Rainmaker.
Trust and Compliance · 5 min read

What "no source, no send" actually means - and why it matters more than spam filters

Citations in cold email aren't a style choice. They're an enforcement mechanism.

When someone asks what makes Rainmaker different, the shortest answer is: "No source, no send."

It sounds like a marketing line. It isn't. It is a description of a function in the product - a function that runs before any email ever reaches your approval queue. Understanding what it actually does, and why we built it, is the clearest way to understand what Rainmaker is.

The problem with AI-generated claims in cold email

Every AI outbound tool generates sentences about your prospects. "Congratulations on your Series A." "I noticed you've been hiring for a new VP of Sales." "Your company recently expanded into the European market."

These sentences are designed to signal research, to create the impression that this email is personal and considered. In a well-executed cold email, they work. In an AI-generated cold email, they introduce a category of risk that most tools do not address: factual error.

AI models do not know things in the way a researcher knows things. They pattern-match from training data and inference. When asked to write about a specific company, the model generates plausible-sounding sentences. Some of them are accurate. Some of them are hallucinated - plausible but false.

If your outbound tool writes "Congratulations on your recent funding round" and the company did not raise a round, you have sent an email that demonstrates the opposite of research. The prospect knows you did not look them up. They know the email was generated. And you signed it.

The spam filter is not the problem. Your reputation is.

What "no source, no send" means in the product

In Rainmaker, every factual claim in an email draft is a SourcedClaim object. Before any email reaches your ApprovalBatch, the system checks that every SourcedClaim has a verifiable source attached - a real, linkable document that substantiates the claim.

If the agent cannot find a source for a claim, one of two things happens:

  1. The claim is rewritten in a way that does not require sourcing (removing the assertion)

  2. The sentence is blocked - it does not appear in the draft

You do not receive an email that contains an unsourced claim. You cannot receive one, because the check happens before the draft is shown to you.

This means that when a draft reaches your approval queue, every factual sentence in it has a live source you can click. The company raised a Series A? There is a link to the announcement. They recently hired a VP of Sales? There is a link to the job posting or the LinkedIn profile. They expanded into the European market?

There is a source.

If you are reviewing a draft and a claim looks wrong, you can click the source and check. The email is auditable before it sends.

Why this protects you (not just your prospects)

The framing around cold email compliance usually focuses on the recipient: their right not to be spammed, their right to opt out, the CAN-SPAM requirements designed to give them recourse.

"No source, no send" is primarily about the sender.

When you send cold email under your name, every claim in that email is a representation you are making. If a claim is false - even inadvertently, even generated by an AI you gave access to your outbound - the misrepresentation is associated with your company, your domain, and your personal name on the signature.

The consequences range from mild (a negative reply) to serious (a prospect who shares the incorrect email publicly, or who files a CAN-SPAM complaint based on deceptive content). In a world where AI-generated outreach is already viewed with skepticism, an email that contains a verifiably wrong fact about the recipient is worse than no email at all.

Source enforcement means that every email Rainmaker drafts can be defended. If a prospect challenges a claim, you have the source. If an email is audited, every sentence has a citation. If you are unsure whether a fact is right, you do not have to trust the AI - you click the link.

How it fits into the broader compliance picture

"No source, no send" is not a compliance certification. It is a product mechanism. But it works alongside a set of other mechanisms that together form what Rainmaker's trust architecture looks like:

  • CAN-SPAM conduct: outbound sends include honest headers, a physical address, and a working unsubscribe mechanism, as required for US commercial email.
  • Opt-out registry: when a prospect opts out, they are added to a cross-tenant registry. They will not receive outreach from any other Rainmaker user, not just the sender who triggered the opt-out.
  • Suppression: standard list suppression plus the cross-tenant registry.
  • Global Privacy Control (GPC) support: the system respects GPC signals.
  • Domain separation: outbound sends from a managed sending address Rainmaker operates for the workspace, or from a mailbox you connect, with bounce and complaint monitoring that pauses campaigns when thresholds are crossed. By default your primary domain does not touch cold outreach.
  • DPA available: a Data Processing Agreement is available inside the product for customers.

What most AI outbound tools offer in this space: an empty compliance page (see Explee's /compliance-faq, which is blank by their own documentation) or no compliance page at all.

The difference matters most when something goes wrong - a prospect files a complaint, a deliverability issue surfaces, a client asks for documentation. Having real mechanisms is different from having a policy statement.

What "blocked" means (it is not a review queue)

This point is worth being precise about.

When we say the email is "blocked," we mean it does not reach your approval queue at all. It is not a queue item flagged for review. It is not a draft in a "needs attention" state. It does not exist as a pending item.

The SourcedClaim check runs before the ApprovalBatch is assembled. An email that fails the check is not shown to you. It is reworked to remove or rewrite the unsourced claim, or it is discarded.

You cannot approve an email that contains an unsourced factual claim, because you will never see one. The check is not advisory. It is structural.

This is the design decision that separates source enforcement from source flagging. Source flagging asks the human to catch the problem. Source enforcement removes the problem before the human sees the draft.

The human approval gate then functions as intended: reviewing the quality and tone of emails that are already factually grounded, not performing a secondary fact-check pass on AI-generated claims.

Questions this post answers

In short.

What does "no source, no send" mean in Rainmaker?

Every factual claim in a Rainmaker email draft is checked against a verifiable source before the draft reaches your approval queue. If no source exists for a claim, the email is blocked - it does not appear in your queue. You cannot approve an email that contains an unsourced factual claim. This is a product function, not a policy: it runs before you see the draft, and you cannot override it.

Can I override the source check?

No. The SourcedClaim check runs before the email draft is assembled into your ApprovalBatch. There is no override path. Emails with unverified factual claims do not reach your approval queue, so there is nothing to override. If you want to include a claim that cannot be sourced, you would need to remove the factual assertion and rewrite it as an observation rather than a claim.

Does adding citations make cold emails less effective?

No. Citations are not appended to the email as footnotes - they are attached to the email as metadata visible in the Rainmaker interface when you review • • • • • the draft. The recipient sees a normal cold email. The sender sees an email where every factual sentence is backed by a clickable source. The experience for the prospect is unchanged; the protection for the sender is real.

What happens to a claim the AI can't source?

One of two things: the claim is rewritten to remove the unverifiable assertion (the agent rephrases the sentence as a question or removes the specificity), or the sentence is dropped from the draft entirely. Either way, the draft that reaches your approval queue does not contain the unsourced claim.

See the receipts on your own domain.

Paste your URL and read the persona and the sourced drafts it writes for your market. No card for the 14-day trial.

You might also read