines
- Body
- Follow-ups · Day +3, +7
- LinkedIn note
- Do not say
- By hand
- 10 to 20 minutes per account to gather facts, prompt, check and edit
- With Edithly
- A couple of minutes per account
Frequently asked questions
What are good AI prompts for sales?
Good sales prompts are specific about inputs and output. They give the AI your offer, two or three verified facts about the prospect, one task and hard constraints such as 'under 90 words, end with one question, no buzzwords'. Prompts for research, cold emails, LinkedIn notes, call openers, objection replies and follow-ups cover most of a seller's week.
How do you write a sales prompt?
Use five parts: role and offer (who you are and what you sell), context (verified facts about the prospect), task (one deliverable), constraints (length, tone, banned phrases) and format (table, bullets or plain text). Then add a check: ask the AI to list any claim it made that isn't in the facts you gave it.
Do these sales prompts work in ChatGPT, Claude and Gemini?
Yes. They're written in plain language with bracketed inputs, so they work in any general AI assistant, including ChatGPT, Claude, Gemini and Microsoft Copilot. Results vary a little by model, but the inputs you paste in decide most of the quality. Keep your best versions in a shared doc so the whole team starts from the same ones.
Why does AI sales copy sound generic?
Because the prompt was generic. 'Write a cold email to a VP of Sales' gives the model nothing specific, so it falls back on safe, average phrasing. Add a real fact about the account, the exact problem you solve, a word limit and a list of banned phrases, and the output changes noticeably.
Do I still need sales prompts if I use Edithly?
Not for outreach and collateral. Edithly stores your product playbook once, then reads each prospect's website when you generate, so the context a prompt would carry is already there. You click a material tile, or ask the assistant for a visual, and get emails, a call script or a deck. General prompts stay useful for other jobs, such as deal reviews.