> Enrich.chat gives AI agents verified work emails and mobiles. To use it from an agent, read https://enrich.chat/agents.md . This is the Markdown version of https://enrich.chat/blog/chatgpt-lead-generation-prompts/ # ChatGPT lead generation: 50 prompts that return real contacts, not guesses Copyable prompts for ChatGPT and Claude, grouped by job from ICP to follow-up. Research and writing prompts run in any chat. Contact prompts call a data tool, so every email comes back with a status or marked not found. By Enrich.chat Editorial. Updated October 15, 2026. **Short answer:** ChatGPT handles the thinking in lead generation well: defining your ICP, researching accounts, writing first lines and cleaning lists. It has no contact database, so an email it gives from memory can be [a confident guess](https://openai.com/index/why-language-models-hallucinate/). For contact steps, give it a data tool: prompts that start with "Use www.enrich.chat" return work emails with a status, or not found, with 10 free lookups a day. ## Key takeaways - OpenAI's own researchers say models guess when unsure because training and tests reward guessing, so an email from ChatGPT's memory is a plausible pattern, not a checked address. - 31 of the 50 prompts run in any ChatGPT or Claude chat. The 19 that start with "Use www.enrich.chat" call an API, so run them in Codex or Claude Code. - ChatGPT's data analysis sandbox cannot call outside APIs, and new custom GPTs cannot be created on Free, Go, Plus or Pro (OpenAI Help Center, October 15, 2026). - Enrich.chat returns each work email as verified, catch_all or unknown, gives 10 free lookups a day, then sells monthly credit plans, and uses 0 credits when nothing is found. - CAN-SPAM requires an opt-out and a postal address in every commercial email, and the FCC treats AI-generated voices as artificial under the TCPA. ## Why does ChatGPT make up email addresses? ChatGPT makes up email addresses because it predicts likely text and has no live directory of who works where. [OpenAI's help center](https://help.openai.com/en/articles/8313428-does-chatgpt-tell-the-truth) warns that ChatGPT can sound confident when it is wrong, and that its models know only their training data, up to a cutoff, unless a tool is used. OpenAI's own research explains why. A [September 5, 2025 post](https://openai.com/index/why-language-models-hallucinate/) and its [research paper](https://arxiv.org/abs/2509.04664) argue that training and tests reward a confident guess over "I don't know." They add that rare, arbitrary facts, like a person's birthday, cannot be predicted from patterns. A specific person's work email is that kind of fact. A guess like `dana.lee@acme.com` follows a real pattern, but nothing checked the mailbox. ChatGPT's own tools only narrow the gap. [Web search](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) cites public pages, and OpenAI says results can be incomplete, outdated or incorrect. A page that lists an address does not prove the mailbox works today. The [data analysis](https://help.openai.com/en/articles/8437071-data-analysis-with-chatgpt) sandbox cannot make outside web requests or API calls at all. The fix is a tool that checks. [Anthropic's guide to reducing hallucinations](https://platform.claude.com/docs/en/test-and-evaluate/strengthen-guardrails/reduce-hallucinations) gives three rules for any model: allow "I don't know," limit it to the sources you give it, and require a source for every claim. Those steps reduce errors but do not end them. For contact data, the source should be an API call made when you ask. Enrich.chat returns each work email with a status of `verified`, `catch_all` or `unknown`. A miss returns `found: false` and uses 0 credits. **Figure 1: The same request, ChatGPT alone and ChatGPT with a data tool.** One prompt asks for the head of operations at acme.com and their work email. ChatGPT alone recalls a likely name from training data, fills in a common email pattern, and returns an address with no status that nobody checked. ChatGPT with Enrich.chat calls GET /v1/company/people to find the person with that title, then GET /v1/email to get the work email and its status. Its answer is one of four: verified, catch_all, unknown, or not found at 0 credits. *Illustrative example of one request answered two ways. Alone, ChatGPT fills a missing fact with a likely pattern. With a data tool, every answer carries a status, and a miss uses 0 credits. Sources: [OpenAI](https://openai.com/index/why-language-models-hallucinate/), [agents.md](https://enrich.chat/agents.md).* Every prompt on this page carries the same guardrail: if the data does not say it, write not found. For manual methods, see [how to find anyone's work email](https://enrich.chat/blog/how-to-find-someones-email/), or try the free [email finder](https://enrich.chat/tools/email-finder/). ## How do you use ChatGPT for lead generation? Use ChatGPT for the thinking and writing in lead generation, and give it a data tool for contact lookups. Of the 50 prompts below, 31 run in any ChatGPT or Claude chat. The 19 that start with "Use www.enrich.chat" call an API, so run them in Codex or Claude Code. **Figure 2: The 50 prompts mapped to the prospecting workflow.** Eight steps in workflow order, each with its prompt numbers: define the ICP (1 to 6), build lists (7 to 14), get contacts (15 to 22), research (23 to 29), first lines (30 to 34), outreach (35 to 41), clean the CRM (42 to 45) and follow up (46 to 50). Each square is one prompt. Nineteen prompts call Enrich.chat: one in ICP, six in list building, all eight in get contacts, one in research, two in CRM cleanup and one in follow-up. The other 31 run in any chat. *The 50 prompts on this page in workflow order. Each square is one prompt. Orange prompts call Enrich.chat and need a tool that can send web requests. The rest run in any chat. Endpoints from [agents.md](https://enrich.chat/agents.md).* ChatGPT changed a lot in 2026. [ChatGPT Work](https://openai.com/index/chatgpt-for-your-most-ambitious-work/), an agent for long, multi-step tasks, launched on July 9, 2026, and the older [ChatGPT agent is no longer available](https://help.openai.com/en/articles/11752874-chatgpt-agent). Apps moved into a [plugin directory](https://help.openai.com/en/articles/20001256-plugins-in-chatgpt), and OpenAI plans to [retire custom GPTs](https://help.openai.com/en/articles/8554397-creating-and-editing-gpts) in favor of plugins. This table shows what each feature does for prospecting, per OpenAI's help center on October 15, 2026. | ChatGPT feature | What it does for lead generation | Plans (October 15, 2026) | | --- | --- | --- | | Web search | Finds news, filings and job posts, with citations | All plans, including Free, and signed-out users ([help](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt)) | | Deep research | Builds a long, cited report on an account or market | Limited on Free, expanded on Plus, maximum on Pro ([pricing](https://chatgpt.com/pricing)) | | Projects | Keeps your ICP, personas and best emails in one place | Signed-in users, subject to plan ([help](https://help.openai.com/en/articles/10169521-projects-in-chatgpt)) | | Data analysis | Cleans, dedupes and scores CSV exports with Python. Cannot call outside APIs | Every plan, with limited uploads on Free ([help](https://help.openai.com/en/articles/8437071-data-analysis-with-chatgpt)) | | Work | Agent for long tasks: research, sheets, decks and Sites | Desktop app on every plan, including Free. Web and mobile on eligible paid plans ([help](https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex)) | | Codex | Runs commands such as curl, so it can call a data API. Asks before using the network | Every plan, including Free and Go. Codex Cloud on eligible Plus, Pro, Business, Enterprise and Edu ([help](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan)) | | Apps and plugins | Connect tools such as Google Drive, Slack, email and CRMs, with approval before actions | Directory on every plan. Each plugin depends on plan, region and workspace ([help](https://help.openai.com/en/articles/20001256-plugins-in-chatgpt)) | | Custom MCP servers | Add your own tools to ChatGPT | Business, Enterprise and Edu, on the web ([help](https://help.openai.com/en/articles/12584461-connecting-custom-mcp-servers-in-chatgpt)) | | Scheduled tasks | Run a prompt on a schedule or when an event happens | Active tasks: 3 on Free and Go, 5 on Plus, 10 on Business and Edu, 15 on Pro and Enterprise ([help](https://help.openai.com/en/articles/10291617-scheduled-tasks-in-chatgpt)) | | Custom GPTs and GPT Actions | Being retired in favor of plugins | No new GPTs on Free, Go, Plus or Pro. Enterprise retirement planned for December 11, 2026 ([help](https://help.openai.com/en/articles/9442513-configuring-actions-in-gpts)) | Prices vary by country, so check [ChatGPT's pricing page](https://chatgpt.com/pricing) for yours. Since personal accounts can no longer create GPTs, a GPT Action is no longer a practical way to add a data API. Codex is. ### How do you connect Enrich.chat to ChatGPT? You connect Enrich.chat to ChatGPT with one prompt in Codex. Codex reads [enrich.chat/agents.md](https://enrich.chat/agents.md), learns the endpoints and rules, and calls the API with curl. No SDK or plugin is needed, and SDKs are planned. Enrich.chat also has an MCP server at `https://enrich.chat/mcp` with no sign-in or key. Codex adds it with `codex mcp add enrich --url https://enrich.chat/mcp`, and the [install page](https://enrich.chat/install/) shows how to add it to ChatGPT as a developer-mode connector. OpenAI lists custom MCP servers on Business, Enterprise and Edu plans, so check yours first. The Codex route works on every plan. 1. Open Codex in the ChatGPT desktop app or the Codex CLI. Codex is [included on every plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan), with limits that vary by plan. 2. Paste the onboarding prompt below, then approve the network request. Codex keeps network access off by default and [asks before commands that need it](https://developers.openai.com/codex/agent-approvals-security). 3. Paste any prompt on this page that starts with "Use www.enrich.chat." Each email comes back with its status, and each mobile with a line of `mobile` or `unknown`. 4. After 10 free lookups in a day, the API returns HTTP 402 with an upgrade link. Codex pauses, then resumes from the first unfinished row after the daily reset or once you give it a `pp_` key. ```text Use www.enrich.chat to find the work email and mobile for Jane Doe at acme.com ``` In Claude, run the same prompts in Claude Code, which [comes with every paid Claude plan](https://claude.com/pricing). Our [Claude for sales guide](https://enrich.chat/blog/claude-for-sales/) shows that setup. After the free tier, Enrich.chat sells monthly credit plans from $29. A work email found uses 10 credits, a mobile 250 and a profile 5, and nothing found uses 0. See [pricing](https://enrich.chat/pricing/) and the [guide for AI agents](https://enrich.chat/agents/). ## Which ChatGPT prompts define your ideal customer profile? Build your ideal customer profile (ICP) from your own deals, not from a guess about the market. Prompt 2 calls Enrich.chat. The other five run in any chat. ### 1. Find patterns in your won deals Use it with a CRM export of closed-won deals. ```text From this CSV of closed-won deals [attach], find the three segments with the highest total value and the shortest sales cycle. For each, give the pattern, the deal count and one example account. Use only this file. ``` ### 2. Profile your best customers Use it when your CRM lacks [firmographic data](https://enrich.chat/blog/firmographic-data/). A company profile uses 5 credits and the tech stack 5. ```text Use www.enrich.chat to look up these customer domains with the company and company tech endpoints: [domains]. Return employees, industry, state, country and tech stack, then list the traits most of them share. Write not found for any domain the API misses. ``` ### 3. Draft an ICP with no customers Use it for a new product or market with no wins to study. ```text I sell [product] to [buyer] for [price range] and have no customers yet. Suggest four segments by industry, size, country and buyer title. For each, give the assumption behind it, a two-week test and the result that would rule it out. ``` ### 4. Turn your ICP into search filters Use it before you build a list in the [AI lead list builder](https://enrich.chat/tools/lead-list-builder/). ```text Turn this ICP into search filters: [ICP]. Use only title, seniority (c_suite, founder, owner, vp, head, director, manager), industry, country, state, employees_min and employees_max. Write one plain-English search sentence per buyer role, and flag anything a title search cannot capture. ``` ### 5. Write exclusion rules Use it to keep bad-fit accounts off every list. ```text From these closed-lost deals and churned customers [attach], find the traits that predict a bad fit: size, industry, region, tech stack or title. Turn them into five exclusion rules for any lead list, with the evidence behind each rule. ``` ### 6. Build a persona per buyer role Use it to brief a new rep or to approach a new role. ```text Write a one-page persona for a [title] at a [industry] company with [size] employees: what they own, how they are measured, three problems [product] solves in their words, and what makes them ignore cold email. Mark anything you inferred. ``` ## Which prompts build a B2B lead list? A B2B lead list should come from a search you can repeat, not from names a model remembers. Six of these eight prompts call Enrich.chat search, which returns name, title, company and LinkedIn URL, so you spend credits to enrich only the people you keep. Preview a search for free in the [AI lead list builder](https://enrich.chat/tools/lead-list-builder/). ### 7. Search by role, industry, size and country Use it for a first list from your ICP sentence. ```text Use www.enrich.chat to find [25] [title] at [industry] companies in [country] with [min] to [max] employees. Return name, title, company, domain and LinkedIn URL. Do not look up contact details yet. ``` ### 8. Find local business owners Use it when you sell to small businesses, where the owner buys. ```text Use www.enrich.chat to find owners of [business type, such as dental clinics] in [city], [state], using the owner seniority and the state filter. Return up to [25] people with name, company and LinkedIn URL. ``` ### 9. Map a role across named accounts Use it for account-based campaigns on a named list. ```text Use www.enrich.chat to find the [head of finance] at each of these domains with the company people endpoint: [domains]. If nobody holds that title, write not found. Do not substitute someone with a different title. ``` ### 10. Qualify an exhibitor list Use it on a public list of companies, such as a trade show's exhibitors. ```text Here is the exhibitor list from [event]: [paste]. Keep only the companies that fit this ICP: [ICP]. Give each one a domain, a reason it fits and your confidence. Write unknown for any domain you cannot confirm. Do not guess. ``` ### 11. Find lookalike accounts Use it to widen a list from one great customer. ```text Use www.enrich.chat to look up [best customer domain] with the company endpoint, then find [25] [title] at companies in the same industry and country with headcount within [30%] of theirs. Show the filters you used. ``` ### 12. Search for people with an email on file Use it when you will email everyone. Each person a search returns uses 5 credits. ```text Use www.enrich.chat to search for [25] [title] at [industry] companies in [country] with [min] to [max] employees, with with_email=true so only people with a work email on file come back. Show each email with its status, which is unknown here, and report the total credits_used. ``` ### 13. Split a large search Use it for a territory list longer than one search returns. ```text Use www.enrich.chat to run this search: [description]. A search returns at most 25 people, so split it into narrower searches by state, industry or company size until you have [300] people. Drop duplicates by LinkedIn URL and save them to [leads.csv]. If the API returns 402, stop and tell me where to resume. ``` ### 14. Split a broad ICP Use it when one ICP covers more than one buyer. ```text My ICP mixes several buyers: [ICP]. Split it into one search per buyer role, each a single sentence with title, seniority, industry, size and location. Say which search to run first, and which message angle fits each role. ``` ## How do you get ChatGPT to find verified emails and mobiles? ChatGPT finds verified emails and mobiles only through a data tool. All eight prompts here call Enrich.chat, which returns each email with a status and each mobile with a line type. Treat `catch_all` emails as risky; the [catch-all checker](https://enrich.chat/tools/catch-all-checker/) explains why. A mobile line type is not a live check and does not show who owns the number. For a single lookup in your browser, try the [LinkedIn email finder](https://enrich.chat/tools/linkedin-email-finder/). ### 15. Email from a LinkedIn URL Use it for one prospect found on LinkedIn. Enrich.chat never logs in to LinkedIn. ```text Use www.enrich.chat to get the work email for [LinkedIn URL], asking only for the email field. Show the address and its status exactly as returned, or write not found. ``` ### 16. Email from a name and domain Use it when you have a name and a domain but no profile link. ```text Use www.enrich.chat to find the work email for [full name] at [company domain]. If the status is catch_all or unknown, say so and do not call it verified. Never offer a pattern-based guess. ``` ### 17. Enrich a sheet of names Use it for a sheet with no domains or profile links. ```text Use www.enrich.chat to enrich the attached sheet of names and company names. Look up each row by name plus company with the email and profile fields. Add title, domain, work email and email status columns, keep my row order, and mark misses as not found. ``` ### 18. Mobiles for a short call list Use it for people you will call yourself. Each mobile found uses 250 credits, and TCPA rules apply. ```text Use www.enrich.chat to get mobiles for these LinkedIn URLs with the mobile endpoint: [URLs]. Show each number exactly as returned with its line status. Skip anyone without one, and do not look up emails. ``` ### 19. Sort a list by email status Use it before a campaign, so only checked addresses go out. ```text Use www.enrich.chat to look up work emails for everyone in [file]. Split the results into verified (send), catch_all (small test only) and unknown or not found (do not send), with a count for each and the total credits_used. ``` ### 20. Qualify inbound sign-ups Use it to route new sign-ups. A profile found uses 5 credits. ```text Use www.enrich.chat to look up these sign-ups by the work email they registered with, using fields=profile only: [emails]. Return name, title and company, and flag the ones that match [ICP rule] as sales-ready. ``` ### 21. Enrich a list on a budget Use it on a long list when you want a hard cap on credits. ```text Use www.enrich.chat to find work emails for [file] in row order, keeping a running total of credits_used. Stop at [5,000] credits or at a 402 response, and tell me which row to resume from. ``` ### 22. Recover a bounced email Use it when an address bounces and you suspect a job change. ```text Use www.enrich.chat to look up [LinkedIn URL, or name plus company domain], whose email [address] bounced, with the email and profile fields. Show the current title, company, email and status as returned, and say if they changed jobs. ``` ## Which prompts research an account before outreach? Account research suits ChatGPT, as long as every fact carries a source. Turn on web search, or deep research for a longer report. Prompt 23 calls Enrich.chat, and the rest run in any chat. To see one company's software, try the free [tech stack lookup](https://enrich.chat/tools/tech-stack-lookup/). ### 23. Company snapshot and tech stack Use it for the top of any account plan. A company profile uses 5 credits and the tech stack 5. ```text Use www.enrich.chat to look up [domain] with the company and company tech endpoints. Summarize headcount, industry, state, country and tech stack in five lines, then give two questions the stack raises for a seller of [your category]. ``` ### 24. Priorities from the annual report Use it for public companies, which state goals and risks in writing. ```text Read [company]'s latest annual report or 10-K: [link]. List the three priorities leadership names and any risks tied to [your category]. Quote the sentence behind each point with its section, and add nothing the report does not say. ``` ### 25. Job posts as buying signals Use it to see which teams are growing. ```text Search [company]'s careers page and job posts from the last [60 days]. Group open roles by team, flag posts that mention [tool or problem], and say what the hiring suggests. Link every post you cite, or say you found none. ``` ### 26. Earnings call summary Use it before a meeting with a public company. ```text Summarize [company]'s latest earnings call or investor update: [link]. Pull out what they said about [costs, growth or a named initiative], with a short quote and timestamp for each. End with one line on why [product] matters now, labeled as inference. ``` ### 27. 12-month news timeline Use it to find a timely reason to reach out. ```text Use web search to list [company]'s news from the last [12 months]: funding, launches, leadership changes, acquisitions, layoffs and expansions. Give one line per event with its date and source link. Leave out rumors and anything undated. ``` ### 28. Vendors they already use Use it to spot a tool you would replace or work with. ```text Find public case studies, press releases and partner pages that name [company] as a customer of a [category, such as CRM] product. List each vendor with its source link and date. If you find none, say so. ``` ### 29. Score accounts against your ICP Use it to decide which accounts get a personal touch. ```text Score each account in the attached sheet from 1 to 5 against this ICP: [ICP]. Use only the sheet's columns. Show the score, the two facts behind it and any missing data, sorted from high to low. ``` ## Which prompts write personalized cold outreach? ChatGPT writes good cold outreach from one true fact and a tight brief. Prompts 30 to 34 write first lines, and prompts 35 to 41 write and test messages. None needs a data tool. Each rules out flattery and fake familiarity, because a wrong personal detail hurts more than a plain opener. ### 30. Openers from one sourced fact Use it when your research turns up one clear trigger. ```text Write three opening lines to [name, title] at [company] based on this fact: [fact and source]. Keep each under 20 words, with no compliments and no questions about their week. Tie the fact to [the problem you solve]. ``` ### 31. Role-based openers Use it for big lists you cannot research one by one. ```text I have no personal facts for these contacts. Write one first line per title: [titles]. Name a problem that role owns at a [industry] company of [size]. Keep it about the job, under 20 words. ``` ### 32. Check first lines for errors Use it as a review pass before any personalized send. ```text Check these first lines against the source column: [table]. Flag any that mention personal life, guess at feelings, state a fact not in the source or could feel like surveillance. Rewrite those using only work facts. ``` ### 33. Reply to a public post Use it when the prospect posted, spoke or wrote in public. ```text Here is a public post or interview by [name]: [text or link]. Write one opening line that responds to a specific point, quotes at most five of their words and does not flatter. Then bridge to [your offer] in one sentence. ``` ### 34. Personalize a sheet with a fallback Use it for bulk personalization without fake details. ```text For each row in the attached sheet, write a first line from the trigger column. If the trigger is empty or older than [90 days], use a role-based line and mark the row generic. Add first_line and type columns. ``` ### 35. A short cold email Use it for the main email of a new campaign. ```text Write a cold email to a [title] at a [industry] company with [size] employees, asking for a 15-minute call about [problem]. Keep it under 90 words, plain text, one ask, and only this proof point: [proof]. Add two honest subject lines. ``` ### 36. A three-step sequence Use it for follow-ups. It builds in the opt-out and postal address CAN-SPAM requires. ```text Write a three-email sequence for a [persona] about [problem]: the core email, a day-4 email with one resource [link], and a polite day-9 close-out. Keep each under 80 words. End each with an unsubscribe line and our postal address: [address]. ``` ### 37. A clean A/B test Use it to learn what works, not just to send more. ```text Write two versions of this email that differ only in the [subject line, first line or call to action]: [paste]. Keep everything else the same. State each version's hypothesis and the reply-rate gap that would be worth acting on. ``` ### 38. Cut an email down to size Use it on any draft that runs long. ```text Cut this email to under [70] words without losing the ask: [paste]. Remove jargon, hype and any claim I cannot prove, and aim for a grade 6 to 8 reading level. Then list what you removed. ``` ### 39. LinkedIn connection note Use it for notes you send by hand. [LinkedIn's User Agreement](https://www.linkedin.com/legal/user-agreement) bans bots that send messages. ```text Write a LinkedIn connection note to a [title] at [company] that mentions [shared context, such as an event or post]. Keep it under 40 words, with no pitch. I will send it myself. ``` ### 40. Call opener and voicemail Use it for calls you dial by hand. ```text Write a 20-second cold call opener and a 15-second voicemail for a [title] at a [company type] about [problem]. Use plain spoken English, one question and no fake familiarity. Add a reply for: just send me an email. ``` ### 41. Skeptical buyer review Use it as the last check before you send. ```text Read this cold email as a busy, skeptical [title]: [paste]. List three reasons you would delete it, then rewrite it to fix them without making it longer. ``` ## Which prompts clean CRM data and handle follow-up? Clean CRM data and steady follow-up turn a list into meetings. Prompts 42 to 45 clean records, and prompts 46 to 50 handle replies. Prompts 44, 45 and 50 call Enrich.chat. Ask for proposed changes before anything is written back to the CRM. HubSpot users can see [HubSpot data enrichment](https://enrich.chat/integrations/hubspot/). ### 42. Find duplicate contacts Use it before an import, so you never enrich the same person twice. ```text Find likely duplicates in the attached CRM export. Match on email, then LinkedIn URL, then name plus company domain. For each group, name the record to keep (latest activity) and the fields to merge. Do not delete anything. ``` ### 43. Normalize job titles Use it when free-text titles break your seniority reports. ```text Map each job title in the attached file to one seniority (c_suite, vp, director, manager, senior, entry, owner, founder) and one function (sales, marketing, finance, operations, IT, HR, other). Add both columns and list titles you could not map. ``` ### 44. Re-check emails before sending Use it on any list built more than a few weeks ago. ```text Use www.enrich.chat to re-check the work emails in [campaign file] before we send, by LinkedIn URL where we have one, otherwise name plus company. Pull any row whose email changed, is no longer verified, or was not found. ``` ### 45. Fill missing company fields Use it when accounts have a domain but no size or industry. ```text Use www.enrich.chat to fill empty employees, industry, state and country fields in the attached accounts file with the company endpoint. Fill only blank cells, never overwrite our data, and list rows where the API disagrees with us. ``` ### 46. Sort replies and honor opt-outs Use it daily. CAN-SPAM allows 10 business days for an opt-out, and same day is safer. ```text Sort these replies into interested, not now, wrong person, unsubscribe and out of office: [paste]. List every unsubscribe address to suppress today, note any referral in wrong-person replies, and draft a short answer to each interested one. ``` ### 47. Follow up after no reply Use it for the second touch. ```text Write a follow-up to this email, unanswered after [5] days: [paste]. Add one new piece of value, such as [a resource or idea]. Keep it under 60 words, with no checking-in line and no guilt. ``` ### 48. Recap a discovery call Use it right after a call, while your notes are fresh. ```text From my notes on a call with [name] at [company]: [notes], write a recap email with their goals in their words, the agreed next step and date, and one open question. Keep it under 120 words. Add no commitments I did not make. ``` ### 49. Re-engage closed-lost deals Use it each quarter on deals you lost. ```text Here are closed-lost deals from [6 to 12] months ago with their loss reasons: [attach]. Group them by reason and write one short re-engagement email per group that answers that reason. Skip anyone who asked not to be contacted. ``` ### 50. Follow champions to new jobs Use it to follow past buyers to their new companies. ```text Use www.enrich.chat to look up these past champions by LinkedIn URL with fields=email,profile: [URLs]. For anyone no longer at the company in my CRM, show the new title and company, get the work email and its status, and draft a short note for me to review. ``` ## What rules apply to outreach you write with ChatGPT? The laws that bind a sales rep also bind outreach drafted by ChatGPT, and the sender is responsible for what goes out. This is a summary, not legal advice. - **Email in the US:** [CAN-SPAM](https://www.law.cornell.edu/uscode/text/15/7704) bans false headers and misleading subject lines. Each commercial email needs a clear ad notice unless the person agreed to it, a working opt-out and a valid postal address. Opt-outs must be honored within 10 business days. The [FTC's guide](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) says the law has no exception for business email and lists penalties of up to $53,088 per email. - **Guessed addresses:** Under [15 U.S.C. 7704(b)(1)](https://www.law.cornell.edu/uscode/text/15/7704), sending a rule-breaking email to addresses generated from permutations of names, letters or numbers is an aggravated violation. - **Calls in the US:** [The TCPA](https://www.law.cornell.edu/uscode/text/47/227) bars calls to mobiles made with an autodialer or an artificial or prerecorded voice without prior express consent. On February 8, 2024, the [FCC ruled](https://www.fcc.gov/document/fcc-makes-ai-generated-voices-robocalls-illegal) that AI-generated voices count as artificial. - **UK contacts:** The [ICO](https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/business-to-business-marketing/) says PECR's email consent rule does not cover employees of limited companies, but UK GDPR does, and business contacts can object at any time. - **OpenAI and LinkedIn:** [OpenAI's usage policies](https://openai.com/policies/usage-policies/), effective October 29, 2025, ban spam, deceit and profiling people's private information without authorization. [LinkedIn's User Agreement](https://www.linkedin.com/legal/user-agreement) bans scraping and bots that add contacts or send messages. The prompts on this page stay inside those lines: honest subject lines, an opt-out and address in every sequence email, same-day suppression, work emails only and calls dialed by hand. Enrich.chat never logs in to LinkedIn. Anyone can remove their data at [enrich.chat/opt-out/](https://enrich.chat/opt-out/), and requests are processed within 15 days. You remain responsible for CAN-SPAM, TCPA and GDPR compliance. For the bigger picture, read [AI lead generation in 2026](https://enrich.chat/blog/ai-lead-generation/). ## Frequently asked questions ### Can ChatGPT find email addresses? Not reliably on its own. ChatGPT has no live directory of work emails, so from memory it may return a plausible pattern, such as first.last@company.com, that nobody checked. [Web search](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) can surface addresses printed on public pages, but a page does not prove the mailbox works today. For a checked answer, run the prompt in Codex with Enrich.chat. Each work email comes back verified, catch_all or unknown, and a miss is marked not found and uses 0 credits. ### What are the best ChatGPT prompts for lead generation? The best prompts give ChatGPT your context and tell it what to do when it does not know. Name the buyer by title, industry, company size and location. Attach real data, such as a CRM export. Ask for a source on every fact, and add the line: if you cannot find it, write not found. Keep research and writing in ChatGPT, and send contact lookups to a data tool. The 50 prompts on this page follow those rules. ### Is it legal to use ChatGPT for lead generation? Yes, but the outreach it writes must follow the same laws as any other. In the US, [CAN-SPAM](https://www.law.cornell.edu/uscode/text/15/7704) requires honest headers and subject lines, a working opt-out honored within 10 business days, and a postal address in every commercial email. [The TCPA](https://www.law.cornell.edu/uscode/text/47/227) limits autodialed and artificial-voice calls to mobiles without prior express consent, and the FCC ruled in 2024 that AI voices count as artificial. This is not legal advice. ### Can ChatGPT send emails for me? Sometimes, depending on the app. [Connected apps](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt) let ChatGPT take the actions each app supports, such as creating or updating information, and ChatGPT may ask for your approval first. Whether it can send mail depends on the app and your workspace. Either way, you are still the sender under CAN-SPAM. For cold outreach, most teams draft in ChatGPT and send from their own email or sales tool, where opt-outs and suppression lists already work. ### Can I use ChatGPT for lead generation for free? Yes. OpenAI lists [web search on every ChatGPT plan](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt), including Free, and [Codex on Free and Go](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan) with lower usage limits, as of October 15, 2026. That covers the research and writing prompts here. For contact data, Enrich.chat gives 10 free API lookups a day per IP with no signup. After that, monthly credit plans start at $29, a work email found uses 10 credits and nothing found uses 0. See [pricing](https://enrich.chat/pricing/). ### Is ChatGPT or Claude better for lead generation? Both work with every prompt on this page, because the prompts rely on context and rules, not on one model. The difference is the tools. In ChatGPT, run the Enrich.chat prompts in Codex, which is on every plan. In Claude, run them in Claude Code, which comes with every paid plan, per [Anthropic's pricing page](https://claude.com/pricing). Neither has a contact database, so both need a data tool for emails. Our [Claude for sales guide](https://enrich.chat/blog/claude-for-sales/) covers Claude in depth. ## Sources 1. [Does ChatGPT tell the truth?](https://help.openai.com/en/articles/8313428-does-chatgpt-tell-the-truth) 2. [Why language models hallucinate](https://openai.com/index/why-language-models-hallucinate/) 3. [Why Language Models Hallucinate (paper)](https://arxiv.org/abs/2509.04664) 4. [Reduce hallucinations](https://platform.claude.com/docs/en/test-and-evaluate/strengthen-guardrails/reduce-hallucinations) 5. [Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) 6. [Data analysis with ChatGPT](https://help.openai.com/en/articles/8437071-data-analysis-with-chatgpt) 7. [Deep research in ChatGPT](https://help.openai.com/en/articles/10500283-deep-research-in-chatgpt) 8. [Projects in ChatGPT](https://help.openai.com/en/articles/10169521-projects-in-chatgpt) 9. [ChatGPT is now a partner for your most ambitious work](https://openai.com/index/chatgpt-for-your-most-ambitious-work/) 10. [ChatGPT Work and Codex](https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex) 11. [ChatGPT agent](https://help.openai.com/en/articles/11752874-chatgpt-agent) 12. [Using Codex with your ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan) 13. [Agent approvals and security (Codex)](https://developers.openai.com/codex/agent-approvals-security) 14. [Connected apps in ChatGPT](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt) 15. [Plugins in ChatGPT](https://help.openai.com/en/articles/20001256-plugins-in-chatgpt) 16. [Connecting custom MCP servers in ChatGPT](https://help.openai.com/en/articles/12584461-connecting-custom-mcp-servers-in-chatgpt) 17. [Scheduled tasks in ChatGPT](https://help.openai.com/en/articles/10291617-scheduled-tasks-in-chatgpt) 18. [Creating and editing GPTs](https://help.openai.com/en/articles/8554397-creating-and-editing-gpts) 19. [Configuring actions in GPTs](https://help.openai.com/en/articles/9442513-configuring-actions-in-gpts) 20. [ChatGPT plans and pricing](https://chatgpt.com/pricing) 21. [Usage policies](https://openai.com/policies/usage-policies/) 22. [Claude plans and pricing](https://claude.com/pricing) 23. [15 U.S. Code 7704 (CAN-SPAM requirements)](https://www.law.cornell.edu/uscode/text/15/7704) 24. [CAN-SPAM Act: A Compliance Guide for Business](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) 25. [47 U.S. Code 227 (TCPA)](https://www.law.cornell.edu/uscode/text/47/227) 26. [FCC Makes AI-Generated Voices in Robocalls Illegal](https://www.fcc.gov/document/fcc-makes-ai-generated-voices-robocalls-illegal) 27. [Business-to-business marketing](https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/business-to-business-marketing/) 28. [User Agreement](https://www.linkedin.com/legal/user-agreement) ## Try this prompt Build a verified lead list from ChatGPT ```text Use www.enrich.chat to find 20 operations directors at US manufacturing companies with 200 to 1,000 employees, add work emails with their status, flag catch-all emails and return a CSV. ``` ## Related - [Claude for sales: workflows and prompts](https://enrich.chat/blog/claude-for-sales/) - [AI lead list builder](https://enrich.chat/tools/lead-list-builder/) - [AI lead generation: what actually works](https://enrich.chat/blog/ai-lead-generation/)