Most senders think spam filtering works like a keyword blocklist. Avoid certain words and your email lands in the inbox. This model has not been accurate for at least a decade.
Modern inbox providers Gmail, Outlook, Yahoo, Apple Mail use multi-layered evaluation systems. They assess sender identity, sending history, recipient engagement, content characteristics, and the real-time behaviour of the specific recipient. Each of these factors contributes to a placement decision.
Understanding how this evaluation actually works changes everything about how you approach email deliverability. It reveals why content changes alone rarely fix inbox placement problems, and why reputation management matters more than subject line optimisation.
The Shift from Keyword Filtering to Behavioural Analysis
Early spam filters in the 1990s and early 2000s were essentially keyword lists. They checked email for words like “FREE,” “GUARANTEED,” and “$$$” and assigned higher spam probability scores when these words appeared.
Spammers adapted quickly. They misspelled words, used images instead of text, and avoided flagged terminology. The keyword filters became inadequate almost as soon as they were deployed.
Inbox providers responded with increasingly sophisticated approaches. By the 2010s, Gmail and Outlook were using machine learning models trained on billions of emails. These models did not check for specific words; they looked at patterns across hundreds of signals simultaneously.
Today, inbox placement decisions at Gmail are primarily determined by sender reputation and recipient engagement data. Content plays a secondary role. A well-reputed domain with strong engagement history can send emails with promotional language and still land in the inbox. A poor-reputation domain sending the most carefully crafted plain-text email may still go to spam.
This is the fundamental reality of modern email filtering that most sender-facing advice fails to address.
Layer 1: Sender Identity Verification
The first layer of evaluation is identity. Before anything else, the receiving server asks: can I verify who sent this email?
Authentication records answer this question. SPF (Sender Policy Framework) confirms the sending server is authorised to send on behalf of the stated domain. DKIM (DomainKeys Identified Mail) confirms the email was signed by the domain owner and was not modified in transit. DMARC ties these together and specifies what to do with email that fails either check.
Why Authentication Failure Is So Damaging
If authentication fails, the inbox provider cannot confirm the email is genuinely from the stated sender. This does not automatically mean the email goes to spam. But it removes a trust signal that would otherwise reduce filtering.
For a new domain with no reputation history, authentication failure is particularly harmful. The only way to distinguish legitimate mail from a new domain versus spam from the same domain is the authentication signal. Without it, conservative filtering applies.
For established domains, authentication failure is damaging but survivable in the short term. However, sustained authentication failures erode reputation over time because they generate poor signals in DMARC aggregate reports that inbox providers accumulate.
Layer 2: IP and Domain Reputation
Once identity is verified, the inbox provider checks reputation. This is the most influential factor in placement decisions.
Reputation operates at two levels simultaneously.
IP reputation reflects the sending history of the specific IP address your email comes from. If you send from a shared IP pool, your reputation is partly determined by other senders on the same pool. If you send from a dedicated IP, your reputation is entirely your own.
Domain reputation reflects the history of the domain in your From address and DKIM signature. At Gmail, domain reputation is the primary signal. Google Postmaster Tools shows this as the Domain Reputation metric — High, Medium, Low, or Bad.
How Reputation Is Calculated
Reputation is not a fixed score. It is a continuously updated model based on the signals generated by your recent sends. Positive signals include opens, clicks, replies, and moves from spam to inbox. Negative signals include spam reports, hard bounces, moves to spam, and deletion without opening.
The ratio of positive to negative signals, weighted by recency, determines reputation. A domain that consistently generates 30% click rates among engaged subscribers has high positive signal density. A domain that sends to a large unverified list and generates 15% bounces and 2% spam reports has high negative signal density.
Reputation is also personalised in Gmail. Each Gmail user has their own engagement history with your domain. Subscribers who have clicked your emails many times will see your emails in Primary even when your overall domain reputation is Medium. Non-openers may see your email in spam even when your overall reputation is High.
Layer 3: Content and Technical Signals
After identity and reputation, the inbox provider evaluates the email’s content and technical characteristics. This is the layer that most senders focus on, but it is the least important of the five.
Technical Signals
Technical signals include the format of the email headers, the HTML structure, the text-to-image ratio, and the domains of any links included.
Malformed headers, unusual formatting, missing required fields — suggest automation tools or infrastructure that spammers commonly use. High image-to-text ratios (predominantly images with little readable text) pattern-match against spam campaigns that put all their content in images to evade text-based filters.
Link domains are checked against reputation databases. If your email contains links to domains that have been associated with spam or malware, the email receives additional filtering pressure regardless of your own domain’s reputation.
Content Signals
Content signals include the presence of certain language patterns, the formatting of the subject line, the presence of an unsubscribe mechanism, and the overall structure of the email.
Modern content filters do not check for specific words. They identify patterns. Urgency combined with financial language combined with a promotional structure is a pattern. Educational content with substantive paragraphs and a clear navigation structure is a different pattern.
However, content signals are secondary. The same content sent from a domain with High reputation typically reaches the inbox. The same content sent from a domain with Low reputation typically goes to spam. The reputation signal dominates the content signal in most cases.
Layer 4: Recipient Behaviour Signals
At Gmail, inbox placement is not purely domain-level. Individual recipient behaviour is also factored in.
Gmail tracks how each recipient has historically interacted with email from your domain. If a recipient has opened your last 15 emails and clicked on several of them, Gmail assigns that recipient a higher probability of wanting future email from you. It places your next email in their Primary inbox accordingly.
If a recipient has never opened any of your emails, Gmail infers they are not engaged with your content. The next email may be directed to Promotions, or in more extreme cases, filtered to spam.
This personalised filtering has significant implications. You can have different inbox placement rates for different segments of your list not because of anything different about the emails you send, but because of the engagement history each subscriber has built with your domain.
It also means that engagement-based segmentation is a deliverability tool, not just a marketing optimisation. Concentrating sends on your most engaged subscribers improves the engagement signals your domain generates, which strengthens domain reputation and improves placement for all subsequent sends.
Layer 5: Real-Time Engagement Data
The fifth layer is the most dynamic. Inbox providers monitor how the current batch of email is being received in real time.
When you send a campaign to 100,000 subscribers, Gmail processes incoming emails and monitors what recipients are doing with them. If the first few thousand recipients open and click at high rates, that positive signal can improve inbox placement for the remaining sends. If the first few thousand generate spam reports and deletions, filtering tightens for the rest of the batch.
This is why sending to your most engaged segments first rather than sending to the full list simultaneously is a best practice for large campaigns. The engaged segment generates the initial positive signals that support better placement across the rest of the batch.
It also explains why campaigns that perform well for active subscribers and poorly for inactive ones are not just a segmentation issue. The positive signals from the active segment are potentially improving placement for the inactive segment that follows. Separating them entirely — sending to inactive subscribers from a different subdomain on a separate day prevents this cross-contamination from working in your favour.
How These Layers Combine into a Placement Decision
The five layers do not have equal weight. Gmail’s filtering system prioritises them roughly in this order:
1. Authentication failure is a hard negative signal. Persistent authentication failures push email toward spam regardless of reputation.
2. Domain reputation is the primary placement driver. High reputation delivers most email to the inbox. Low or Bad reputation delivers most email to spam.
3. Recipient engagement personalisation adjusts placement for individual recipients within the domain reputation context.
4. Real-time engagement signals from the current batch adjust placement as the send progresses.
5. Content and technical signals are tiebreakers. They matter when reputation and engagement signals are neutral or ambiguous.
This hierarchy explains why fixing content does not fix inbox placement when the root cause is domain reputation. And why improving reputation through verified lists, engaged segments, and consistent sending is the most reliable path to sustained inbox placement.
Why Gmail and Outlook Evaluate Differently
Gmail and Outlook use different filtering architectures, which is why a domain can have excellent Gmail delivery and poor Outlook delivery (or vice versa).
Gmail prioritises domain reputation and recipient engagement personalisation. Domain-level signals dominate.
Microsoft 365 / Exchange Online Protection uses a different reputation database, integrates Spamhaus and other blocklists more directly into its filtering, and applies organisation-level policies set by corporate IT administrators. Enterprise Outlook users may have additional content filtering layers applied by their company’s security team.
This means monitoring Google Postmaster Tools alone is insufficient. Microsoft SNDS (Sender Network Data Service) provides IP-level reputation data for Microsoft’s mail system that is completely separate from what Postmaster Tools shows.
What Senders Can Actually Control
Given how filtering actually works, senders have direct control over some factors and indirect influence over others.
Direct control:
- Authentication records (SPF, DKIM, DMARC)
- List quality (through verification and suppression management)
- Sending volume and patterns (warm-up, ramp controls)
- Content and technical email structure
Indirect influence:
- Domain reputation (by managing the signals your sends generate)
- IP reputation (by managing your shared pool or warming your dedicated IP)
- Recipient engagement personalisation (by concentrating sends on engaged subscribers)
No direct control:
- The specific algorithms inbox providers use
- Individual filtering decisions for specific recipients
- Policies applied by corporate IT administrators for enterprise Outlook users
Understanding this distinction prevents wasted effort. Optimising content and subject lines has limited effect on inbox placement when the root issue is domain reputation. Improving domain reputation requires the inputs that directly influence it: verified lists, engaged segments, and consistent authenticated sending.
Key Takeaways
- Modern inbox placement decisions use five layers of evaluation: sender identity, IP and domain reputation, content and technical signals, recipient behaviour signals, and real-time engagement data.
- Domain reputation is the primary driver of inbox vs spam placement. Content signals are secondary and have limited impact when reputation is the root issue.
- Gmail uses personalised filtering. Subscribers who have historically engaged with your emails get better inbox placement than non-engagers from the same domain, sending the same email.
- Real-time engagement from the first portion of a send influences placement for the rest of the batch. Send to your most engaged subscribers first.
- Gmail and Outlook use different evaluation systems. Monitoring only Postmaster Tools misses Outlook-specific reputation signals visible in SNDS.
- Senders have direct control over authentication, list quality, and sending patterns. They have indirect influence over reputation through those inputs. Content is the least powerful lever in the filtering hierarchy.
Frequently Asked Questions
Yes, but reputation matters more. Subject lines mainly influence opens, while sender reputation has a stronger impact on inbox placement.
Inbox placement can vary based on each recipient’s engagement history. Active subscribers are more likely to receive your emails in the inbox.
Only marginally. Content can influence filtering, but fixing sender reputation and engagement signals is more effective.
Yes. Opens, clicks, replies, and other positive engagement signals can help inbox providers recognize that recipients value your emails.
Yes. Prioritize authentication, bounce rates, complaints, and engagement first. Once reputation is healthy, optimize subject lines and content for higher engagement.
Conclusion
Understanding how inbox providers actually evaluate email changes the priorities for every deliverability decision. Authentication, reputation, and engagement signals dominate. Content is a secondary factor.
This means the highest-leverage deliverability investments are also the right ones: verify your list to improve bounce signals, segment by engagement to concentrate positive signals, authenticate properly to establish verified identity, and monitor reputation weekly to catch problems early.
Content optimisation matters, but it belongs at the end of the priority list, not the beginning.
