Email content affects deliverability — but not in the simplistic "spam words" model that dominated deliverability advice a decade ago. Modern ISP spam filters (Gmail, Yahoo, Microsoft Exchange Online Protection) and corporate gateway filters (Proofpoint, Barracuda) use sophisticated multi-signal evaluation that weights sender reputation and recipient engagement far more heavily than content signals alone. Understanding what content signals these filters actually evaluate in 2026 — and which content practices still matter, and which are myths — is the basis for content decisions that support rather than undermine deliverability.
How Modern Spam Filters Actually Work
Consumer ISP spam filters (Gmail, Yahoo, Outlook.com) and corporate gateway filters (Proofpoint, Barracuda, Mimecast) evaluate incoming messages using fundamentally different approaches, and understanding the difference is essential for content decisions that work across both filter types.
Consumer ISP filters: Gmail's spam filter is primarily a reputation and engagement-based system. The most weighted signals are: sender domain reputation (Postmaster Tools domain reputation tier), recipient engagement history (has this specific recipient previously engaged with email from this sender?), and spam rate signals from other Gmail users (how many other Gmail users marked this sender's mail as spam?). Content signals are evaluated but are secondary to reputation signals for senders with established domain history. A message from a sender with High Gmail domain reputation will reach the inbox even if it contains some content elements that might trigger scoring on a reputation-neutral sender. Conversely, a sender with Low domain reputation will land in spam regardless of content quality.
Corporate gateway filters: Proofpoint, Barracuda, and Mimecast apply rule-based and machine learning content scoring that weights content signals more heavily than consumer ISP filters do — particularly for senders without established reputation history at that corporate gateway. Cisco Talos SenderBase, Proofpoint's IP reputation, and Barracuda's sender reputation all feed into gateway filtering. Content scoring (HTML structure, image-to-text ratio, link analysis, attachment analysis) plays a meaningful role in corporate gateway decisions, especially for new senders or senders without previous delivery history to that corporate domain.
The practical implication: content optimisation matters most for (a) new senders without established reputation history, (b) senders whose reputation is at Medium or Low (where reputation cannot overcome content scoring), and (c) senders with significant B2B audiences served through corporate gateways. For established senders with High reputation sending to consumer ISPs, content optimisation matters at the margin — it is not the primary deliverability lever.
HTML Structure: What Filters Evaluate
Email HTML is evaluated by spam filters for structural patterns that correlate with spam campaigns rather than legitimate commercial email. The structural signals that consistently affect content scoring:
Excessive nesting and complexity: Spam templates frequently use deeply nested table structures (tables within tables within tables, 10+ levels deep) because they were designed to render consistently across old email clients before modern HTML support. Contemporary corporate gateway filters flag this structural pattern. Modern commercial email templates should use relatively flat HTML structure — 2-3 levels of nesting maximum — that is also easier to maintain and renders reliably across all email clients.
Hidden or white-text content: Spam filters actively scan for content rendered invisible to the human reader — white text on white background, zero-font-size text, display:none content that is not legitimate preheader text. Any invisible content that appears to be stuffing keywords or manipulating content-to-image ratios is a strong spam signal. Preheader text (the preview text visible in email client list views) should use standard HTML comment or visible-then-hidden techniques that email clients recognise, not CSS tricks that appear manipulative to spam filters.
Excessive external resources: Messages that load 20+ external images via separate HTTP requests, reference many external CSS resources, or include numerous third-party tracking scripts generate higher spam scores at some corporate gateway filters. Each external resource requires an HTTP connection that the filter may analyse — a message with 30 external image requests looks like a high-tracking-density marketing template rather than a legitimate communication.
Malformed or non-standard HTML: Severely malformed HTML (unclosed tags, broken attributes, invalid nesting) appears in spam campaigns where templates are assembled programmatically without validation. Well-formed, HTML-standards-compliant email templates produce lower content scoring than malformed ones. Validate templates with the W3C HTML validator before deployment.
Image-to-Text Ratio and Image-Only Emails
Image-only emails — messages where the entire visible content is embedded in a single large image, with no text content outside the image — consistently score poorly on corporate gateway content filters and have degraded inbox placement at many ISPs. Spam filters apply this scoring because image-only emails are a common spam technique: the content is in an image that spam filters cannot read for keyword analysis, while the surrounding HTML contains no evaluable text. Additionally, recipients who receive an image-only email in a mail client with images disabled see a completely empty email — a bad user experience that generates both complaints and disengagement.
The minimum text-to-image ratio for acceptable content scoring at major corporate gateways: 60% text, 40% images by word count (not pixel area). This means a message with 5 product images should have approximately 150+ words of text content to reach a 60/40 balance. The text does not need to be long paragraphs — product descriptions, feature bullet points, navigation links, and footer content all count toward the text total.
For predominantly visual campaigns (fashion retail, photography, design services) where images are central to the brand communication, the approach: maintain 60%+ text ratio in the HTML structure while using large images, include meaningful alt text for every image (alt text is text content that spam filters read even when image loading is disabled), and ensure the email communicates its core message and CTA effectively for recipients who see no images.
Link Domains and URL Patterns
The domains used in email links are one of the most consistently evaluated content signals across both consumer ISP and corporate gateway filters. Link domain signals that affect spam scoring:
URL shorteners: Links using Bit.ly, TinyURL, t.co, or similar URL shortening services are a strong spam signal at Proofpoint, Barracuda, and many corporate gateways. Spam campaigns use URL shorteners to obscure the destination domain from content filtering. Legitimate commercial email should never use URL shorteners — use full URLs with proper UTM parameters for tracking instead.
Mismatched link domains: When the From: domain is brand.com but all links point to completely different domains (ad-network.com, tracking-platform.net), spam filters flag the domain mismatch as suspicious. Use a click-tracking domain that is a subdomain of the sending domain (click.brand.com or track.brand.com) rather than a generic third-party tracking domain.
Newly registered or low-reputation link domains: If the link domain (or click-tracking domain) is newly registered (less than 6 months old) or appears on domain reputation blocklists (SURBL, URIBL), content scoring is elevated. Use a tracking subdomain of the established sending domain — the sending domain's age and reputation transfers to the subdomain, providing the history that newly registered tracking domains lack.
Excessive link count: Messages with 50+ unique URLs are occasionally flagged by corporate gateway filters that analyse all URLs in a message for reputation. This is particularly relevant for product catalogue emails with individual product links. Consider reducing link count by using category links rather than individual product links, or by aggregating tracking through a primary click URL rather than unique URLs per product.
Subject Lines: Myths and Reality
Subject line spam scoring is one of the most misunderstood aspects of email content and deliverability. The myth: certain words in the subject line (FREE, URGENT, !!!, all caps) automatically send emails to spam. The reality: for established senders with good reputation, subject line keyword scoring contributes very little to Gmail's filtering decision. For new senders or senders with poor reputation, subject line signals matter slightly more but remain secondary to reputation signals.
The subject line signals that do matter:
Deceptive subject lines: Subject lines designed to mislead the recipient about the message content — "Re: Your recent inquiry" when there was no inquiry, "Payment confirmation" for a marketing email, "Important account update" for a promotional send — are specifically prohibited by CAN-SPAM and generate disproportionate complaint rates when recipients feel deceived. The deliverability impact from complaint rates is far more significant than from any keyword scoring.
Excessive punctuation and caps: "FREE FREE FREE!!!! ACT NOW!!!!" generates spam scoring at SpamAssassin (still used by some corporate gateways and regional ISPs) even if it no longer strongly affects Gmail or Yahoo. More importantly, it generates higher open rates initially (clickbait) followed by much higher complaint rates from recipients who feel manipulated — a net negative deliverability outcome.
Emoji: Emoji in subject lines do not systematically affect spam scoring at major ISPs but may affect rendering on some corporate email clients (Outlook on Windows renders some emoji as placeholder characters). Test emoji rendering across the target email client mix before deploying emoji-heavy subject lines to corporate audiences.
Spam Words: Do They Still Matter?
The "spam words" list approach to content optimisation — avoiding words like "free," "guarantee," "no risk," "limited time offer" — was relevant in the SpamAssassin era when keyword-based content filtering dominated. At Gmail and Yahoo, keyword content scoring has been largely replaced by engagement-based and reputation-based filtering. At Proofpoint and Barracuda (corporate gateways), some keyword scoring remains in place but is weighted far below sender reputation signals.
The practical answer: spam words matter at the margin for corporate gateway content scoring, particularly for new senders with no reputation history at the target corporate domain. For established senders with good reputation, avoiding "free" in a legitimate promotional email is unnecessary and counterproductive (if "free shipping" is the offer, say so clearly). For new senders warming a domain or sending cold email to corporate audiences, avoiding highly-loaded promotional language in early sends reduces the risk that content scoring adds to the reputation-building friction of a new sender.
The more important content concern for established senders is not keyword scoring but engagement quality. Content that recipients find genuinely useful generates opens and clicks — positive signals that outweigh any marginal keyword scoring. Content that recipients find misleading or disappointing generates complaints — negative signals that dramatically outweigh any content quality score. Write for the recipient; the spam filter follows.
Content Scoring Tools: SpamAssassin and ISP Filters
SpamAssassin: The open-source spam filter used by many corporate mail servers and regional ISPs. SpamAssassin scoring is based on a combination of keyword rules, header analysis, DNS-based reputation checks (Spamhaus, SURBL), and authentication checks. A SpamAssassin score above 5.0 is likely to be filtered as spam at servers using default SpamAssassin configuration. Mail-tester.com uses SpamAssassin scoring as part of its test report — a useful proxy for corporate gateway content scoring even though it does not replicate Proofpoint or Barracuda specifically.
GlockApps spam score: GlockApps's content analysis component tests against both SpamAssassin and ISP-specific filters, providing a composite content score alongside inbox placement results. Using GlockApps for both content scoring and placement testing in a single tool provides efficient pre-campaign quality checking.
Postmaster Tools spam rate (indirect content signal): The Gmail Postmaster Tools spam rate, while not a direct content scoring tool, is the most operationally relevant content quality signal available — it shows whether recipients are marking the programme's email as spam. Spikes in the spam rate correlated with specific campaign deployments identify campaigns (and therefore their content) as generating more complaints than baseline. This is the most actionable content quality feedback loop available.
Content Best Practices for Deliverability
The content best practices that consistently support deliverability across all filter types:
- Use full URLs, never shorteners. Every link should show the actual destination domain — use tracking parameters (UTM) for analytics rather than URL shorteners that obscure destinations.
- Maintain 60%+ text-to-image ratio. Include meaningful text alongside images. Write useful alt text for all images — it is both accessibility best practice and content scoring positive signal.
- Use a branded click-tracking subdomain. Replace generic tracking domains (clicks.esp.com) with a branded subdomain (click.yourbrand.com) that shares the sending domain's reputation.
- Match the From domain across From, Reply-To, and primary links. Consistency across these identifiers reduces mismatch signals that affect corporate gateway scoring.
- Write for the recipient, not the filter. Content that genuinely interests recipients generates engagement signals that outweigh any content scoring concern. Content that disappoints or misleads generates complaints that are the most severe deliverability signal of all.
- Test every new template before deployment. Use mail-tester.com or GlockApps to score each new template against SpamAssassin and ISP content filters. Investigate any score below 7/10 before deploying at production volume.
Email content in 2026 matters for deliverability — but not in the keyword-avoidance model that dominated email marketing advice a decade ago. The content signals that matter are structural (HTML quality, image-text ratio, link domain trust), behavioural (does the content generate engagement or complaints from real recipients?), and contextual (does the content match what the subscriber opted in to receive?). Optimise for all three, test before deployment, and content will support rather than undermine the reputation-based deliverability that sustains commercial email programme performance.