Deliverability repair

How Inbox Providers Decide What Is Spam: The Signals That Matter Most

How Gmail, Outlook and Yahoo decide what's spam: the signals they weigh, ranked by impact, with Google and Yahoo's current sender rules and what you can control.

Updated 9 min readBy Trevor DavisDeliverability & Sender Reputation

Inbox providers decide what's spam by evaluating sender reputation, authentication records, and recipient engagement signals, not by scanning your subject line for trigger words. The keyword myth persists because it was partially true fifteen years ago. Today, Gmail, Outlook, and Yahoo run layered filtering systems that weigh how recipients behave when they receive your mail far more heavily than what your email says. If you're optimising your copy while ignoring the system underneath, you're solving the wrong problem.

Key takeaways

  • Authentication is the gate: Gmail wants SPF or DKIM from every sender, and SPF, DKIM, DMARC and one-click unsubscribe from anyone sending 5,000 or more messages a day to Gmail.
  • Spam complaints are the strongest signal a reader can send, and both Google and Yahoo say to keep your spam rate below 0.3%, with Google recommending under 0.1%.
  • Engagement decides a lot of placement, so subscribers who've quietly stopped opening drag down delivery for everyone else on your list.
  • Content matters less than people think: structure, broken links and image-only emails cause more trouble than trigger words.

What Signals Do Inbox Providers Actually Evaluate?

Inbox providers collect dozens of signals on every message and score them against models trained on billions of daily sends. But not all signals carry equal weight. Some are binary gates that must pass before anything else is evaluated. Others are behavioural indicators that accumulate over time and shift your reputation gradually. From the reviews we run (the deliverability guide covers the full process), the senders who struggle most are treating all signals as equally important, obsessing over word choice while their DMARC record is misconfigured.

Here's how the major signals rank, based on published provider guidelines and what I've seen in the lists I've worked on. The weights are my judgement, not a published scoring model.

The Spam Signal Ranking Table

| Signal | Weight | What Providers Check | Your Control Level | |---|---|---|---| | Authentication (SPF, DKIM, DMARC) | High | Valid DNS records, cryptographic signatures, policy alignment | Full | | Complaint rate | High | Percentage of recipients clicking "Report Spam" per send | High | | Engagement signals | High | Opens, clicks, replies, deletes-without-reading, moves to spam/inbox | Moderate | | Sender reputation (domain + IP) | High | Historical sending behaviour aggregated across all signals | Moderate | | List quality and bounce rate | Medium-High | Hard bounce percentage, spam trap hits, invalid address ratio | High | | Sending consistency | Medium | Volume patterns, frequency regularity, sudden spikes or drops | High | | Content signals | Medium | Spam-like formatting, link density, image-to-text ratio, URL reputation | High | | Infrastructure and IP reputation | Medium | Shared vs. dedicated IP history, IP warmup status, reverse DNS | Moderate | | Unsubscribe mechanism | Medium | Presence of List-Unsubscribe header, RFC 8058 compliance, processing speed | Full | | Header and technical compliance | Low-Medium | Proper MIME formatting, valid Message-ID, RFC 5322 compliance | Full |

Use this as a diagnostic priority list. If your authentication is failing, no amount of engagement optimisation will help. If your spam rate is at or above 0.3%, content quality is irrelevant. The signals compound in order.


How Does Authentication Affect Spam Filtering?

Authentication is the gate. If your mail fails SPF, DKIM, or DMARC checks, inbox providers treat it as potentially spoofed, and downstream signals never get a chance to help you. Google's sender guidelines require SPF or DKIM from everyone sending to Gmail, and SPF, DKIM and DMARC from anyone sending 5,000 or more messages a day. Yahoo asks the same of its bulk senders, without publishing a volume number. Outlook applies similar evaluation, though its documentation focuses more on IP reputation alongside authentication.

From what I have seen, authentication problems are the most common "invisible" deliverability issue. A DNS change, an ESP migration, or a new third-party tool can quietly break DKIM alignment. The mail still sends. It might even deliver for a while. But the negative signal accumulates, and by the time you notice placement dropping, the reputation damage has been compounding for weeks. The fix is mechanical: verify SPF includes for every authorised sending service, confirm DKIM signing is active and aligned, and publish a DMARC policy starting at p=none with reporting enabled. Our Gmail and Yahoo bulk sender rules guide covers the specifics.


Why Do Complaint Rates Matter More Than Content?

Because a complaint is the strongest signal a recipient can send. When someone clicks "Report Spam," they're telling the provider directly that they didn't want that email. Providers treat this as ground truth. No engagement metric and no content optimisation overrides a pattern of recipients actively flagging your mail.

Google publishes explicit thresholds in its sender guidelines: keep your spam rate below 0.1%, and avoid ever reaching 0.3%. Yahoo's own page says to keep your spam rate below 0.3%, and Outlook watches complaints through its SNDS and JMRP programs. Complaint rates are usually a list problem, not a content problem. People hit "Report Spam" when they don't remember subscribing or when the unsubscribe link is hard to find. The fix almost always lives in your acquisition practices or your list hygiene process, not in your email copy.


How Do Engagement Signals Shape Inbox Placement?

Engagement is where filtering gets behavioural. Inbox providers track what recipients do after your mail is delivered: whether they open it, click links, reply, forward it, delete it immediately, or let it sit unread. Providers aggregate these signals across your entire recipient base to form a picture of how wanted your mail is.

Gmail uses engagement signals to determine placement within its tabbed inbox. A message that generates opens, replies, and clicks is more likely to land in Primary. A message that consistently gets deleted without opening drifts toward Promotions or Spam. The filtering is per-recipient and per-sender, which means your reputation with Gmail is a distribution across every person you send to, not a single number.

This is where the leak happens. A portion of your list has simply stopped engaging. Those recipients aren't complaining or bouncing. They're just ignoring the mail. That pattern of being ignored, across enough recipients, quietly erodes your placement for everyone, including your most engaged readers. Sunsetting inactive subscribers isn't just a list hygiene exercise. It's a deliverability intervention.


What Role Do Content and Structure Actually Play?

Content signals matter, but they sit below authentication, complaints, and engagement in the hierarchy. Providers scan message content for structural patterns rather than individual keywords. Heavy image-to-text ratios, excessive link density, URL shorteners, and links pointing to domains with poor reputations all raise flags. So does HTML that's malformed or overly complex.

The keyword myth is largely outdated. Modern spam filters use machine learning models trained on billions of messages, and those models evaluate content in context. A well-authenticated sender with strong engagement can use the word "free" in a subject line without consequences. A sender with a damaged reputation will struggle to reach the inbox even with perfectly neutral copy. From my experience, content problems that actually trigger filtering tend to be structural: a template that renders poorly, a broken link pointing to a flagged domain, or an email that's entirely an image with no text fallback. These are fixable in an afternoon.


How Do Sending Patterns and Infrastructure Factor In?

Inbox providers watch your sending behaviour over time. Consistent volume on a predictable schedule builds trust. A sudden spike, like sending to a dormant segment you haven't mailed in six months, looks suspicious and often triggers throttling. This is especially true when that spike is accompanied by higher bounce rates or complaints, which it usually is.

IP reputation adds another layer. On a shared IP through your ESP, your reputation is partially shaped by the behaviour of other senders on that infrastructure. A dedicated IP gives you full control but means you own every consequence. Our piece on sender reputation metrics covers the tradeoffs in detail. Technical compliance rounds out the picture. Proper MIME formatting, valid Message-ID headers, working List-Unsubscribe headers with RFC 8058 support, and correct reverse DNS are all baseline expectations. They don't earn you inbox placement on their own, but failing any of them gives providers a reason to downgrade your mail.


How Do All These Signals Interact?

No single signal determines whether your email reaches the inbox. The signals compound. Strong authentication is necessary but not sufficient. Low complaint rates help, but not if your engagement is poor across a large segment of your list. Good engagement protects you from content-level filtering, but it can't overcome a broken DMARC record.

The compounding works in both directions. A sender with passing authentication, low complaints, strong engagement, and consistent sending patterns builds a reputation that acts as a buffer. Minor issues, like an occasional content flag or a small bounce spike, get absorbed because the overall signal profile is healthy. Conversely, a sender with marginal authentication and middling engagement has no buffer. One bad send, one list import with stale addresses, one campaign that generates a complaint spike, and the whole thing tips over.

From what I have seen, the teams that maintain strong deliverability are the ones who monitor the full signal stack, not just the metric that hurt them last time. They check authentication quarterly, review complaint rates after every send, segment by engagement level, and maintain consistent sending volumes. It's not glamorous work. It compounds quietly into reliable inbox placement, which compounds into a functioning owned audience. If your emails are already going to spam and you can't see why, that's what our deliverability repair work is for.


Frequently Asked Questions

What signals do inbox providers use to filter spam?

Inbox providers use a layered set of signals that includes authentication checks (SPF, DKIM, DMARC), spam complaint rates, recipient engagement patterns (opens, clicks, replies, deletes), sender reputation at the domain and IP level, list quality indicators like bounce rates and spam trap hits, sending volume consistency, content structure, and technical header compliance. Authentication and complaint rates carry the most weight. Content keyword scanning, while still a factor, is far less important than most senders assume. The signals are evaluated together and compound over time, so a weakness in one area accelerates damage from weaknesses in others.

How does Gmail decide if an email is spam?

Gmail evaluates incoming mail against a multi-layered filtering system. It starts with authentication. If SPF, DKIM, and DMARC pass, the message moves to reputation and behavioural evaluation. Gmail shows you your spam rate and reputation in Postmaster Tools, and it factors in how recipients interact with your mail at an individual level. High engagement from a specific recipient makes future delivery to that person more likely. High complaint rates across your sends push mail toward spam for everyone. Gmail also uses machine learning models that evaluate content patterns, link reputation, and structural characteristics, but these operate downstream of the reputation and engagement layers.

Can good content still land in spam?

Yes. Content quality doesn't override authentication failures, elevated complaint rates, or poor sender reputation. If your DMARC record is misconfigured, your mail may be rejected or filtered regardless of how well-written it is. If your spam rate reaches Google's 0.3% line, expect much more of your mail to be filtered or rejected, even if every word is perfect. Good content helps by driving engagement, which feeds positive signals back into your reputation. But it's one input among many, and it sits below authentication, complaints, and reputation in the filtering hierarchy.

What could email do next for your business?

If you have an offer, an audience and a part of your email you’d like to improve, let’s talk about where to start.

Let’s talk about your email

Prefer to write? Use the contact form.