YouTube Thumbnail A/B Test
How YouTube's Test & compare works, what to test first and how to read the result without fooling yourself.
Two Variants of the Same Thumbnail
An A/B test changes one thing at a time. Here the only difference is the expression and where the number sits.
No credit card required · 5 free credits each month
A thumbnail A/B test shows two or more versions of a thumbnail to similar viewers and compares the results. YouTube Studio has a built-in feature for this, called Test & compare. This page walks through one test from start to finish, explains how to judge the result, and says when a test isn't worth running.
What YouTube's built-in test does
With Test & compare, you upload up to three thumbnails for a video. YouTube shows them evenly to viewers and, at the end, names a winner. A detail that surprises people: the winner is chosen by watch time share, not by click-through rate alone. A thumbnail that gets many clicks from viewers who leave quickly can lose to one that gets fewer clicks from viewers who stay.
- A test can take from a few days to about two weeks.
- It can end without a clear winner. That's a result too: the versions didn't differ enough to matter.
- You decide what to do with the result. Nothing forces you to use the winner.
- Access depends on your channel having advanced features turned on. Look for the option when you edit a video in YouTube Studio.
YouTube has extended the feature over time, including to titles, so check Studio for what your channel can currently test. Rules like these change, and YouTube Help has the current version.
Running one test, step by step
The example: a video about reorganizing a small kitchen. The current thumbnail shows the finished kitchen with the words "SMALL KITCHEN".
- Choose a video with steady impressions. A test needs enough viewers to compare. A video that is seen only a handful of times a day won't give a clear answer.
- Decide what one question you're asking. Here: does a close-up of the messy "before" cupboard work better than the finished kitchen?
- Make versions that differ in that one thing. Keep the text, colors and layout the same. Change only the image.
- Upload them in Test & compare. Add up to three. Two is easier to read.
- Leave the video alone. Don't change the title or promote it differently in the middle of the test. That muddies the result.
- Read the result when it ends. Note the winner, or the fact that there wasn't one, and write down what you changed.
What to test first
Change one variable at a time, so you know what caused the difference. Start with the ones that change the thumbnail most.
| Variable | Version A | Version B | What you learn |
|---|---|---|---|
| The subject | The finished result | The problem or the "before" | Which story pulls people in |
| Text | No text | Three words | Whether words help or crowd it |
| Face size | Small face | Large face | Whether a close-up matters for your audience |
| Background | Real location | Plain dark background | Whether a clean background helps |
| Accent color | Yellow | Red | Whether the color change makes a difference |
Reading the result fairly
- A small difference with few viewers can be noise. If the result says there's no clear winner, believe it.
- One test shows one thing about one video. Repeat a finding on a second video before you build a rule on it.
- Look at watch time and click-through rate together. A thumbnail that attracts the wrong viewers isn't a win.
- The result tells you what worked for your audience this time. Another channel's winner may not be yours.
Three outcomes and what to do with each
- A clear winner. Keep it. Then ask what the winner changed and whether the same change might help your other videos. Test it on a second video before you treat it as a rule.
- No clear winner. Your audience didn't care about the difference. That's useful: you can stop spending time on that variable and test something bigger, such as the subject of the image.
- A mixed result. One version gets more clicks and another holds viewers longer. Think about which viewers you want, and consider whether the thumbnail promises more than the video delivers.
Keep a test log
After a few tests, memory fails. A simple log keeps the lessons. For each test write the video, the date, what changed, what the versions were, the result and one line on what you'll do next time. A spreadsheet is enough. After ten tests, patterns show up that no single test could show you, such as "close-ups seem to matter on my tutorials, text doesn't".
Testing titles
YouTube has extended Test & compare to titles as well, though what your channel can test depends on its access, so check Studio. The same rule applies: change one thing and let the test finish. The thumbnail and the title are read together, so when you test one, keep the other fixed. The YouTube title vs thumbnail page explains how the two interact, and the YouTube title generator can supply title versions to try.
If you can't use the built-in test
You can compare by hand: use one thumbnail for a period, switch to the other for the same length of time, and compare. It's a weaker test, because the two periods differ in more than the thumbnail. Weekends, a trending topic or a change in how YouTube shows the video can all move the numbers. Keep the periods equal, avoid the first days after upload and treat the result as a hint, not proof. YouTube thumbnail CTR explains how to read the numbers.
When a test isn't worth it
- Your channel gets few impressions. The result will be too noisy to trust.
- Both versions are weak. Fix the basics first with the YouTube thumbnail checklist.
- You're about to change the title as well. Do one thing at a time.
Making the versions
Versions for a clean test should differ in one thing. If you use Titles.video, remember that each new image is built fresh and can differ in more than one way, so two generated thumbnails compare two concepts, not one variable. That's fine for a first look at which direction works. For a single-variable test, edit one base image by hand, or use Improve mode on the same photo and then change only the text. The steps are on the YouTube thumbnail maker page, and YouTube thumbnail design lists the principles worth testing.
Frequently Asked Questions
How do I A/B test a YouTube thumbnail?
Use Test & compare in YouTube Studio if your channel has access. Upload up to three thumbnails that differ in one thing, leave the video alone and read the result when the test ends.
What does the YouTube thumbnail test measure?
YouTube picks the winner by watch time share, not by click-through rate alone. A thumbnail that gets clicks from viewers who leave quickly can lose to one that keeps viewers watching.
How long does a thumbnail test take?
From a few days to about two weeks, depending on how many people see the video. It can also end without a clear winner.
How many impressions do I need?
YouTube doesn't publish a fixed number. The more impressions the video gets, the clearer the result. Channels with few impressions often get inconclusive tests.
What should I change between versions?
One thing at a time, such as the subject, the text, the face size or the background. If several things change, you can't tell which one made the difference.
