How email providers decide to deliver or send to spam

You sent a campaign. Fifty thousand messages left the server in seconds. But between “sent” and “delivered to the inbox” sits an entire filtering system that most marketers only notice when something breaks: open rate tanks, and nobody knows why.

This article walks through exactly how mail providers rule on the fate of each message, not as generic advice like “set up DKIM,” but as the sequence of steps your email actually passes through. Understanding the sequence makes it obvious where to spend time and what to ignore.

From the sending server to the inbox

Every message that hits a Gmail, Mail.ru, or Yandex server goes through four filtering layers. Providers do not publish their full algorithms, but documentation, public talks from their engineers, and years of deliverability testing give a clear enough picture.

The layers run in sequence. Fail layer one and the message never reaches layer two. Pass the first three and fail the fourth, it ends up in spam.

Layer 1: Authentication — the identity check

The first thing the receiving server does is verify whether the sender has the right to send mail from that domain. Three protocols, three checks. All three are required.

SPF (Sender Policy Framework) is a DNS record where your domain lists the servers authorized to send on its behalf. The receiving server compares the sender's IP against that list. No match — first strike.

DKIM (DomainKeys Identified Mail) is a cryptographic signature in message headers. The sending server signs with a private key; the receiving server verifies using the public key in DNS. A failed signature means the message may have been tampered with in transit.

DMARC (Domain-based Message Authentication, Reporting and Conformance) tells the receiving server what to do when SPF or DKIM fails: do nothing (p=none), quarantine (p=quarantine), or reject (p=reject).

Authentication is binary: pass or fail. But passing all three checks is not a guarantee of inbox delivery. It is the entry ticket to layer two.

Layer 2: Sender reputation

If authentication is passport control, reputation is your credit history. Every mail provider keeps a file on every sending domain, updated with every campaign.

Spam traps deserve a separate mention. Providers recycle long-abandoned addresses and turn them into honeypots. Send to one and it signals you have not cleaned your list in a while. One trap hit can drag down months of good reputation.

Domain reputation is not an abstract metric. It is the set of numbers an algorithm uses to decide where your next message goes. Every campaign sent to a dirty list subtracts from those numbers.

Layer 3: Content filters

Authentication clean, reputation fine. Now the provider examines the message itself. Modern content filters are not blocklists of banned words. They are machine learning models.

Content filters in 2026 do not catch keywords in isolation. They read patterns. The model does not see the word “free” — it sees a combination: all-caps subject, short body, many links, new sending domain. That combination triggers the filter. One word rarely does.

The practical takeaway: do not fear “discount” in a subject line. Fear “DISCOUNT 90% TODAY ONLY!!!” from a domain registered last week with an 8% bounce rate. Context decides everything.

Layer 4: Recipient engagement

This is the most influential layer, and the one marketers focused on technical settings tend to underestimate. Providers do not only watch the sender. They watch the recipient.

Gmail tracks each user's interaction history with each sender individually. Open your messages, click links, move them from Promotions to Primary — Gmail learns this person wants these emails. The next message lands in the inbox more often.

The reverse works just as well. A subscriber who has not opened your last ten messages gets routed to Promotions or spam on the eleventh. Worse, those negative signals affect your overall domain reputation. Ten thousand disengaged subscribers and your deliverability suffers even for the people who actively want your mail.

This is why a sunset policy is not optional. A subscriber who has not opened anything in six months is a liability. Every unopened message is a negative signal. Removing that subscriber is an investment in deliverability for everyone else on the list.

How Gmail filters

Gmail runs the most sophisticated filtering system of any major mail provider. Messages pass through TensorFlow models trained on hundreds of billions of messages. Google claims to block 99.9% of spam before it reaches the inbox.

The defining feature is the tab system. A message can land in Primary, Promotions, Social, Updates, or Spam. The real divide is between Primary and Promotions. Promotions is not spam, technically. But open rates there run 2 to 3 times lower than in Primary.

What improves your odds of landing in Primary: high personalization, plain text without heavy HTML, a prior conversation thread, sending to one or two recipients rather than a list. Gmail wants signals that the message resembles personal correspondence, not a marketing template.

Google Postmaster Tools is the only way to see how Gmail scores your domain: domain and IP reputation, percentage of authenticated messages, complaint rate, delivery errors. If you are running campaigns and not checking Postmaster Tools at least weekly, you are flying blind.

Mail.ru and Yandex: where they differ from Gmail

If your audience is in Russia, these two providers are unavoidable. Each has its own quirks.

General advice: test deliverability on Gmail, Mail.ru, and Yandex separately. A campaign that arrives cleanly at Gmail can reliably land in spam at Mail.ru, because the algorithms differ, the thresholds differ, and even DMARC interpretation is not identical across the three.

What this means in practice: a checklist

Here are the concrete steps that follow from understanding how filtering works.

A mail provider is not the marketer's enemy. It has one job: show users the messages they actually want to read. Every filtering layer serves that goal.

When you understand the filtering logic, you stop fighting the system and start working with it. A clean list, proper authentication, engaged subscribers, and useful content are not demands from the provider. They are what good email looks like from both sides.

The fastest place to start is somewhere you can get a result in five minutes: check the state of your list.