> 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/can-chatgpt-find-email-addresses/

# Can ChatGPT find email addresses? What it reads, what it guesses and how to check

ChatGPT has no directory of email addresses. It can read a page, guess from a pattern or call a tool, and each route gives a different kind of answer. Here is what each one can and cannot return.

By Enrich.chat Editorial. Updated October 16, 2026.

**Short answer:** ChatGPT does not hold a directory of email addresses. It can quote an address printed on a page it opens, guess one from a company's email pattern, or call a connected tool that returns real records. A guess can look right and still bounce, and [OpenAI's privacy policy](https://openai.com/policies/row-privacy-policy/) says not to rely on the factual accuracy of its output, so check every address before you send.

## Key takeaways

- ChatGPT has three routes to an email address: read a page, guess from a pattern, or call a tool. Only a page or a tool gives you a source to point to.
- In a February 2026 workshop paper, ChatGPT returned the right address in 443 of 500 attempts when given a researcher's web page, and a wrong one 46 times.
- Asked cold, with no page, the same study saw ChatGPT decline 335 of 500 requests and answer 9 wrongly, so a refusal is common but not guaranteed.
- Guesses bounce when the pattern differs, the domain is catch-all or the person left. In January 2026, 20.6% of US wage and salary workers had been with their employer a year or less.
- You can check an address without sending mail: syntax, MX records, then the server's reply. A catch-all domain says yes to everything, so a pass is evidence, not proof.

## Can ChatGPT find email addresses?

ChatGPT does not hold a directory of email addresses, so when you ask for someone's address it can come back with one in only three ways. It can read an address printed on a page it opens. It can guess an address from a company's email pattern. Or it can call a tool or connector that returns real records. Figure 1 shows what each route gives you and what it cannot do.

**Figure 1: Three ways ChatGPT can return an email address.** Three rows compare the routes. Route A, reading a page, returns an address printed on a page it opened and cannot find one nobody published. Route B, guessing a pattern, returns a likely address with no proof it exists and cannot tell you whether the mailbox is real. Route C, calling a tool such as a connected app or an API, returns a record with a status or a not found result and cannot return more than its source holds.

*The three routes by which ChatGPT can return an email address, and the limit of each. Routes drawn from OpenAI's help pages on [web search](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) and [connected apps](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt).*

The root cause is how the model works. [OpenAI's privacy policy](https://openai.com/policies/row-privacy-policy/) says ChatGPT writes answers by predicting the words most likely to come next, that those words may not be the most accurate, and that you should not rely on the factual accuracy of its output. That is fine for a summary. It is a problem for an address, where one wrong character sends mail to the wrong place.

The rest of this page takes each route in turn, then covers address checks and the law. For a table of ChatGPT features by plan, see our guide to [ChatGPT lead generation prompts](https://enrich.chat/blog/chatgpt-lead-generation-prompts/).

## Can ChatGPT read an email address from a web page?

Yes. With search or browsing on, ChatGPT can open a page and quote an address printed there. It cannot find an address that nobody published, and the page itself may be old or wrong.

OpenAI says [web search is available on every ChatGPT plan](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt), including for people who are not signed in. The same page warns that results and citations can be "incomplete, outdated, or incorrect," and it tells you to open a cited source and check that it supports the answer.

When the address is on the page, ChatGPT is good at pulling it out. A [workshop paper from February 2026](https://www.ndss-symposium.org/wp-content/uploads/madweb2026-20.pdf) gave ChatGPT (GPT-5, through its Pro interface) the home page of each of 100 security researchers, including pages that disguise the address, for example by spelling it out in words. Across 500 attempts it returned the right address 443 times and a wrong one 46 times. It declined 11 times.

So about one answer in ten was wrong even with the page in front of it. The paper is small, covers one kind of person and one model setting, and its authors framed the result as a phishing risk. Read it as a rough guide, not a benchmark for sales contacts.

Two limits apply. Many printed addresses are shared inboxes such as press@ or sales@, which do not name a person, as our [email format finder](https://enrich.chat/tools/email-format-finder/) page explains. And reading one page for one person you chose is not the same as sweeping sites for addresses in bulk. Our guide to [scraping and the law](https://enrich.chat/blog/is-linkedin-scraping-legal/) explains where that line sits.

## Can ChatGPT guess an email address from a company's pattern?

Yes, and the guess often looks right, which is the trap. A guess is a likely address, not a fact, and ChatGPT will not always say which one you got.

Many companies build every work address from one rule, such as first.last@company.com, so one real address often reveals the rest. Our guide to [business email formats](https://enrich.chat/blog/email-address-formats/) lists the common rules, and the [email permutator](https://enrich.chat/tools/email-permutator/) lists the variants for a name. ChatGPT does the same job in one step.

Memory is a weaker source than the pattern. A 2022 study by [Huang, Shao and Chang](https://arxiv.org/abs/2205.12628) found that language models do memorize some addresses from their training data. But the models were weak at linking a name to its address, so the chance of pulling out a specific person's address was low.

The 2026 workshop paper shows what that looks like in practice. Asked cold, with no page, ChatGPT declined 335 of 500 requests, answered 156 correctly and answered 9 wrongly. In its error breakdown, the paper puts invented name-format addresses at about 10% of ChatGPT's errors. Those people were security researchers with public home pages, so the result says little about an ordinary sales contact.

OpenAI's own research explains why guesses slip through. Its [September 5, 2025 post on hallucinations](https://openai.com/index/why-language-models-hallucinate/) says models guess when unsure because training and testing reward a lucky guess over saying "I don't know." The post says GPT-5 hallucinates significantly less but still does. It adds that arbitrary, rare facts cannot be predicted from patterns alone. A company's pattern is predictable. The exceptions are not.

You can narrow the risk in the prompt. Ask ChatGPT to label every address as printed on a page, with the link, or as a guess, and to write not found when it has neither. That does not make a guess true, but it stops a guess from passing as data. Our [prompt library](https://enrich.chat/blog/chatgpt-lead-generation-prompts/) uses that rule throughout.

## Why do guessed email addresses bounce?

A guessed address bounces when the pattern is wrong for that person, when the domain accepts everything and the mail fails later, or when the person has left. Figure 2 shows the four common cases.

**Figure 2: Four ways one guessed email address fails.** A fictional guess, dana.lee@example.com, built from a first.last pattern, can fail four ways. The company may use a different pattern for that person, so the server answers 550 5.1.1. The domain may be catch-all, so the server says yes and the mail may bounce later or sit in an unread inbox. The person may have left, so the old address goes dark. Or two people may share a name, or the company may use a second domain.

*Four ways one guessed address fails. The person and address are fictional. The 550 5.1.1 reply is from [Google's SMTP error list](https://knowledge.workspace.google.com/admin/support/troubleshooting/gmail-smtp-errors-and-codes), and the late bounce from [RFC 5321](https://www.rfc-editor.org/rfc/rfc5321.html#section-6.1).*

**Pattern variants.** Companies bend their own rules: nicknames, initials, a number for a second person with the same name, or an older format kept by early hires. Our [email format finder](https://enrich.chat/tools/email-format-finder/) guide documents each exception with sources. When the mailbox does not exist, a Google mail server [answers 550 5.1.1](https://knowledge.workspace.google.com/admin/support/troubleshooting/gmail-smtp-errors-and-codes), which Google defines as an account that does not exist.

**Catch-all domains.** A catch-all server says yes to every address, real or not. [RFC 5321](https://www.rfc-editor.org/rfc/rfc5321.html#section-6.1) says a server that accepts a message and then fails to deliver it must send a failure notice, so the bounce arrives after the send. A guess at such a domain passes the server check. Our [catch-all checker](https://enrich.chat/tools/catch-all-checker/) page explains the setup and the risk.

**Job changes.** A work email lives and dies with the job. The US Bureau of Labor Statistics found that [20.6% of wage and salary workers](https://www.bls.gov/news.release/tenure.nr0.htm) had been with their employer for 12 months or less in January 2026. That count includes new hires, so it overstates job-to-job moves, but it shows how fast an address list goes stale. Our guide to [B2B data decay](https://enrich.chat/blog/b2b-data-decay/) compares the published rates.

**Wrong person or domain.** Two people can share a name, and a company can run a second domain, so a guess can land in the wrong inbox.

A bounce costs more than one failed message. [Google's sender guidelines](https://support.google.com/a/answer/81126) tell senders to automatically unsubscribe recipients who have multiple bounced messages. Our guide to [email bounce rate](https://enrich.chat/blog/email-bounce-rate/) covers the limits that mailbox providers set.

## How can you check an email address without sending mail?

Run three checks in order: the address syntax, the domain's MX records, and the mail server's reply for that recipient. A pass raises your confidence. It is not proof, because a catch-all domain says yes to everything. Figure 3 shows the steps.

**Figure 3: Three checks that need no message, and what each cannot tell you.** A four-step ladder. Step one checks syntax and proves only the shape of the address. Step two looks up MX records and proves the domain takes mail, not that this mailbox exists. Step three reads the mail server's reply for the recipient, but a catch-all domain says yes to every address and a block or timeout gives no answer. Step four gives a verdict of valid, invalid, accept-all or unknown, and a pass is evidence, not proof.

*The checks a verifier runs without sending a message, and the limit of each. Sources: MX lookup and recipient replies from [RFC 5321](https://www.rfc-editor.org/rfc/rfc5321.html#section-5.1), check fields and labels from [Hunter's API documentation](https://hunter.io/api-documentation/v2#email-verifier).*

1. **Syntax.** Check that the text is shaped like name@domain and uses allowed characters. This catches typos only.
2. **MX records.** [RFC 5321, section 5.1](https://www.rfc-editor.org/rfc/rfc5321.html#section-5.1), says a sender first looks up the domain's MX records, the DNS entries that name its mail servers. If there are none, the sender falls back to the domain's own address record. If no route works, the mail cannot be delivered.
3. **The server's reply.** A verifier connects to the mail server, names the recipient with the RCPT TO command and reads the answer. Under [section 3.3 of the same RFC](https://www.rfc-editor.org/rfc/rfc5321.html#section-3.3), a 250 reply means the server accepts the recipient, and 550 means it is known not to be deliverable. The verifier then disconnects before any message goes out.

The last step has limits. A catch-all domain answers 250 for every address. Some servers delay a first-time sender with a temporary 4xx reply, which [RFC 6647 calls greylisting](https://www.rfc-editor.org/rfc/rfc6647.html), and others block checks. Hunter's [email verifier documentation](https://hunter.io/api-documentation/v2#email-verifier) shows the effect. It returns separate fields for the format check, MX records, SMTP connection and SMTP result. It warns that SMTP results on an accept-all server can be false positives, and it reports when a server blocked the check.

An older SMTP command, VRFY, asks a server whether a user exists. [RFC 5321 lets sites turn it off](https://www.rfc-editor.org/rfc/rfc5321.html#section-7.3), and a site that does must answer 252, which is neither yes nor no. In practice it rarely helps.

Because of these limits, a good verifier reports a label, not a verdict. The table shows the labels Hunter uses. Other verifiers use similar words.

| Label | What happened | What to do |
| --- | --- | --- |
| valid | The check found the address deliverable | Include it, and still watch bounces |
| invalid | The check found it not deliverable | Drop it |
| accept_all | The server accepts every address, so this one cannot be confirmed | Hold it unless you have a second signal, such as a colleague's address with the same pattern |
| unknown | The check failed or the server blocked it | Retry later, or hold it |

Do not use a real send as the test, because every bounce counts against your [sender reputation](https://enrich.chat/blog/email-bounce-rate/). For a quick check, use the free [email finder](https://enrich.chat/tools/email-finder/), which takes a name and company and returns an address with a status, or the [catch-all checker](https://enrich.chat/tools/catch-all-checker/) for a domain. To learn who owns an address you already have, read our [work-email lookup guide](https://enrich.chat/blog/who-owns-this-email-address/).

## Can ChatGPT use a tool or connector to get real email data?

Yes. A connected app, an MCP server or an API gives ChatGPT a source that returns records, so it can report what the source says instead of what it predicts. The tool holds the data. ChatGPT only relays it.

OpenAI describes [connected apps](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt) as links between ChatGPT and outside services, and says ChatGPT may ask for your approval before it reads information or takes an action. A [plugin](https://help.openai.com/en/articles/20001256-plugins-in-chatgpt) can bundle apps and skills for one task. Which apps you can use depends on your plan, region, workspace and role, so check your own account.

Email-finder vendors now ship their own routes. Hunter's help center says its [ChatGPT app](https://help.hunter.io/en/articles/15702684-how-to-connect-hunter-to-chatgpt) can find addresses, list a company's contacts and verify an address, and that your Hunter credits still apply. Firecrawl's [guide](https://www.firecrawl.dev/articles/chatgpt/find-work-email) describes a ChatGPT plugin that passes name-and-company lookups to data providers and bills them to a Firecrawl account. Both are vendors describing their own products, read on October 16, 2026, not tests.

Custom code is the other route. The [Model Context Protocol](https://modelcontextprotocol.io/specification/latest/server/tools) lets a server expose tools that a model can call. Its specification says tools are model-controlled and that a person should always be able to deny a tool call. OpenAI's [developer documentation](https://developers.openai.com/api/docs/guides/tools-connectors-mcp) says it asks for approval before sharing data with a remote MCP server by default, and warns that a malicious server can send out sensitive data it sees. Use servers you trust, and keep approvals on until you do. Our guide to the [best MCP servers for sales](https://enrich.chat/blog/best-mcp-servers-for-sales/) compares the options.

GPT Actions are a fading route. OpenAI's [GPT help page](https://help.openai.com/en/articles/8554397-creating-and-editing-gpts) says personal accounts on Free, Go, Plus and Pro cannot create new GPTs, and that OpenAI plans to retire custom GPTs in favor of plugins. Do not build a new workflow on them.

Whatever route you pick, ask the tool for three things: a status on each field, an honest not found, and no guessed fill-in. Without those, you have only moved the guess to a different place.

## What does OpenAI say about browsing, connectors and personal data?

OpenAI's pages say four things that matter here: its output may be wrong, search results can be out of date, apps may see what you type, and its rules bar collecting people's private information without authorization. The table lists each page and the copy we read.

| OpenAI page | What it says | Copy we read |
| --- | --- | --- |
| [Usage policies](https://openai.com/policies/usage-policies), effective October 29, 2025 | Bars attempts to aggregate, monitor, profile or distribute individuals' private or sensitive information without authorization. Also bars spam and impersonation. | Archived October 15, 2026 |
| [Privacy policy](https://openai.com/policies/row-privacy-policy/), updated February 6, 2026 | ChatGPT predicts likely words, so do not rely on its factual accuracy. People can ask OpenAI to correct or remove inaccurate information about them. | Archived October 12, 2026 |
| [Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) | Search may rewrite your question into queries sent to search providers. Results and citations can be incomplete, outdated or incorrect. | Archived October 15, 2026 |
| [Connected apps in ChatGPT](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt) | An enabled app may see relevant context from your chats. Data shared with an app follows that app's terms. On Free, Plus, Go and Pro, OpenAI may train on app information if "Improve the model for everyone" is on. Business, Enterprise and Edu: not by default. | Archived October 4, 2026 |
| [MCP in the Responses API](https://developers.openai.com/api/docs/guides/tools-connectors-mcp) | Asks for approval by default before data goes to a remote MCP server. Warns that a malicious server can send out sensitive data. | Read October 16, 2026 |

These are OpenAI's own rules for its services, not a ruling on whether your use is lawful. The practical lesson is to type only what the task needs. A name you paste into ChatGPT can reach an app or a search provider, and OpenAI's pages describe both.

## Is it legal to use ChatGPT to find and email people?

Using ChatGPT does not change the law. The rules follow the email you send and the personal data you hold. This is not legal advice.

**GDPR in the EU and UK.** The European Commission defines [personal data](https://commission.europa.eu/law/law-topic/data-protection/data-protection-explained_en) as information about an identifiable person and lists an address like name.surname@company.com as an example. The UK regulator says the [same holds for business contact details](https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/business-to-business-marketing/): UK GDPR applies even when the person acts for a business, you need a lawful basis, and in most cases you must give privacy information within one month if the data came from another source. The ICO marks that page as under review. Its [accuracy principle](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/data-protection-principles/a-guide-to-the-data-protection-principles/accuracy/) also expects reasonable steps to keep personal data correct, so label a guessed address as a guess and do not store it as fact.

**CAN-SPAM in the US.** The FTC says the law [makes no exception for business-to-business email](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business). Each message needs honest headers and subject lines, an ad label, your postal address and an opt-out you honor within 10 business days. The FTC lists penalties of up to $53,088 per email. The statute also makes an already unlawful message an [aggravated violation](https://www.law.cornell.edu/uscode/text/15/7704) if the sender knew the addresses came from a tool that combines names, letters or numbers into numerous permutations. So never send every variant of a guess.

For more depth, see our guides to [cold email data](https://enrich.chat/blog/cold-email/), [buying email lists](https://enrich.chat/blog/buying-email-lists/) and [calling and texting mobiles](https://enrich.chat/blog/tcpa-b2b-cold-calling/), which a different law covers. No data tool makes a send compliant. You decide whether you may contact each person.

## How does an agent look up a work email instead of guessing?

An agent looks up an email by sending a name and a company website to a contact API and reporting exactly what comes back, including a miss. With Enrich.chat that is one GET request or one MCP tool call, and a lookup that finds nothing uses 0 credits.

1. **Start with a name and a website.** Use Dana Lee at example.com, a fictional person on a reserved test domain. Enrich.chat takes a company website such as example.com, not a company name.
2. **Call the API.** The agent sends `GET /v1/email?name=Dana%20Lee&domain=example.com`, or calls the `find_work_email` tool on the MCP server at `https://enrich.chat/mcp`. The free tier needs no key: 10 lookups a day per IP address, no signup.
3. **Read the result.** A match returns the address, a status and `credits_used`. The status is `verified` (Enrich.chat's records mark the mailbox as verified), `catch_all` (the domain accepts every address, so the mailbox cannot be confirmed) or `unknown`. Treat every contact as a strong lead, not a guarantee.
4. **Branch on the result.** The table below shows what the agent writes for each case.
5. **Report and stop cleanly.** The agent returns a table of name, company, email and status, plus the total `credits_used`. On HTTP 402 it stops and tells you. On 429 it waits for `retry_after` seconds.

Here is the shape of a match on a catch-all domain. The person and address are fictional.

```
GET https://enrich.chat/v1/email?name=Dana%20Lee&domain=example.com

{
  "found": true,
  "work_email": {"value": "dana.lee@example.com", "status": "catch_all"},
  "pattern": "first.last@example.com",
  "credits_used": 10,
  "free_lookups_remaining": 9
}
```

And the shape of a miss. Nothing is used, and the agent must not fill the gap.

```
{"found": false}
```

| Result | What the agent writes | What you do |
| --- | --- | --- |
| found, status verified | The address, labelled verified | Add it to your list. You still decide whether you may email the person. |
| found, status catch_all | The address, labelled risky because the domain accepts every address | Hold it, or send a small test batch after a second signal. |
| found, status unknown | The address, labelled unconfirmed | Hold it. |
| found is false | Not found, with nothing invented | Try a LinkedIn URL if you have one, or skip the row. |

On a miss, the email endpoint can also return a field named `guess`: the company's most common format applied to the name. The [Enrich.chat agents file](https://enrich.chat/agents.md) tells agents it is not data, so the agent should call it an unconfirmed guess or leave it out. That label is the difference between a guess and a result. A LinkedIn URL works as a lookup key too, and Enrich.chat never logs in to LinkedIn.

Paste this prompt into an agent that can call the API or the MCP server:

```text
Use www.enrich.chat to find the work email for [full name] at [company website]. Show the address and its status exactly as returned. If nothing is found, write not found and do not guess an address from a pattern.
```

A work email found uses 10 credits, and nothing found uses 0. A list of 200 names needs at most 2,000 credits, which fits the Growth plan at $29 a month. See [pricing](https://enrich.chat/pricing/) for every plan.

To connect it, add the MCP server at `https://enrich.chat/mcp` to a client that supports custom MCP servers, or let an agent that can send web requests read [enrich.chat/agents.md](https://enrich.chat/agents.md). The [install page](https://enrich.chat/install/) has steps for ChatGPT, Claude, Cursor and Codex. OpenAI sets which ChatGPT plans can add a custom connector, so check yours first. Our [guide for AI agents](https://enrich.chat/agents/) and the [email finder API](https://enrich.chat/api/email-finder/) page cover the endpoints. Other email finders can fill the same role, and our [comparison of email finder tools](https://enrich.chat/blog/best-email-finder-tools/) covers them.

## Frequently asked questions

### Can ChatGPT find someone's email address?

Not reliably on its own. ChatGPT has no directory of email addresses. It can quote an address printed on a page it opens, guess one from a company's email pattern, or call a connected tool that returns real records. Only a printed address or a tool result gives you a source. A guess can look right and still bounce, so check it before you send. Our guide to [finding anyone's work email](https://enrich.chat/blog/how-to-find-someones-email/) covers the other methods.

### Can ChatGPT guess an email address from a company's format?

Yes. Give it a name and a company website and it can apply a common pattern such as first.last@ and return a plausible address. That is a guess, not a lookup. The address may belong to someone else, follow a different pattern or sit on a catch-all domain that accepts everything. Ask ChatGPT to label each address as printed on a page or a guess, then check the guess. See [business email formats](https://enrich.chat/blog/email-address-formats/).

### Why does ChatGPT sometimes refuse to give an email address?

OpenAI's [usage policies](https://openai.com/policies/usage-policies) bar attempts to aggregate, monitor, profile or distribute individuals' private or sensitive information without authorization. In a 2026 workshop study, ChatGPT declined 335 of 500 cold requests for researchers' addresses, and 11 of 500 when it was given the researcher's web page. The study used one model setting, so treat refusals as a guardrail you cannot count on.

### Can ChatGPT search my own inbox for an email address?

It can if you connect an app for your mail account. OpenAI says [connected apps](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt) let ChatGPT find and read information in services you already use, and that it may ask for approval before reading or acting. It sees only what your account permissions allow. That is a different job from finding a stranger's address, which no inbox can answer. Review each app's permissions in Settings, then Plugins.

### How do I check an email address without sending a message?

Check the syntax, confirm the domain has MX records, then ask the mail server whether it accepts that recipient and disconnect before any message is sent. A verifier runs these steps for you. The result is evidence, not proof: catch-all domains accept every address and some servers block checks. Try the free [email finder](https://enrich.chat/tools/email-finder/) or the [catch-all checker](https://enrich.chat/tools/catch-all-checker/), and email only addresses that pass.

### Is it legal to use ChatGPT to find email addresses and email people?

The law follows what you do with the address, not the tool. In the US, [CAN-SPAM](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) covers business email and needs honest headers, a postal address and an opt-out honored within 10 business days. In the EU and UK, an address that names a person can be personal data, so you need a lawful basis and must say where you got it. This is not legal advice.

## Sources

1. [Privacy policy (EEA, UK and Switzerland)](https://openai.com/policies/row-privacy-policy/)
2. [Usage policies](https://openai.com/policies/usage-policies)
3. [Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt)
4. [Connected apps in ChatGPT](https://help.openai.com/en/articles/11487775-connected-apps-in-chatgpt)
5. [Plugins in ChatGPT](https://help.openai.com/en/articles/20001256-plugins-in-chatgpt)
6. [Creating and editing GPTs](https://help.openai.com/en/articles/8554397-creating-and-editing-gpts)
7. [Why language models hallucinate](https://openai.com/index/why-language-models-hallucinate/)
8. [MCP tool in the Responses API: risks and approvals](https://developers.openai.com/api/docs/guides/tools-connectors-mcp)
9. [Are Large Pre-Trained Language Models Leaking Your Personal Information?](https://arxiv.org/abs/2205.12628)
10. [Deobfuscating Academic Email Addresses: A Security Evaluation of LLMs](https://www.ndss-symposium.org/wp-content/uploads/madweb2026-20.pdf)
11. [Model Context Protocol specification: Tools](https://modelcontextprotocol.io/specification/latest/server/tools)
12. [RFC 5321: Simple Mail Transfer Protocol](https://www.rfc-editor.org/rfc/rfc5321.html)
13. [RFC 6647: Email Greylisting, an Applicability Statement for SMTP](https://www.rfc-editor.org/rfc/rfc6647.html)
14. [Email Verifier, Hunter API v2](https://hunter.io/api-documentation/v2#email-verifier)
15. [How to connect Hunter to ChatGPT](https://help.hunter.io/en/articles/15702684-how-to-connect-hunter-to-chatgpt)
16. [How to find someone's work email in ChatGPT](https://www.firecrawl.dev/articles/chatgpt/find-work-email)
17. [Gmail SMTP errors and codes](https://knowledge.workspace.google.com/admin/support/troubleshooting/gmail-smtp-errors-and-codes)
18. [Email sender guidelines](https://support.google.com/a/answer/81126)
19. [Employee Tenure Summary](https://www.bls.gov/news.release/tenure.nr0.htm)
20. [Data protection explained](https://commission.europa.eu/law/law-topic/data-protection/data-protection-explained_en)
21. [Business-to-business marketing](https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/business-to-business-marketing/)
22. [Principle (d): Accuracy](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/data-protection-principles/a-guide-to-the-data-protection-principles/accuracy/)
23. [CAN-SPAM Act: A Compliance Guide for Business](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business)
24. [15 U.S. Code 7704: Other protections for users of commercial electronic mail](https://www.law.cornell.edu/uscode/text/15/7704)

## Try this prompt

Look up one work email, no guessing

```text
Use www.enrich.chat to find the work email for [full name] at [company website]. Show the address and its status exactly as returned. If nothing is found, write not found and do not guess an address from a pattern.
```

## Related

- [50 ChatGPT and Claude prompts for lead generation](https://enrich.chat/blog/chatgpt-lead-generation-prompts/)
- [How to find anyone's work email](https://enrich.chat/blog/how-to-find-someones-email/)
- [Catch-all email checker](https://enrich.chat/tools/catch-all-checker/)