Plain Text vs HTML in Cold Email: Deliverability, Authenticity and What to Actually Send

The format question sounds cosmetic — plain text or HTML? — but in cold email it decides two outcomes before a word is read: whether the message reaches the inbox, and whether it reads as a personal letter or a marketing blast. This guide walks through the actual mechanics on both counts and lands on a concrete recommendation: what your first-touch emails should be built from, byte by byte.

First, the honest framing: it's minimal vs designed, not text vs HTML

Almost no business email today is literally text/plain. When you write from Gmail or Outlook, the client sends a multipart message whose visible part is simple HTML — paragraphs, maybe a link, a signature. So the practical question is not «HTML or not» but «how much HTML»: a lean, typed-looking note at one end, a designed template with headers, images, columns and buttons at the other.

That reframing matters because each end belongs to a different genre of email, and everyone — recipients, filters, regulators — treats the genres differently. Designed templates are the native format of newsletters and promotions sent to subscriber lists. Lean notes are the native format of one-to-one business correspondence. Cold outreach to a named decision-maker is, by its nature, claiming to be correspondence; dressing it as a promotion contradicts the claim in the first glance.

This is where the LDM angle is not a preference but a definition: addressed B2B outreach means a specific person at a specific legal entity, written to for a specific reason. The format should look like what the message claims to be. When it does, the deliverability and perception effects below mostly take care of themselves.

The deliverability mechanics: what filters read in your markup

Modern filtering is dominated by sender reputation — who you are and how your mail has behaved — but message structure is still an active input, and for a cold sender without years of history it carries real weight. Filters and corporate gateways evaluate the ratio of markup to text, image-to-text balance, the number and nature of links, tracking pixels, remote resources, and how closely the message resembles known bulk-mail templates. A 40KB HTML payload carrying 80 words of actual text is structurally a promotion, whatever the words say.

The B2B twist: your recipients sit behind corporate gateways — Microsoft Defender for Office 365, Proofpoint, Mimecast — which are more conservative than consumer Gmail. These systems sandbox messages, rewrite and pre-click links, strip or block remote images, and score «marketing-shaped» structure aggressively. Every additional link is another URL to scan and another chance to trip a category filter; every image is a blocked box; every button is a bulk-mail fingerprint. Lean messages give these systems less to object to.

Two structural elements deserve specific caution. Tracking pixels — the invisible image that reports opens — are both a bulk fingerprint for filters and increasingly useless as a metric, since privacy proxies and scanners inflate opens beyond usefulness; in first-touch cold email you can simply drop open tracking. Link tracking is more defensible (replies and clicks are what matter), but run it through a clean branded tracking domain, because a hover revealing a third-party redirect chain undermines both filter scores and human trust. And keep it to one link; zero is even safer if your CTA is a reply question — which, in a first touch, it usually should be.

Orientation numbers, honestly framed: teams that strip first-touch emails from designed templates down to minimal format routinely see meaningful placement improvements at corporate domains — and reply rates follow, since healthy cold B2B campaigns live in the 3–8% reply band only when they reach the inbox at all. The effect size varies by audience and sender history; the direction is consistent.

• Markup weight: a first touch should be a few KB, not tens of KB
• Images: zero in first touch — logos and banners are blocked, bulk-flagged, or both
• Links: one, contextual, on a branded tracking domain; no social icon rows
• Open-tracking pixel: drop it in cold first touch; the data is polluted anyway
• Buttons and columns: bulk fingerprints — use a plain sentence with a plain link
• Attachments: never in a first touch; that is the malware pattern gateways exist to catch

The perception mechanics: what the format tells a human

Recipients classify email by sight before reading it. A designed header, brand colors and a button announce «this was sent to a list» — and trigger list-mail behavior: skim, archive, maybe unsubscribe, reply never. A typed-looking note from a named person announces «someone wrote to me» — and triggers correspondence behavior, where replying is the socially normal response. Since the entire economics of cold outreach runs on replies, the format choice is effectively a choice of which behavioral script the recipient runs.

This is also an authenticity claim you must keep consistent. A plain note whose first line is obviously mail-merged («Hi {FirstName}, I loved your recent post») fails the same test from the other side: the format promised a human, the content revealed a robot, and the mismatch reads as deceptive rather than merely lazy. Minimal format raises the bar on your copy — which is the correct direction for addressed outreach anyway: real personalization from real account signals, or honest segment-level framing.

There is a legitimate boundary case: some content genuinely needs structure. A benchmark table, a short case summary with figures, a product visual. The answer in first touch is: don't send it — link to it or offer it. «Want the two-page comparison? I'll send it over» is both a lighter email and a better CTA than embedding the comparison. Structure belongs in later touches, after a reply established the thread, where a table or a PDF arrives as requested material rather than unsolicited packaging.

The rendering reality: where heavy HTML mechanically breaks

Even if a designed email dodges filters and skeptical readers, it still has to survive B2B rendering conditions — and they are hostile. Desktop Outlook, still standard in exactly the corporate environments cold B2B targets, renders HTML with a legacy engine that mishandles much of modern CSS: layouts shift, spacing collapses, fonts substitute. Templates that look perfect in a design tool arrive subtly broken where your buyers read them.

Images are blocked by default in many corporate clients, so an image-led email arrives as a skeleton of empty frames with your pitch trapped inside an unloaded banner. Dark mode — now common on phones and increasingly on desktop — inverts and recolors designed emails unpredictably: dark text on dark backgrounds, logos in invisible boxes. A minimal message is immune to all of this by construction: text renders as text everywhere, in every mode, on every client from a 2010 Outlook install to a phone lock screen.

The maintenance economics point the same way. Designed templates need testing across dozens of client-mode combinations and re-testing every time a client updates its engine. A minimal format needs none of that — the effort goes into list quality and copy, which is where addressed outreach earns its returns anyway. Robustness, not aesthetics, is the underrated argument: the format that cannot break is the one that always delivers your first impression intact.

The recommended format, byte by byte

Here is the concrete spec for a first-touch cold email that both filters and humans read as legitimate correspondence. Structure: multipart message with a proper plain-text part and a minimal HTML part that mirrors it — short paragraphs, default fonts, no CSS beyond what a mail client itself would produce. Length under roughly 150 words. One link at most, in a sentence, on a branded domain; no images, no attachments, no buttons, no pixel. Signature in four lean text lines: name, role and company, one verification link, opt-out route. Subject in sentence case, conversational, no promotional vocabulary.

Compliance folds neatly into the same format. CAN-SPAM requires truthful sender identification, a non-deceptive subject, a physical postal address and a working opt-out — all of which fit in two quiet signature lines and read as professionalism rather than boilerplate. Under GDPR, one-to-one outreach on legitimate interest should state plainly who you are and make objecting effortless; a human sentence («not relevant? say so and I won't write again») does this better than a newsletter footer, and converts irritation into a harmless no instead of a spam complaint.

Later touches can relax the spec gradually: a requested PDF after a reply, a table in a third message where it genuinely serves, even a styled proposal once a conversation exists. The rule that survives the whole sequence: format follows relationship. The colder the contact, the leaner the message.

In LDM this spec is the default rather than a discipline to remember: templates are plain-text-first with spintax and account-signal variables, campaign QC flags heavy markup, excess links and missing opt-outs before send, tracking runs on branded domains, and per-account send limits keep volume patterns looking human. The format does the quiet work; the copy and the targeting do the rest.

Testing the question on your own audience

General mechanics favor minimal, but your audience has its own mix of gateways, clients and expectations — so verify on your traffic. The test design: same list split randomly at send time, same subject, same copy, two builds — your current template versus the minimal spec above. Judge on replies, classified by sentiment, and on placement where you can measure it; opens are too polluted by prefetching to settle anything.

Watch per-provider splits, not just totals: a format effect often concentrates at corporate domains behind strict gateways while consumer Gmail shrugs. And give the test volume — at addressed-outreach scale, a few hundred deliveries per arm across two or three campaign waves before calling it. The most common result, for what it's worth: minimal wins or ties on replies everywhere, and wins clearly on placement at exactly the enterprise domains that matter most.

• Split at send time, 50/50, same list, same copy — only the build differs
• Primary metric: replies (sentiment-classified); secondary: placement by provider via seed tests
• Ignore raw opens — prefetchers and scanners have made them noise
• Segment results by recipient provider: corporate gateways vs consumer inboxes
• Minimum a few hundred deliveries per arm before concluding
• Retest after any major change in sender domain, audience or template

FAQ