How has AI and filtering changed email in the 2020s?

In the 2000s, Gmail and other providers relied heavily on Bayesian filtering, which learned from word patterns and content signals. Spammers fought back by hiding blocks of innocent text at the bottom of spam emails to poison the filters.

By the 2020s, mailbox providers shifted to engagement-based filtering. Instead of asking "Is this message spam?", they now ask "Does this specific user usually open emails from this sender?" Gmail, Outlook, and Yahoo all track whether you typically read newsletters from a particular domain, click links, reply, or ignore them completely.

This change matters because content alone no longer decides inbox placement. You can write a perfect subject line with zero spam triggers and still land in the spam folder if your recipients don't engage. Mailbox providers watch individual subscriber behavior: if someone hasn't opened your last 20 emails, your next one probably goes to spam (or gets graylisted) for that person, even if it lands in the inbox for everyone else.

The shift also introduced per-user filtering. Two subscribers on the same ESP can see your email in completely different folders based on their individual history with your brand. This is why segmenting by engagement level became critical in the 2020s. Sending to everyone hurts deliverability for engaged subscribers when non-openers drag down your overall engagement metrics.

What changed for senders: list hygiene went from optional to mandatory. Sunset policies (automatically removing non-openers after a set period) became standard practice. Targeting campaigns based on open and click history became table stakes. And warming up new sending domains or IPs became far more important because reputation now builds through sustained engagement, not just avoiding spam complaints.

Machine learning models now analyze patterns mailbox providers never looked at before: how quickly users delete unopened emails, whether they search for emails from you later, whether they move your messages to specific folders, and even whether they pause scrolling when your subject line appears in their inbox preview. The filters got smarter, which means senders have to be smarter too.

If you're seeing placement issues despite following all the technical rules (SPF, DKIM, DMARC set up correctly), engagement is almost always the culprit. Check your open rates by subscriber segment. Anyone who hasn't opened in 90 days should be on a re-engagement path or suppressed entirely. Worth running your list through a validation service if it's been sitting untouched for a while (we clean them if you're stuck ;)). The 2020s reward senders who respect their subscribers' actual behavior, not senders who blast everyone and hope for the best.