Email Marketing With AI-Generated Video

The single most common mistake in email marketing with video is embedding an actual video tag and hoping. Outlook won't play it. Gmail's web client won't play it. Most corporate clients strip the element entirely and leave a blank space where your hero was. HTML5 video in email has been theoretically supported and practically unusable for a decade, and nothing about 2026 changed that.

What works is unglamorous: a still or a short animated loop, a visible play affordance, and a link to a page where the video actually lives. The technical debate has been settled for years. What's worth writing about now is the production side — video-in-email used to be a quarterly special because a bespoke 20-second clip per campaign wasn't worth the effort. When the clip is a 15-minute generation, it becomes a normal part of every send.

This is the workflow: what to make, how to make it, how to wire it, and what to measure. Versely produces the assets; your ESP sends the mail. Two separate systems, and this assumes you have the second.

What actually renders in an inbox

Before producing anything, know the constraint set. Four options, and only two of them are real.

Approach	Renders in Outlook	Renders in Gmail	File weight	Verdict
HTML5 <video>	No	No (web)	n/a	Don't
AMP for Email video	No	Limited, sender allowlisted	Low	Niche
Animated GIF	Yes (first frame only in some Outlook builds)	Yes	300KB–2MB	Default choice
Static poster + play button	Yes	Yes	40–150KB	Safest choice

The practical rule: animated GIF with a strong first frame. Outlook builds that don't animate show frame one, so if frame one carries the message you lose nothing. Design for that — a complete, legible image with the value proposition visible, animation as a bonus.

Keep the GIF under about 1MB. Above that you're adding mobile load time for marginal benefit, and some filters treat heavy image payloads as a spam signal. Three to five seconds, 8–12 fps, and a hard crop usually gets you there.

The per-campaign production run

Here's the actual sequence, start to finish, for one campaign.

Step 1 — Decide what the video is for. There are only three useful jobs for video in an email: show the product doing something photos can't, put a human face on a message that reads cold, or demonstrate a process. If your video does none of those, send the email without it.

Step 2 — Write a 20–35 second script. Shorter than you think. The email's job is the click; the video's job is on the landing page. What goes in the email is the teaser.

Step 3 — Generate the video. For product-led sends, reference-driven generation from your actual product images keeps the packaging honest. For founder or team messages, an avatar keeps the same face across every send in a sequence — the AI avatar generator covers the talking-head case and the AI product video generator covers the product case.

Step 4 — Cut the 3-second email loop. The most visually distinct three seconds, not the opening — openings are usually a slow reveal. Pick the beat where something recognizable happens.

Step 5 — Export the GIF and the poster frame. Same source, two assets. The poster is your fallback and your social static.

Step 6 — Composite the play button. A play triangle over the image is the single highest-impact detail; it signals "video" and lifts click-through in every test I've run. Bake it into the image rather than overlaying HTML, which breaks in half the clients.

Step 7 — Wire the link. The GIF links to a landing page with the full video above the fold, autoplaying muted. Don't link to YouTube from a marketing email if conversion matters — you've handed the visitor to a recommendation feed.

Step 8 — Set the alt text. Roughly half of Outlook recipients see images off by default, and your alt text is the email in that case. Write a sentence, not a filename.

That's a two-hour run the first time and closer to 40 minutes once the workflow is saved.

Where video belongs in the lifecycle

Not every send needs it. Ranked by how often it earns its place:

• Welcome sequence, email 1. Highest engagement moment in the lifecycle. A 25-second founder intro does real work here.
• Product education / onboarding. The strongest use case — showing a setup process beats describing it and deflects support tickets.
• Abandoned cart, second touch. Not the first, which should be a plain reminder. The second can carry a 15-second demo answering the doubt that caused the abandonment.
• Launch and announcement sends. Obvious fit, and the one most teams already do.
• Re-engagement. A face and a voice cut through better than another discount code.
• Transactional and routine newsletters. Almost never. Don't put marketing video in a receipt.

The pattern is the same as everywhere else: video wins where it removes doubt or adds humanity, and loses where it's ornament.

The personalization question

Generated video makes per-segment variants cheap, which raises an obvious question: how far should you personalize?

Segment-level is almost always the right granularity — five variants for five customer types is a good use of an afternoon. One-to-one personalized video at list scale is technically feasible via the API but gets mixed reactions, and recipients who notice a machine-generated personal video often react worse than to a good generic one.

Where per-recipient dynamic media clearly works: data the recipient already knows you have. A usage summary, an order-specific product shot, a renewal date. That reads as service, not as a trick. AI email marketing personalization covers the segmentation logic in more depth.

Measuring it properly

Video in email has a measurement trap: the metric that moves is click-through, and click-through improvements are easy to fake with a more aggressive play button. Track the chain.

• Image-on rate for your list. If it's low, your video strategy is really an alt-text strategy.
• Click-through on the video block specifically, not the email overall. Most ESPs support per-link tracking.
• Landing page play rate. Clicks on the GIF that don't become plays mean the page is misaligned with the email's promise.
• Watch-through on the page, which tells you whether the content is good rather than whether the thumbnail was enticing.
• Conversion, not clicks. A block that lifts CTR 30% and conversion zero is a curiosity-gap problem.
• Deliverability. Watch for inbox-placement changes after introducing heavier payloads.

Run the first test as video-block vs no-video-block on the same send, split evenly. Not a sequential before-and-after — email metrics drift with list health and season, and sequential comparison will lie to you.

Making it repeatable

The reason this stops happening after two campaigns is that step 3 requires someone with a creative tool open. Fix that by saving the run as a reusable workflow: fixed brand references, fixed aspect and duration, variable script and product. Then the per-campaign work is writing 30 seconds of script and picking a product — genuinely a coffee-break task.

For teams sending weekly, put it on a schedule. The workflow runs, the assets land, the email producer picks them up. That's the difference between video-in-email as a program and video-in-email as a thing you did once in Q2. The email plus video promotion loop covers how those assets feed back into the rest of the channel mix.

FAQ

Can you embed a real video in an email?

Not reliably. HTML5 video doesn't render in Outlook or the Gmail web client, and AMP for Email video requires sender allowlisting with limited support. The working approach in every client is an animated GIF or static poster with a play button, linked to a landing page hosting the actual video.

How large should an email GIF be?

Under about 1MB, ideally 300–600KB. Three to five seconds at 8–12 fps with a tight crop usually lands there. Heavier files add mobile load time, and unusually large image payloads can affect inbox placement with some filters.

Where should the video link go — YouTube or my own page?

Your own landing page, with the video above the fold and autoplaying muted. Linking to YouTube from a conversion-focused email hands your recipient to a recommendation feed full of other people's content. Use YouTube for discovery, not as the destination of a paid-for email click.

Which emails benefit most from video?

Welcome sequences, product onboarding, second-touch cart recovery, and launches. Video earns its place when it shows something photos can't, humanizes a cold message, or demonstrates a process. It doesn't belong in transactional mail or as decoration in a routine newsletter.

Is personalized video at scale worth it?

Segment-level variants — five versions for five customer types — are cheap to produce and reliably useful. Per-recipient personalized video is technically possible via the API but gets mixed reactions, and the novelty decays. Dynamic media built from data the recipient knows you have, like an order or usage summary, is the version that consistently reads as service rather than gimmick.

---

Build one campaign's assets end to end before you commit to a program — script, generate, cut the loop, ship it against a no-video control. If it earns its click, save the run as a workflow and the next twelve campaigns cost you a script each.