Does video on a landing page increase conversions? What the evidence shows, when it hurts, and how to test it.

By Zach Wennstedt, founder · Updated November 6, 2026 · 9 minute read

Does video increase landing page conversions?

Sometimes, and the honest answer depends on what the video does for the visitor. If a visitor already understands the offer, a video is one more thing between them and the button. If the offer is hard to explain in a headline, a short video might close the gap. Those are different jobs, and most “add a video” advice treats them as one.

Here is what we could verify. There is one large observational comparison (Unbounce, 2021), one famous vendor A/B test (Vidyard, 2011), and a lot of repeated percentages that lead back to one of those two or to nowhere. We checked each source live on November 6, 2026 and link them all below. Where a number comes from a chart we couldn’t read, we say so instead of guessing.

What did the biggest comparison find?

In August 2021 Unbounce published an analysis of its customers’ landing pages, comparing pages with no video, an embedded video and a background video. It used three separate time frames between late 2020 and early 2021, each with approximately 35,000 pages, so a seasonal quirk wouldn’t drive the result. It split the pages by industry, by conversion goal (form fill or click-through), by traffic channel and by device. Here is what it reported:

• Form-fill pages: very little difference with or without video. Background video did slightly better; embedded video did not.
• Click-through pages: a negative difference the authors called even more striking than the form-fill gap.
• Mobile and desktop: video pages had lower conversion distributions on both.
• Industries: across ten industries video was generally neutral or negative. Real estate showed a strongly negative effect. Finance and insurance had a higher median with video, but most pages in that group performed similarly either way.

Read the limits before you quote it. This is vendor data from one landing page platform, not a controlled experiment. Pages that use video may differ from pages that don’t in ways the analysis can’t see (traffic quality, offer, price). The exact median conversion rates appear only in charts, so we don’t quote them. The background-video cohort was small and the background-video click-through group was too small to report. Unbounce’s own conclusion was measured: video isn’t always bad, so test pages with and without it instead of ripping out video that works.

What the data does support is narrower than “video hurts” and the opposite of “video lifts.” Video is not a reliable conversion lever by itself.

Why do the big video lift claims fall apart?

The most-cited positive result is a 2011 Unbounce case study of Vidyard, a video platform, testing its own sign-up page. Per the case study, the no-video page converted at 6.5%, the page with an embedded video at 11% (a 69% lift) and the page with the video in a lightbox pop-up at 13% (a 100% lift). It is a real test and a useful idea, because the click-to-play lightbox beat the always-visible embed. But look at what the page does not say: no traffic numbers, no test dates, no statistical significance. It was written by a company that sells video, about a product that sells video, 15 years ago.

Even the details drift. A third-party case-study summary we found lists different numbers for the same test (6%, 12% and 9%), which is a good reminder that copied statistics decay. Other “video lifts conversions by X%” figures we searched for traced back to vendor blogs without published methods. Our rule on this site is the same as in our page speed article: sort evidence by how much it proves, and don’t build a decision on a number you can’t open.

When does video help?

The evidence doesn’t hand us a rule, so these are hypotheses worth testing, not findings. They share one trait: the video does work the page can’t.

1. The offer is hard to picture. Software, a new kind of service, a physical product people want to see in use. A 45-second demo can replace three paragraphs.
2. Trust is the barrier. A short customer or owner clip lets a visitor hear a real person. Our trust signals article covers what evidence there is for that kind of proof.
3. The visitor chooses to watch. Vidyard’s lightbox result, whatever its flaws, points the same way: click-to-play lets the interested visitor in without taxing everyone else.
4. The page is long and complex. A summary video near the top can orient people before they scroll.

When none of those is true, the headline, the proof and the form usually deserve the space first.

When does video hurt?

When it slows the page. Speed is the one mechanism with randomized evidence. In Vodafone’s 50/50 test on paid traffic, the faster version improved Largest Contentful Paint (LCP) by 31% and sales by 8%, with lead-to-visit up 15% and cart-to-visit up 11%. Google’s web.dev guidance says the first frame of a video, once painted, can be the LCP candidate, and that a video with autoplay begins downloading immediately, even if it’s outside the first screen. A heavy hero video can put your LCP past the 2.5 seconds Google calls good. Core Web Vitals are judged at the 75th percentile of real visits, so your slowest quarter of phones sets the grade.

When it competes with the action. Unbounce’s sharpest negative was on click-through pages, where the job is one click. A video that holds attention on the video is not holding it on the button.

When it autoplays with sound, or tries to. Chrome always allows muted autoplay, but allows sound only after the visitor has interacted with the site or met other conditions, so don’t assume playback will start. Chrome’s guidance is to check the promise returned by play() and show a Play button if it’s rejected. A video that fails silently is a dead rectangle.

When it’s the wrong content. A founder talking for three minutes, a stock-footage montage, a video with no captions for the visitor in a waiting room with the sound off. If you wouldn’t put the same words as text above the fold, they don’t get better in a player.

Which way of adding video is safest?

The method matters as much as the decision. These are the four we see, ranked by how much risk they add to speed and to the page’s main job. The speed notes follow web.dev’s video guidance.

Whichever you choose: give the poster an explicit width and height so the layout doesn’t shift, add captions, and keep the video file small. One more thing from web.dev: preloading a poster that isn’t actually your largest element can slow the real LCP, so measure instead of assuming.

How do you test video on your own page?

This is the part most “does video work” articles skip, and it’s the only way to get an answer about your visitors. The approach is the same one we use on every page: change one thing, split traffic at random, wait for enough data, and judge by the outcome you’re paid on. Our A/B testing page explains the method, and the A/B test simulator lets you see how noisy small samples are.

1. Write the hypothesis. “For paid visitors on this page, a click-to-play demo under the headline will raise form submissions because the offer is hard to picture.” If you can’t say why, wait.
2. Test video against no video, not against a redesign. Same headline, same form, same offer. Only the video block changes.
3. Pick the metric before you start. Calls, booked jobs or qualified leads, not plays or watch time. A video can be watched a lot and still lose.
4. Check the traffic. Optimizely’s documentation gives a worked example: a 15% baseline, a 10% relative lift to detect and 95% significance needs about 8,000 visitors per variation. Lower baselines or smaller lifts need more. If your page gets a few hundred visits a month, a clean split test isn’t realistic; decide on speed, clarity and the evidence above, and use before-and-after with care.
5. Guard speed. Record field LCP for each version. If the video arm is slower, you aren’t testing video; you’re testing video plus delay, and the Vodafone result tells you which way that leans.
6. Run full weeks. Include weekends and paydays, and don’t stop the moment one arm pulls ahead.
7. Segment mobile and desktop. Unbounce saw the video effect on both, but your mix may differ.

If the test says no difference, you’ve learned something useful: you can cut the video and the megabytes, or keep it for brand reasons knowing it isn’t costing you.

What do we do on service-business pages?

For a plumber, a law office or a local clinic, the first screen should carry the offer, the proof and the call or form. We put video lower, click-to-play, with a real poster and captions, and only when it does a job: a short walkthrough of how the service works, or a real customer. We would not autoplay anything above the form without a test behind it.

We make video ourselves, too, so we aren’t anti-video. We produced a 26-second TV spot for Search Converts, and video production is part of what we offer. Each claim in our spot came from our own live site, and we haven’t published conversion results for it, because we haven’t earned any claim like that. You can see how a spot like that is built at Madwackd’s Vault. A spot for broadcast or social and a video on a landing page are different tools with different tests.

If you want a second pair of eyes on whether video belongs on a particular page, a landing page optimization review starts with the page’s speed, its message and its form, and video comes after those.

What should you do this week?

1. List every page with a video and its field LCP on mobile.
2. Switch any autoplay-with-sound or heavy hero video to click-to-play with a sized poster and preload set to none.
3. Pick your highest-traffic page with a video and write a one-sentence hypothesis for why it should be there.
4. If traffic allows, run video against no video for four weeks, judged on calls and qualified leads.
5. If traffic doesn’t allow a test, fix speed first and decide on clarity.