What every B2B cold email needs, in the US and Canada
A cold pitch to a business is a commercial email. In the US you can send the first one without permission. In Canada you need consent before it. In both, every message must identify you, give contact details and offer a working opt-out. AI wording changes none of this.
"The law makes no exception for business-to-business email." (FTC: CAN-SPAM Act, a compliance guide for business)
| Rule | US (CAN-SPAM) | Canada (CASL) |
|---|---|---|
| Permission before the first email | Not needed | Needed; for a stranger, usually a work address they published (next section) |
| Honest sender and subject | Required: the subject "must accurately reflect the content of the message" | False or misleading sender or subject information is reviewable conduct (Competition Act, s. 74.011) |
| Say it's an ad | Required, "clearly and conspicuously"; you choose the wording | Not one of CASL's required items (s. 6(2)) |
| Your address | A valid physical postal address; a registered P.O. box works | A mailing address, plus a phone number, email or website |
| Opt-out | A clear way to opt out, working at least 30 days after you send | Free, by reply or link, valid at least 60 days (s. 11) |
| Honoring it | Within 10 business days | Within 10 business days (s. 11(3)) |
| Maximum penalty | Up to $53,088 per email | C$10 million per violation for a business (s. 20(4)) |
US column: FTC compliance guide. Canada column: Canada's Anti-Spam Legislation: the law on consent before sending commercial emails and texts to people in Canada. More and ISED's "Getting consent to send email". A reply address counts as an opt-out: the FTC accepts "a return email address", so a closing line like "Reply 'no thanks' and I won't write again" works, if you honor it.
Using a tool or an agency doesn't move the duty. The FTC says you "can't contract away your legal responsibility". CASL covers anyone who sends, or causes or permits a message to be sent (s. 6(1)).
Can I cold email a business in Canada?
Yes, if the person published their address for work and your pitch fits their job. CASL treats that as implied consent when the person:
"has conspicuously published, or has caused to be conspicuously published, the electronic address to which the message is sent" (CASL, s. 10(9)(b))
The same paragraph adds two conditions: no statement next to the address saying they don't want unsolicited messages, and a message "relevant to the person's business, role, functions or duties". An office manager's email on a clinic's contact page, pitched IT support for the clinic, fits. The same address pitched a personal loan doesn't.
Two traps:
- The B2B exclusion isn't for strangers. The regulations exempt messages between organizations only "if the organizations have a relationship" (SOR/2013-221, s. 3(a)(ii)).
- You carry the proof. "A person who alleges that they have consent ... has the onus of proving it" (CASL, s. 13). Save the page and date where you found each address. A bought list rarely shows where each person published theirs.
A referral opens one door. If someone with a relationship to both of you refers you, the regulations allow one first email without consent, if it names the referrer and says it follows their referral (SOR/2013-221, s. 4(1)).
CASL applies when "a computer system located in Canada is used to send or access" the message (s. 12(1)), so a US firm emailing a Toronto office is covered. Message contents and consent periods: CASL consent.
How to set up AI cold outreach, step by step
- Pick a short list you can truly help. One kind of business, one role (owner, office manager, operations lead), a few dozen companies. AI can't rescue a list of people who don't need you.
- Find addresses the clean way. The company's contact or team page, the person's own bio, a business card. Note where and when. Skip address scrapers: the FTC lists "harvesting email addresses or generating them through a dictionary attack" among acts that can bring criminal penalties. Don't point bots at LinkedIn either; its User Agreement bans "bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages".
- Send from your own domain, set up properly. Turn on SPF, DKIM and DMARC (your email host's settings, or your IT person). Use a real reply address someone reads. Add a footer: business name, postal address, one-line opt-out.
- Research with AI, one prospect at a time. Paste their public About page or a recent post. Ask for two facts that matter to your service, each with the sentence it came from. Then open the source and check.
- Draft with AI, edit by hand. A few sentences: why them, what you'd fix, one question. Subject lines that describe the email; no fake "Re:". Never let AI state results you can't prove.
- Send at the pace you can answer. From your normal mailbox or a sending tool, a small batch a day. Read the next section before buying "unlimited inboxes".
- Follow up twice, then stop. Each follow-up is another commercial email, with the same footer. Stop at once on any "no" or opt-out, and in all cases within 10 business days.
- Count replies and meetings, not opens. Log each send, reply and opt-out in a sheet or CRM, so no one hears from you after saying no. Apple's Mail Privacy Protection "prevents senders from seeing if you've opened the email message they sent you" (Apple Support), so open rates mislead.
Why step 4 says check: AI tools state wrong things with confidence.
"Claude can occasionally produce responses that are incorrect or misleading. This is known as 'hallucinating' information" (Claude Help Center)
Treat facts from any AI tool the same way. A made-up "congrats on your new location" costs you the prospect; a made-up claim about your own results can make the email misleading.
Will my cold emails go to spam?
They're more likely to if people mark them as spam or your domain isn't set up. Gmail and Outlook.com publish minimum rules for their personal inboxes (Google Workspace Admin Help: Email sender guidelines; Microsoft: Outlook's requirements for high-volume senders). Most B2B prospects use company mail, which runs its own filters.
| Provider | Applies to | Key requirements |
|---|---|---|
| Gmail | Anyone sending to personal Gmail accounts | SPF or DKIM; spam rate reported in Postmaster Tools below 0.3% |
| Gmail | More than 5,000 messages a day to personal Gmail | SPF, DKIM and DMARC; one-click unsubscribe on marketing mail |
| Outlook.com, Hotmail, Live | Domains sending over 5,000 a day | SPF and DKIM pass; DMARC at least p=none. Since May 5, 2025, failing mail goes to Junk or is rejected |
"If messages from your domain are frequently reported as spam, future messages from you are more likely to be marked as spam." (Google Workspace Admin Help: Email sender guidelines)
Varying the wording doesn't change that. Microsoft's recommended practices for senders include: "Use accurate subject lines, avoid deceptive headers, and ensure your recipients have consented to receive your messages."
Your mailbox host has its own rules. Google Workspace's Acceptable Use Policy bars using it "to generate, distribute, publish or facilitate unsolicited mass email". Microsoft 365 caps each user at 10,000 recipients a day and says "Exchange Online customers who need to send legitimate bulk commercial email (for example, customer newsletters) should use third-party providers that specialize in these services" (Microsoft Learn: Exchange Online limits). Spreading thousands of cold emails across many look-alike inboxes is still a mass send.
Example
Illustrative; names and numbers invented.
A US IT support firm writing to dental offices
The owner lists 40 practices within 30 miles and takes office managers' emails from practice websites. AI reads each site and notes, for example, "two locations, online booking". He checks each note, edits each draft, and sends 8 a day from his own domain with his address and a "reply 'no thanks'" opt-out line. Two follow-ups, a week apart. Three managers reply; two say no and get nothing more.
A Toronto bookkeeper writing to contractors
She uses addresses contractors published on their own websites, skips any site that says "no solicitations", and pitches bookkeeping to the owner named as the contact. She saves each page and date. A data vendor offers 5,000 contractor emails for C$400; she can't show where any of them were published, so she passes.
FAQ
Is cold emailing legal?
In the US, yes, if every email follows CAN-SPAM. In Canada, only with consent, usually a published work address and a pitch for their role.
Can AI write my cold emails?
Yes, as a first draft. You're responsible for every claim in it, and AI can invent facts, so check each one before sending.
How many cold emails can I send a day?
No US or Canadian law sets a number. Microsoft 365 caps a user at 10,000 recipients a day. Send only what you can answer.
Can I buy a B2B email list?
In the US, it isn't banned, but every email must follow CAN-SPAM. In Canada, you must prove consent, which a bought list rarely shows.
Different for you?
- Emailing or texting Canadians who already contacted you: CASL consent.
- Emailing past customers, not strangers: email and text marketing.
- Calling or texting bought leads: TCPA consent.
- A prospect replied and you need to answer fast: speed to lead.
- You want every reply answered and booked for you: Conversion Agent.