Drop version A and version B, name your topic, and see both in the feed next to the real top videos — then keep the one that doesn't blend in. Live tests come after publishing; this one comes first.
Thumbnail A
Thumbnail B
Only one version? Use the thumbnail tester. Need ideas for the text? Browse thumbnail text from high-performing videos.
A live A/B test needs real impressions: YouTube Studio's own Test & compare shows your versions to viewers after the video is published and reports which one earned more watch time. That is the only test that measures clicks. A pre-publish test does a different job — it cuts your ideas down to the versions worth testing live, before any impressions are spent on a weak one.
A thumbnail that looks great on its own can disappear next to the videos it actually competes with. Put both versions in the same feed of real results for your topic and see which one your eye finds first.
Comparing two versions full-screen, side by side on a white page
Both in a feed of the top videos on the topic, at phone size
If A and B differ in background, text, face and colour at once, you learn nothing you can reuse. Change the one element you are unsure about — the background colour, the text or the subject — and keep the rest.
B is a completely different design
Same shot and text; B uses a light background instead of a dark one
Before a video is live nobody can measure its click-through rate. What you can measure is whether a version looks like everything around it. If one version is darker, flatter or greyer than most of its neighbours, it is the riskier one.
Picking the version you like more on your own screen
Picking the version that differs more from the feed, then checking it is readable on a phone
Rewrites are illustrative examples, not measured results.
Yes. Comparing two versions in the feed and the measurements are free with no sign-up. An optional AI review of a design uses credits; a free account includes 15.
YouTube's Test & compare runs after you publish and shows your versions to real viewers, so it measures actual results. This runs before you publish, against the videos yours will compete with, so you can drop weak versions before they cost impressions. Use both.
No. It shows both versions in YouTube-style feeds and measures how much each differs from the real thumbnails around it in brightness, contrast and colour. Click-through rate can only be measured on a live video.
No. Both images are read and measured in your browser and never sent to our servers.
They are real YouTube results for your topic from the YouTube Data API, ordered by how far each beat its own channel's median views.