Video in the help center: high demand, heavy upkeep
Customers say they'd rather watch than read, and for show-me problems they're right. But video is the most expensive content you'll ever maintain, and search can't watch it. A decision rule.

At some point a leader on your team will open a competitor's help centre, find a tidy library of two-minute walkthrough videos, and come back asking why yours is a wall of text. It's a fair question. It just isn't the right one. The right question is: which of our problems is video actually good at, and who is going to keep fifty clips true after the next redesign?
The demand is real
Start by conceding the popular point, because the data behind it is consistent.
That comes from a small marketing-flavoured survey — 266 respondents — so hold it loosely, but Wyzowl has asked versions of this question for twelve years and the preference has never flipped. And behaviour backs the stated preference:
For a certain class of problem, the preference is also simply correct. "Where is the export button in the new layout" is a show-me problem. A thirty-second screen recording answers it faster than three paragraphs and a screenshot ever will. When the content of the answer is motion — a sequence of clicks, a physical setup, a gesture — video isn't a nice-to-have format. It's the native one.
What video quietly costs
Here is the part the competitor's tidy library doesn't show you. When a text article goes stale, you edit the sentence and hit save. When a video goes stale, there is no diff. You re-record, re-edit, re-caption, re-upload — or, in practice, you don't, and the clip sits there confidently demonstrating a UI that no longer exists. Screenshots age quietly; screen recordings age loudly, in motion, with a voiceover asserting things that are no longer true.
We've written about the maintenance debt of a text knowledge base — the review cadences and ownership that nobody budgets. Now scale the problem:
Every item on a knowledge shelf is a standing maintenance liability, and a video is the most expensive kind per item: highest production cost, highest update cost, and the only format you cannot patch. If your product ships UI changes monthly, a video library without a re-recording budget is a stale-content factory with good production values.
Screenshots age quietly. Screen recordings age loudly, in motion, with a voiceover asserting things that are no longer true.
Search can't watch your video
The second hidden cost is findability — already the place where self-service goes to die:
A video makes that worse unless you do deliberate work, because search engines, help-centre search bars, and — increasingly important — your AI agent all read text. An untranscribed video is invisible to every retrieval system you own. As we argued in your AI inherits your knowledge debt, the bot answers from what it can read; a walkthrough that exists only as pixels and audio contributes nothing to the shelf your AI reads aloud. The transcript isn't an accessibility checkbox. It is the searchable, retrievable, patchable asset — the video is the friendly rendering of it.
And the stakes of a failed self-service attempt keep rising with your youngest customers:
The decision rule
Use video where motion is the content; use text where structure is. Concretely:
- Video earns its keep on: multi-step UI flows, first-run setup, physical installation, anything customers currently answer with "can you just show me?"
- Text stays king for: reference material, policies, troubleshooting trees, anything customers skim, search, or need to copy-paste from. Nobody scrubs a video for a tax-form field name.
- Every video ships with its transcript on the page — for search, for the AI, for the customer on a train without headphones.
- Every video gets an owner and a review-by date, exactly like an article, plus one honest extra field: the UI surfaces it shows. When those surfaces change, the clip goes into the re-record queue automatically.
- Keep them short and chaptered. The preference data is for clips measured in seconds and low minutes, not webinars.
- Pair every video with text steps on the same page, so the page survives its video's death — a stale clip you had to pull shouldn't take the answer down with it.
- Budget video at a multiple of text in your maintenance plan. If the multiple isn't affordable at your release cadence, make fewer videos. An accurate wall of text beats a beautiful library of lies.
Run your existing library through the KB health audit with those rules in hand, and read where knowledge bases go to die before you commission the next fifty clips. The customers asking to be shown deserve videos that show the truth — which is to say, videos somebody is paid to keep true.