Video on Your Website: What Actually Works

Video on a website works when it loads fast, plays muted, carries captions, and is the right length for where it sits. Done well, video lifts dwell time and engagement. Done carelessly, it slows your page, annoys visitors, and quietly costs you sales. Here is what the evidence actually says.

Should video autoplay with sound?

No. Every major browser — Chrome, Safari, and Firefox — blocks autoplay with sound by default (Chrome Developers; WebKit; MDN, autoplay docs). The only reliable autoplay is muted. The safe, cross-browser pattern is a muted, inline, looping background clip, with a static poster image as the fallback.

This isn't just a technical limit. Usability research from the Nielsen Norman Group warns that "users don't appreciate being surprised by video or audio content that begins playing without their consent" (NNGroup, video usability). So background video should be ambient — no sound, no narration, just atmosphere.

For anything substantive — a product demo, an explainer, a testimonial — use click-to-play with a strong thumbnail. People who choose to press play are already more engaged. We treat that split as a build rule, not a stylistic choice: ambient loops autoplay muted, real content waits for a click.

How long should a website video be?

Length depends entirely on where it lives. Wistia analysed more than 13 million videos and 79 million hours of viewing for its 2026 State of Video report, and the pattern is clear: shorter wins above the fold, longer works below it for visitors who have already self-selected.

• Hero / background loop: 10–30 seconds, seamless, ambient.
• Product demo: 1–3 minutes. Videos in this range average roughly 50% engagement (Wistia, 2026).
• Explainer: 2–5 minutes. How-to content holds attention especially well — around 74% engagement (Wistia, 2026).
• Testimonial: 30 seconds to 2 minutes. They lose viewers after the two-minute mark.
• Deep dive / webinar: below the fold only. Long videos earn just 9% call-to-action click-through, but the people who stay are highly qualified (Wistia, 2026).

The instinct to put one long, do-everything video at the top of the page is usually wrong. Lead with something short. Save the depth for visitors who scroll.

Do captions actually matter?

More than most people expect. Captions are one of the highest-return additions you can make to a video.

Always-on captions increase view time by 12% and lift completion rates so that 80% more viewers watch to the end (3Play Media, aggregating Facebook and social-video data).

And captions aren't only for people who can't hear: about 80% of caption users are not Deaf or hard of hearing (Ofcom, via 3Play Media). They help in noisy places, silent offices, on muted phones, and for non-native speakers. On Facebook, captioned video has shown 135% more organic reach (3Play Media). Search engines can read caption text too, so it helps you get found.

There's a legal floor as well. Synchronised captions for pre-recorded video with audio are a WCAG 2.2 Level A requirement — the baseline standard for an accessible website. This is exactly the kind of detail we build in by default rather than bolt on later.

What's the cost of a YouTube embed?

Heavier than it looks. A standard YouTube iframe pulls in roughly 540 KB of JavaScript on initial page load (Chrome Lighthouse, third-party facades). On a page that should feel instant, that's a real drag on your loading speed.

The fix is a "facade" — a lightweight placeholder that only loads the full player when someone clicks. The widely-used lite-youtube-embed, built by a Chrome DevTools engineer, cuts that initial load to about 3 KB: a 224× improvement (paulirish/lite-youtube-embed). Lighthouse specifically recommends this approach.

For a short hero loop, we go further and self-host an MP4 with a WebM alternative — zero third-party JavaScript, and the WebM file is around 30–40% smaller than MP4 at the same quality. This is exactly the kind of number we design around when a page's speed is on the line.

Which type of video converts?

Match the video type to the job:

• Animated explainers suit abstract products and "how it works" sections. 96% of people say they've watched an explainer to learn about a product, yet only 23% of marketers use animation versus 51% live-action (Wyzowl, 2026) — so it's a real chance to stand out.
• Talking heads build trust. Consumers favour authentic over polished, and faces draw the eye — in one Instagram study, images with human faces earned 38% more likes and 32% more comments (ResearchGate, "Faces Engage Us"). The principle carries to thumbnails.
• Screencasts sell software. 80% of people say they've downloaded an app after watching a demo video (Wyzowl, 2026).

Honest note: there's no published A/B test proving a specific conversion lift from background hero video, and the famous "80% lift" figure traces to a defunct company and can't be verified. We don't repeat numbers we can't stand behind. What the data does support is that 82% of video marketers report video increased dwell time (Wyzowl, 2026, self-reported) and 89% of consumers say video quality affects how much they trust a brand (Wyzowl, 2026).

FAQ

Does video slow down my website? It can. A standard YouTube embed adds about 540 KB on load (Chrome Lighthouse), and an uncompressed background loop can be several megabytes. With a facade, self-hosted compressed files, and lazy loading, the cost is minimal until a visitor actually wants to watch.

Should I autoplay my hero video? Only muted, and only if it's purely ambient. Browsers block autoplay with sound, and visitors dislike being surprised by it (NNGroup). Anything with a message belongs behind a click-to-play thumbnail.

Do I really need captions? Yes. Captions lift view time by 12% and help 80% more viewers finish (3Play Media), most caption users aren't hard of hearing, and synchronised captions are a WCAG 2.2 Level A accessibility requirement.

How long should a homepage video be? For a hero loop, 10–30 seconds. For a click-to-play explainer, aim under three minutes — shorter holds more viewers above the fold (Wistia, 2026).

We build fast, conversion-tuned websites informed by exactly this research — muted autoplay, captions by default, facades over heavy embeds, the right length in the right place. If you're weighing up video for your site and want a straight answer about what's worth it, book a call.