The best ChatGPT prompts for sales prospecting share one structure: assign the model a role, paste in real prospect data, state a single task, and specify the output format. A prompt like "write a cold email to a CFO" produces generic filler. A prompt that includes the prospect's actual job posting, a verified trigger event, and a 90-word limit produces something you can send.
Below are 23 copy-paste prompts organized by the job to be done: account research, ICP scoring, cold email drafts, call prep, objection handling, follow-up, and pipeline review. They work in ChatGPT (GPT-5), Claude, and Gemini. Replace the bracketed placeholders with your own material, and always paste in source data rather than asking the model to remember facts about a company. AI use in sales is now the norm, not the edge: Salesforce's seventh edition State of Sales report, based on a survey of more than 4,000 sales professionals, found that 87% of sales organizations already use some form of AI for tasks like prospecting, forecasting, and drafting emails. The differentiator is no longer whether you use these tools. It is whether your prompts feed them real signal.
How to structure a sales prompt
Both OpenAI's and Anthropic's prompt engineering guides converge on the same skeleton. Four parts:
- Role. Tell the model who it is: "You are an SDR preparing outreach" or
"Act as a skeptical revenue analyst." Role framing sets vocabulary and judgment.
- Context. Paste the actual data: the prospect's homepage text, the job
posting, the call transcript, your CRM notes. The model is only as accurate as what you give it. Never ask it to recall facts about a specific company from memory.
- Task. One task per prompt. "Summarize, then draft, then critique" in one
message produces mush. Chain prompts instead.
- Format. Specify length, structure, and constraints: "under 90 words,"
"return a table," "quote the exact lines you used." Format constraints are what make output checkable.
Three habits multiply the value of every prompt below. Ask for variants (three openers beat one). Ask the model to flag uncertainty instead of guessing. And verify every claim before it reaches a prospect: a model will confidently invent a funding round that never happened.
Account research prompts
1. Company snapshot from source material
`` You are an SDR preparing outreach to [company]. I am pasting the text of their homepage, latest press release, and one current job posting. Using only this material, summarize: what they sell, who they sell to, how they likely make money, and three recent changes that could open a sales conversation. Flag anything you are unsure about instead of guessing. ``
2. Earnings and annual report scan
`` Act as a revenue analyst. Here is an excerpt from [company]'s latest earnings call or annual report: [paste]. List their stated priorities, the risks they name, and any initiative that suggests they would buy [your category]. Quote the exact lines you relied on so I can verify them. ``
3. Persona research from a job description
`` I sell [product] to [role] at [industry] companies. Based on the responsibilities in this job description [paste], list the five outcomes this person is judged on, the likely metric behind each, and where [product] plausibly helps. Label each connection "plausible" or "proven." ``
4. Trigger scan across a list
`` Here are news summaries from the last 90 days for my top 10 accounts: [paste]. For each account, name the single strongest buying trigger and why it qualifies, or write "no trigger" if nothing does. Do not invent events. ``
ICP scoring prompts
Scoring only works when the underlying ICP definition is written down. If yours is not, start with the market layer of the Tenbound Pipeline Architecture Standard before you automate anything.
5. Turn your ICP into a rubric
`` Help me turn my ideal customer profile into a scoring rubric. Our 10 best customers look like this: [paste firmographics, use case, and the trigger that made them buy]. Propose 5 to 7 scoring criteria with weights that sum to 100, plus a one-line rule for scoring each criterion 0, 1, or 2. ``
6. Score a batch of accounts
`` Score the following accounts against this rubric: [paste rubric]. Accounts: [paste one row of data per account]. Return a table: account, score per criterion, weighted total, and the single missing piece of information that would most change the score. ``
7. Argue for disqualification
`` Here is our ICP: [paste]. Here is an account brief: [paste]. Make the strongest case for disqualifying this account. If the case is weak, say so and explain what evidence would change your view. ``
Cold email prompts
8. First-touch draft from verified facts
`` Write a first-touch email to [name], [title] at [company]. Context I have verified: [paste 2 or 3 facts and the trigger event]. Constraints: under 90 words, one idea, one question at the end, no buzzwords, no "I hope this finds you well," and reference only the facts I gave you. Give me 3 variants with different opening lines. ``
9. Rewrite my own draft
`` Rewrite this cold email I drafted: [paste]. Keep my facts and my ask. Cut it to under 80 words, simplify every sentence, and remove any claim I did not support with the context provided. Then list what you removed and why. ``
10. Subject lines
`` Here is the final email: [paste]. Give me 10 subject lines of 5 words or fewer. No clickbait, no fake "re:" replies, no all caps. Rank them by how specific they are to this prospect. ``
11. Follow-up sequence
`` Build a 4-step follow-up sequence behind this first email: [paste]. Steps at day 3, day 7, day 14, and day 21. Each step must add something new (a stat, a customer example, a sharper question), never "just bumping this." Under 60 words per step. Use only these proof points: [paste]. ``
Call prep prompts
12. One-page call brief
`` Build a one-page brief for my meeting with [name], [title] at [company]. Inputs: [paste research notes, CRM history, and the email thread]. Sections: who they are, why they likely took the meeting, three discovery questions tied to their situation, the objection I should expect, and the next step I should propose. ``
13. Discovery questions
`` I sell [product]. The prospect's situation: [paste]. Write 8 discovery questions that test whether the problem is real, funded, and urgent. No yes/no questions. Order them from safe to bold. ``
14. Role-play the prospect
`` Role-play as [title] at a [industry] company evaluating [product category]. You are skeptical but fair. Your context: [paste]. I will practice my opening. Respond as the prospect would. After each of my turns, break character and give me one specific coaching note. ``
Objection handling prompts
15. Build an objection map
`` List the 8 most likely objections to this pitch: [paste pitch and pricing]. Classify each as misunderstanding, real limitation, or brush-off, and draft a 2-sentence response. For real limitations, do not spin: state the honest trade-off. ``
16. Diagnose a live objection
`` A prospect replied: [paste the exact reply]. Diagnose what they are actually saying. Then draft two responses: one that keeps the conversation alive and one that closes the loop politely and asks what future event would make this relevant. Under 60 words each. ``
17. Competitor comparison
`` A prospect asked how we compare to [competitor]. Here is what I know to be true: [paste verified differences]. Draft a response that concedes what they do well, states where we differ, and offers a way to test the difference. Do not disparage them and do not add claims beyond my list. ``
Follow-up prompts
18. Post-call recap
`` Here are my call notes or transcript: [paste]. Draft a recap email under 120 words: what we agreed, what I owe them, what they owe me, and a proposed next step with a date. Match their level of formality. Quote their own words for the key pain point. ``
19. Re-engage a quiet prospect
`` This prospect went quiet. Last two messages: [paste]. Draft 3 different re-engagement notes: one that adds new information, one that asks a direct closing question, and one that offers to stop reaching out. Under 50 words each. ``
20. Closed-lost nurture plan
`` This deal closed lost because [reason]. Their likely future trigger: [event]. Draft a 3-touch nurture plan over the next two quarters. Each touch must give the prospect something useful, with no ask until the final touch. ``
Pipeline review prompts
21. Weekly pipeline hygiene
`` Here is my pipeline export with stage, age, value, last activity, and next step: [paste]. Flag every opportunity with no documented next step or more than [X] days of inactivity. Then name the three deals where this week's effort most changes the quarter, with your reasoning. ``
22. Forecast challenge
`` I claim these deals close this quarter: [paste list with notes]. Interrogate each one: is there a named decision process, an agreed date, access to the economic buyer, and a slip risk? Return a table: deal, biggest risk, one action to take this week. ``
23. Loss-pattern analysis
`` Here are my last 20 lost opportunities with stage lost and stated reason: [paste]. Group them by the earliest cause, not the final CRM reason. Then tell me the one change to targeting or qualification that would remove the largest group. ``
What not to automate
The failure mode in 2026 is not bad prompts. It is sending unverified model output to real buyers. Keep humans on four jobs:
- Fact verification. Every company fact, statistic, and name in outbound
copy gets checked against the source before sending. Models fabricate specifics with confidence.
- The send decision. A person approves what reaches a prospect, at least
until a message pattern has a proven track record with a defined segment.
- Live conversations. Discovery calls, negotiations, and relationship
repair are human work. Use the model to prepare, never to impersonate.
- Judgment calls with reputation risk. Pricing exceptions, competitive
claims, and anything touching a customer's data or legal terms.
The pattern behind all 23 prompts is the same: the model drafts and analyzes, you supply the truth and make the call. Teams that hold that line get speed without the credibility tax.