A pasted YouTube link is a tax on your own reach#
For B2B software, LinkedIn is not one channel among several; it is the channel. Wistia's 2026 State of Video, built from the videos it hosts, puts 81% of teams posting video to LinkedIn, ahead of YouTube at 76%, and names LinkedIn B2B's top video-sharing channel6. That is where your buyers already scroll. The mistake is in how most teams put a video in front of them: they paste a YouTube link into the post and call it done.
That link is a reach tax, and the reason is in LinkedIn's own feed mechanics. Since 2020 the feed has ranked updates partly on dwell time, the seconds a member spends on an update, through a model that predicts the probability of a skip1. LinkedIn's engineers spelled out why they trust dwell over likes and clicks: most members view an update without ever clicking, liking, or commenting, so those actions are missing for the silent majority; they are binary rather than graduated; and a click can be a "click bounce" that means the opposite of interest. Dwell time, by contrast, is "always measurable" and a "real-valued measure of engagement."
Read that back with a YouTube link in hand. The best case for a pasted link is a click, and a click sends the viewer off LinkedIn to youtube.com, ending the dwell the post is ranked on. A native upload does the opposite: it autoplays in place, and every second watched is dwell the feed counts in your favor. The pillar's channel matrix already flags that a native upload out-reaches a pasted link; the dwell model is why. Upload the file. If you need the YouTube version discoverable, drop that link in the first comment, where it will not compete with the clip for reach.
The LinkedIn upload spec, in one table#
LinkedIn accepts a wide envelope and rewards a narrow part of it. The full spec, current as of July 2026, with the B2B in-feed pick called out:
| Property | What LinkedIn accepts | The B2B in-feed pick |
|---|---|---|
| Aspect ratio | 16:9, 1:1, 4:5, 9:16 | 1:1 or 4:5 |
| Length | 3 seconds to 30 minutes | 15-30 seconds |
| Resolution (square) | 360x360 to 1920x1920 | 1080x1080 |
| Resolution (4:5) | 360x450 to 1080x1350 | 1080x1350 |
| File size | 75 KB to 500 MB | comfortably under |
| Container / codec | MP4, H.264 or VP8 | MP4 / H.264 |
| Frame rate | under 30 fps | 30 fps |
Those numbers are LinkedIn's2; the right-hand column is the opinion. Two cells decide the reach. Length first: LinkedIn recommends 15 to 30 seconds so a clip qualifies for every placement, and the retention data agrees that a short clip is watched while a long one is merely sampled. Vidyard's 2024 benchmark of 943,305 videos found 65% of viewers stay to the end of a sub-minute clip against 20% past twenty minutes5.
Aspect is the cell teams get wrong. A 16:9 clip is the shape of a YouTube master, not a feed clip. At the fixed width of the feed column, a 1:1 or 4:5 video is simply taller than a 16:9 one, so it occupies more of a vertical phone screen and holds the scroll a beat longer, which the dwell model rewards directly. The square cut is not a matter of taste; it is more pixels of screen time per scroll. The launch-clips piece carries the full cross-platform cut table; this is its LinkedIn column, with the reasoning laid under it.
Muted autoplay makes your first two lines the pitch#
A LinkedIn video autoplays in the feed with the sound off. LinkedIn says so plainly and, for that reason, calls captions strongly recommended because "videos autoplay without sound"3. So the first view of your demo is silent, and two things carry it.
The first is the post copy above the video. LinkedIn collapses that text after the first two or three lines behind a "see more" fold, so those lines are not a caption on the video, they are the headline that decides whether the muted clip gets watched at all. Lead them with the outcome or the problem in a plain sentence, not with "excited to share." The second thing carrying the silent view is the video itself: its opening frame and its burned-in captions. The wind-up open, a logo card and a face promising to walk you through it, spends the one silent second the feed grants you; open on the product mid-task instead, with the words already on screen.
Those words have to be burned into the pixels. LinkedIn re-encodes what you upload and a reshare can strip a sidecar caption file, so the only text that reliably survives is the text rendered into the frame, and the case for captions as pixels covers the timing and the rendering in full. The opening move is its own small craft, and the catalog of hooks that survive a muted scroll applies to the LinkedIn frame as much as to any feed.
The document post for the weeks you don't film#
Not every week produces a new recording, and LinkedIn has a native format for the gap that most B2B teams underuse: the document post. Upload a PDF and LinkedIn lets members read it inside the feed, page by page, without leaving the platform to do it4. It is not video, but it plays by the same ranking rule. Every page a reader turns is another second spent on the update, so a multi-page document banks dwell the way a native clip does and never sends the reader off-platform to do it.
For software, the document post has a quiet advantage: it is made of the stills you already have. A demo pipeline that captures named screenshots of each step, the empty state, the one action, the populated result, has a five-page document sitting in its output folder. Turn "here is the workflow" into five frames with a line of text on each and you have a post that carries a product story with no narration, no audio, and nothing to mute. On a week with no new feature to film, that is the format that keeps the account posting without a shoot, and it feeds the same dwell signal the video cut does. It is the answer to the false choice between filming every week and going quiet.
One recording feeds every LinkedIn slot#
Line up what LinkedIn actually wants from a single demo of one workflow: a native square or 4:5 cut, 15 to 30 seconds, captions burned in, for the in-feed post; a 16:9 master for the YouTube link you drop in the comments; and a document post of that same flow's still frames for the weeks between films. Three deliverables, one capture.
Cutting those three by hand is three exports every time, and it is worse than that, because LinkedIn keeps adjusting what the feed rewards and your product keeps shipping, so each deliverable is re-made on both clocks. Treat the demo as a spec instead of a captured file and the arithmetic changes: a square reframe becomes a render at a different output size rather than a crop that discards a third of the frame, and the document-post stills regenerate from the same flow when the UI moves. This is how aidemo, our open-source engine, is built: an agent authors the storyboard, a deterministic run replays it in real Chrome, and one take renders at whatever aspect a slot asks for and exports the named stills a document post needs. Its honest limits are real: it captures a browser tab and nothing outside one; the storyboard is written as text by an agent, not clicked together in an editor; and there is no visual timeline to nudge clips on. That makes it the wrong tool for a one-off phone-shot clip and the right one when the same workflow has to fill three LinkedIn slots and stay current for a year. Who is watching still sets the edit, and on LinkedIn that is usually a buying committee; the seven axes that separate a B2B cut from a self-serve one decide the tone before any of this spec work begins.
Sources#
- LinkedIn Engineering - Understanding LinkedIn's use of dwell time for feed ranking (2020)
- LinkedIn Help - Video ad specifications (aspect ratios, length, resolution, file size)
- LinkedIn Marketing - Video ads (autoplay muted, captions strongly recommended)
- LinkedIn Marketing - Document ads (read documents directly in the feed)
- Vidyard - Video in Business Benchmark Report (sub-minute completion 65%)
- Wistia - State of Video: video marketing statistics (81% of teams post to LinkedIn)
FAQ#
Should I upload video natively to LinkedIn or post a YouTube link?#
Upload it natively. LinkedIn's feed has ranked partly on dwell time since 2020, and a pasted YouTube link's best outcome is a click that sends the viewer off LinkedIn, ending the dwell the post is measured on. A native upload autoplays in the feed and every second watched counts in your favor. If the YouTube version needs to be findable, put that link in the first comment rather than the post body, so it does not compete with the native clip for reach.
What size should a LinkedIn video be for B2B software?#
Square 1:1 or 4:5, not 16:9, for the in-feed post. At the feed's fixed column width a square or vertical clip is taller than a widescreen one, so it fills more of a phone screen and holds the scroll longer. Keep it to 15 to 30 seconds, ship it as an MP4 in H.264 under 500 MB, and save the 16:9 version for the YouTube master. LinkedIn accepts anything from 360x360 up to 1920x1920 for a square clip.
Do LinkedIn videos need captions?#
Yes. LinkedIn video autoplays in the feed with the sound muted, so the first view is silent and the on-screen text is the whole message. Burn the captions into the pixels rather than relying on a sidecar file, because LinkedIn re-encodes uploads and a reshare can drop a separate caption track. A clip that depends on its audio to make sense is a clip most of the feed will scroll past without ever hearing it.
