A recorded demo proves that a workflow exists and shows what the interface looks like when everything goes right. It cannot prove accuracy, speed, coverage or fit with your code, because the data, the timing and the edits are all under the vendor's control. Use it to build a test plan, not to decide.
What a recording does show
Recordings are good at a few things. You can see how many steps a task takes, how findings are presented, what a developer would see, and how much jargon the product uses. That is real information about design, and it costs you nothing to watch.
It also lets you pause and read. Pause on any screen with a list of findings and note the columns, the filters and what is not shown.
What it cannot show
| Claim you may infer | Why a recording cannot support it |
|---|---|
| The scan is fast | Cuts and edits hide wait time |
| Results are accurate | The sample project was chosen to look good |
| It works with my language | One demo project covers one stack |
| Fixes are correct | The fix shown may have been prepared in advance |
| It scales | Demo repositories are small |
None of this implies bad faith. It is simply how demos are produced.
Turn the recording into questions
While you watch, write down every statement that begins with "it automatically" or "in seconds". Each becomes a task for your trial. For example:
- "It automatically finds the vulnerable function." Ask to see it on your repository, with a function you picked.
- "It opens a fix in seconds." Time it on your build, with your tests.
- "It prioritises what matters." Ask for the reasoning behind each priority on a sample you know.
Then send the list to the vendor and ask them to run these live, with your input, rather than replay the video.
Live versus recorded
A live session on your data adds three things a recording lacks: your code, unscripted questions and real latency. If a vendor will only show recordings, that is worth noting. If they will run a live session but only on their sample, ask again for yours. Your own free scan, on a manifest you choose, is often the quickest way to see actual output.
A short checklist for watching one
Before you press play, write down your three biggest worries. While you watch, mark the timestamp where each one is addressed, or not. Afterwards, send the vendor only the unanswered ones. This keeps the follow up call short and focused on what the recording could not show, rather than on a repeat of the tour.
Where Safeguard fits
Safeguard provides a free scan at gold.safeguard.sh/scan, so you can see real output on your own manifest before any call. Compare what you see with any recording, from any vendor.
Quick check
Are recorded demos misleading?
Not by nature. They show the intended flow well. The risk is reading them as evidence of accuracy or speed, which they cannot show.
What should I ask for instead of a recording?
A live run on a repository you choose, with your questions, timed, and with output you can export.
Last reviewed: 19 October 2026.
