How do mailbox providers filter and classify emails?

Ever wonder why the same email lands in the inbox at Gmail but goes straight to spam at Outlook? That's not random. Every mailbox provider (MBP) runs every incoming message through a series of filters, and the order and weighting of those filters varies by provider.

Here's how the filtering process generally works, from first contact to final placement.

Step 1. Authentication checks

The first thing an MBP wants to know is whether you are who you say you are. It checks for SPF, DKIM, and DMARC records on your sending domain. If any of those are missing or broken, the message is immediately suspicious. Some providers will reject it outright. Others will let it through but flag it as untrusted, which hurts placement.

Step 2. Sender reputation

Once authentication passes, the MBP looks at your sending history. This covers your IP address reputation, your domain's track record, how often your past emails triggered spam complaints, and how many bounced. Think of it like a credit score for your sending behavior. A sender with a strong, consistent history gets more benefit of the doubt. A new IP or domain with no history gets scrutinized harder (which is why warming up a new sending IP matters so much).

Step 3. Content analysis

So the MBP scans the email itself. Machine learning models look for patterns that match known spam, phishing attempts, or malware signatures. This includes things like suspicious links, unusual HTML structure, misleading sender names, and certain phrases historically tied to spam. It's not a simple keyword blocklist anymore. Modern filters are trained on billions of messages and they're pretty good at spotting patterns a human wouldn't immediately notice.

Step 4. Engagement signals

This is the layer that surprises most senders. MBPs watch what their users actually do with emails from your domain. If people consistently open your emails, click, and reply, that's a strong positive signal. If they delete without opening, or worse, hit "mark as spam," the MBP notes that and filters your future mail more aggressively. Your reputation is essentially rebuilt email by email, inbox by inbox. (This is also why sending to a stale list can tank your deliverability even if your authentication is perfect.)

But each provider weights these four layers differently. Gmail leans heavily on engagement data. Outlook puts significant weight on its own blocklist and content filters. That difference in weighting is exactly why the same campaign can perform differently across providers.

If you're not sure how your authentication is holding up, you can check your setup with our free Email Header Analyzer. It shows you exactly what an MBP sees when your message arrives.