Every email-marketing tutorial cites the 60/40 rule: at least 60% text, no more than 40% images. Like most rules in this category, it was a useful approximation in 2008–2012 and has been overtaken by smarter filters. Modern systems rarely score on a literal pixel ratio. They score on whether the email communicates without images — and what the images actually say when read by OCR.
Pure image-only emails are penalised everywhere. Heavy image-to-text ratio is mildly penalised. Modern filters OCR the images and score the extracted text. Alt text matters. Single-pixel tracking images are fine — but the more of them you embed, the more obvious the marketing-automation pattern.
Why image-only emails are penalised
For roughly a decade, spammers shipped messages where 100% of the visible content was a single image — the body of the email was "Hi" or empty, and a JPG contained the entire sales pitch. The reason was simple: filters in 2005 couldn't read images, so the message bypassed all word-based scoring.
Filters caught on by the late 2000s. Today, an image-only email triggers a heavy penalty across every major provider:
- SpamAssassin rule
MIME_HTML_ONLY+HTML_IMAGE_ONLY_*can stack 3+ points alone. - Gmail's ML treats hidden-text patterns (no rendered text when images blocked) as a strong manipulation flag.
- Outlook routes most image-only emails to Junk by default.
The penalty is for hiding intent — not for using images. Use as many images as you want, as long as the email still communicates when images are blocked.
The OCR layer: filters read your images
Gmail, Outlook and the major commercial filters all OCR embedded and inlined images. The extracted text feeds the same scoring model as the body text. This catches the classic spam evasion of putting trigger words inside an image:
- A pharmacy ad with the entire product description as a JPG gets scored as if those words were typed.
- Crypto pump "guarantee" phrasing inside a graphic banner triggers crypto-spam classifiers.
- QR codes have been OCR'd and their target URL scored since ~2022 — phishing-by-QR is a known vector and filters catch it.
OCR isn't perfect — heavily stylised text or low-resolution imagery can slip through. But assuming filters can't read your images is a 2010 assumption.
Alt text matters
The alt attribute on <img> tags was always good practice for accessibility. It is now also scored by filters for two reasons:
- Alt text is shown when images are blocked. Filters check whether the with-images-blocked rendering communicates meaningfully — alt text is part of that rendering.
- Alt text is read by screen readers. Filters serving accessibility-focused mailboxes weigh its presence.
Practically: every meaningful image should have descriptive alt text. Decorative images can use empty alt (alt="") — explicitly empty is better than missing the attribute. Tracking pixels should always be alt="".
Single-pixel tracking images
The 1×1 transparent GIF used for open tracking is universal across marketing-automation platforms. Filters know it's there and don't penalise it directly — but the presence of a tracking pixel is itself a signal that the message is automated mail, not personal correspondence.
For cold outreach, where the goal is to look like a personal message, removing the open-tracking pixel removes one broadcast-marketing signal. Apple's MPP has already destroyed open-rate accuracy as a metric. The pixel provides increasingly useless data and a small reputation cost.
Apple Mail Privacy Protection pre-fetches every tracking pixel, inflating opens to ~100% on Apple devices. Gmail does similar pre-fetch on some accounts. Open rate as a metric is essentially noise in 2026 — the cost of the tracking pixel now outweighs the data value.
What ratio actually works in 2026
From measured testing across 20+ seed mailboxes:
- Pure text: baseline. Inboxes well from warm domains.
- ~80% text, 20% image (one banner or two icons): baseline. No measurable penalty.
- ~50% text, 50% image: small but consistent penalty (~3-5% lower inbox rate).
- ~25% text, 75% image (heavy newsletter style): moderate penalty (~10-15% lower inbox rate). Common for retail marketing — accepted as "Promotions" class at Gmail.
- Image-only: heavy penalty. ~30-50% lower inbox rate across consumer providers.
For cold B2B outreach, target the first or second category. For consumer marketing, the 25/75 ratio is acceptable but you should expect Promotions placement at Gmail.
Mobile rendering
Roughly two-thirds of email opens happen on mobile in 2026. Mobile rendering issues are a separate engagement problem (recipient deletes without reading = negative engagement signal) but they don't directly affect spam scoring. What matters for filters:
- Images sized appropriately — embedded 4000px-wide images that force horizontal scroll are a sign of unprofessional template.
- Responsive HTML with
max-width: 100%on images. - Text remains readable when images are blocked (default on many mobile email apps in cellular mode).
Practical guidelines
- Never send image-only. Always include real text the email communicates with even if every image is blocked.
- Aim for 3:1 text-to-image ratio for B2B. 1:1 to 2:1 for retail, accepting Promotions placement.
- Add descriptive alt text to every meaningful image.
- Don't hide trigger words inside images — OCR catches it.
- For cold outreach: skip the tracking pixel. The data is noise and the pixel is a broadcast signal.
- Test rendering with images-blocked. Most ESPs have a preview mode for this. If the email is unreadable, fix it.