Plain Text vs. HTML: Which Is Better for Cold Email?

Short answer

Use plain text — or HTML so minimal it's indistinguishable from plain text. For a one-to-one cold email, branded headers, images, buttons, and CSS lower reply rates, push mail toward Gmail's Promotions tab, and add spam-filter risk. Keep formatting near zero and treat open tracking as optional, not default.

What "plain text" actually means

Most people arguing about this are arguing about two different things. There's the visual question — does the email look designed? — and there's the technical question of what MIME parts your sending tool puts on the wire.

Every email is either text/plain, text/html, or multipart/alternative, which carries both versions and lets the client pick. A lot of cold email software sends multipart/alternative even when your draft has no formatting at all, because the editor wraps your paragraphs in HTML tags behind the scenes. So the message you think is plain text may technically be HTML with a plain-text twin attached.

That's usually fine. Filters don't hate HTML — they hate the things HTML is used for. A multipart message whose HTML part is a handful of paragraph tags and one link scores almost identically to true plain text. A message that is HTML-only, with a 60 KB body full of table layouts and inline styles, does not.

One rule that does matter: if you send multipart/alternative, the plain-text part has to actually match the HTML part. Tools that stuff a generic "view this email in your browser" line into the text part, or leave it empty, are handing filters a cheap signal. SpamAssassin has had rules targeting HTML-only mail and mismatched or missing text alternatives for years, each worth a fraction of a point — individually trivial, collectively the difference between a 2.1 and a 4.8.

Why decorated HTML hurts cold outreach specifically

Newsletters can afford design because the recipient opted in and expects a broadcast. Cold email can't, for four reasons.

It breaks the premise. You're claiming to be one person who found their company and wrote to them. A logo header and a rounded CTA button says otherwise in the first half-second, before a word is read. The reader doesn't consciously think "this is bulk" — they just file it with the other bulk.

It changes how Gmail categorizes you. Promotions-tab classification leans on structural signals: image-to-text ratio, multiple outbound links, tracking pixels, list headers, commercial phrasing. A stripped message with one link and no images has fewer of those signals than a template does. It's not a guarantee — Gmail's Promotions tab has its own logic — but format is one of the levers you control cheaply.

It inflates message size. A plain cold email is 2–4 KB. A template with inline CSS and a base64 logo can hit 80–120 KB. Gmail clips messages over roughly 102 KB, showing a "[Message clipped]" link — and your unsubscribe line usually lives below the fold that gets cut.

It renders badly more often than you think. Outlook on Windows still uses a Word-based rendering engine for parts of HTML layout; dark mode inverts backgrounds and turns transparent-background logos into gray boxes. Plain text renders correctly everywhere, forever.

What plain text actually costs you

Be honest about the trade-offs instead of pretending there are none.

You lose the open pixel. Open tracking requires a 1×1 image, and an image requires HTML. If you go true text/plain, your open rate is zero by definition. In practice that's a smaller loss than it sounds: security scanners at Proofpoint, Mimecast, and Microsoft Defender fetch images and follow links before the human ever sees the message, so open numbers are noisy at best. We covered the mechanics in why cold email open rates lie. If replies and verified site visits are your real metrics, dropping the pixel costs you a vanity number and removes a filter signal at the same time.

You lose click attribution unless you're willing to rewrite links. Link rewriting is its own deliverability question — a tracking domain shared with hundreds of other senders is a reputation liability. If you track clicks, use a subdomain of your own sending domain with its own CNAME, and keep it to one tracked link per message.

You lose visual hierarchy. For a 120-word email that shouldn't matter. If your cold email genuinely needs bullet points and bold subheads to be understood, the email is too long.

The formatting rules that matter more than the format

Once you've decided on "minimal," these are the specifics that move the needle:

• No images at all — not a logo, not a headshot, not a signature banner. Image-only or image-heavy bodies trip well-known filter rules.
• One link, two at most, and one of those can be your unsubscribe line.
• Never mismatch anchor text and destination. "Click here" pointed at a domain that doesn't appear in the text is a classic phishing pattern. Show the actual domain, or link a word that names it.
• Skip the tracking pixel by default. Turn it on for a specific test, not for every campaign.
• Keep the body under about 150 words. Short bodies are also small bodies, which keeps you far from any clipping threshold.
• Use simple paragraph breaks. No nested divs, no inline styles, no font declarations — those come from pasting out of Word or Google Docs, so paste as plain text first.
• Signature: name, company, one line of contact, and the physical address CAN-SPAM requires. Text only. No vCard attachment, no social icons.
• Follow-ups reply in-thread with the original quoted below. Once the thread starts, the format is locked to whatever the first message was — another reason to start clean.

How to test it instead of arguing about it

Two things are worth measuring, and one isn't.

Worth measuring: your spam score before send. Run the draft through SpamAssassin scoring and look at which rules fire, not just the total. If an HTML-related rule is contributing, you'll see it by name and can strip the offending element. Also worth measuring: actual inbox placement, using seed mailboxes across Gmail, Microsoft 365, and a couple of smaller providers. Send the plain version and the styled version from the same warmed mailbox on the same day and compare where they land. That's a proper inbox placement test, and it answers the question in an afternoon.

Not worth measuring, usually: a format A/B test on reply rate. If your baseline reply rate is 4% and the true effect of stripping HTML is 2 points, you need roughly 2,000 leads per variant to detect it with any confidence. Most campaigns split-testing format on 400 contacts are reading noise. Placement tests are cheap and conclusive; reply-rate tests on format are expensive and usually aren't.

Where Sendvanta fits

Sendvanta's editor produces clean, minimal markup by default with a matching plain-text part — no template wrapper, no injected branding, no forced pixel. Every draft can be scored with SpamAssassin before it sends, so you see the exact rules firing rather than a vague traffic light, and seed-mailbox placement tests tell you which tab you're landing in across Gmail and Microsoft 365. Because engagement classification is bot-aware, turning the open pixel off costs you less than it would elsewhere — verified visits and replies are what the reporting is built around anyway. Mail goes through your own Gmail, Microsoft 365, or SMTP mailboxes, never a shared IP. The free plan is $0 forever with 1,000 active leads and 3,000 emails a month; paid plans start at $29/mo. More detail on what's included.

Methodology: this guidance reflects MIME behavior, SpamAssassin rule categories, and Gmail and Microsoft 365 rendering as observed in seed-mailbox placement testing as of February 2026.