The spam risk report

Before you send, AgentsMail can read your campaign the way a filter — and a reader — would, and tell you what is likely to hurt it. You run it from the campaign screen, and it looks at three things: the subject, the preview text and the body. It is advice, never a gate. Nothing it says is written into your campaign, and no result has ever stopped a send. You read it, you decide.

What it actually looks at

Two very different kinds of check, and it matters that you can tell them apart.

Counts, done instantly and for free

These are not judgements — they are measurements, and filters weigh them far more reliably than they weigh tone:

Check	Why it matters
HTML weight	An oversized email gets clipped by some inboxes, hiding your unsubscribe link
Number of links	A wall of links reads like a link farm
Number of distinct domains	Many destinations in one email is a classic spam shape
Unsubscribe link present	Its absence is the single most damaging omission
Postal address present	Expected by law in most places, and by filters
Words you repeat	The same term five times or more stops being emphasis and starts being a pattern

An editorial read

This part is an AI analysis of how the email is written, and it returns eight signals in two families. Four are about technique — how it is written:

Signal	What it means
Clickbait subject	The subject promises more than the email delivers
Artificial urgency	A deadline that is not real
Unsupported claims	Promises nothing in the email backs up
Subject / content mismatch	The subject announces one thing, the body is about another

That last one is the one to take most seriously: a reader who feels tricked is the one who clicks "report as spam", and complaints are what actually damages your ability to reach inboxes. Four are about the nature of the content — adult content, gambling, a phishing-like tone, and crypto solicitation. Inboxes treat these harshly however soberly they are written.

How to read the result

The report gives a level — low, moderate, high or critical — for the campaign, and one for each of the three elements. The header takes the worst of them. When it is a single element that pulls the level up, the report names that element: you may see high at the top while the overall distribution still leans moderate, and the line tells you which of the subject, preview text or body is responsible. Every signal is a probability, and it is shown as one. You read "probable, 81%", never "your subject is clickbait". When the analysis is itself unsure, the report says so next to the result rather than hiding it. It is an estimate about how your email reads — treat it as a second opinion, not a verdict.

Three signals raise the level on their own: adult content, gambling and crypto solicitation. Those put a floor at high whatever the rest of the report says, because they affect sending reputation beyond your own account. A phishing-like tone does not — it covers "a verification is needed" and "an unexpected refund", which is half of all legitimate account email.

What it costs

Each run spends one analysis credit from your balance. A run that fails costs nothing — the credit goes back. If your balance is empty, the report says so instead of returning an empty result.

What it cannot tell you

It is not a spam score from a mailbox provider. Nobody outside Gmail knows what Gmail will do with a given message.
It says nothing about your setup. A verified domain, a real sender address and a clean suppression list matter more than any wording, and they are checked elsewhere.
It does not read your audience. Sending relevant email to people who asked for it is what keeps you out of the junk folder; no analysis of the text can compensate for the opposite.

Why a send is refused — the checks that do block a send
How a campaign works
Reading campaign statistics