If I Started A/B Testing Thumbnails in 2026, I'd Do This

[00:00] I have AB tested over 200 thumbnails this year and I want to share with you the most important things I've learned so you don't make the same mistakes as me. At first, I want to start off with a mistake that too many people make, that you should

[00:12] not make because it will actually damage the information you get from using YouTube's AB testing tool. So if you don't know, YouTube has a tool where you can test multiple thumbnails at the same time up to three and then it will show them to the audience and then it will tell you

[00:26] which one won. Having a metric called watch time share percentage, which is a combination of not just how many people clicked but also how long they watched and they do this so that you don't just get

[00:40] incentivized to make clickbait because if they only showed you this, then you would just make the most clickbaity thumbnails possible. So what is this important mistake that you shouldn't make? When you initially run a test, at first it will just show you that there's no results.

[00:53] They don't have enough data yet. But if you check back in like an hour later, what you'll see is that it will give you some initial results. So maybe one thumbnails at 33, watch time share percentage, maybe one's at 41 and maybe

[01:06] one's like 28, okay? And so the big mistake here is that people see that wow, the second thumbnail has way higher than the other two. So let me just end the test now and move on to another test or move on with my life.

[01:19] The issue here is that if you were to wait it out, you would see that these numbers equal as or change as the YouTube test gets more data. And so if you were to give it another couple hours, maybe it turns out that this one's at 41

[01:31] percent while this one's at 28 and this one's at 33. And so the big mistake here is ending the test instead of letting it run till completion. Because this tool was actually pretty well made when it gives you a result, it's statistically

[01:45] significant. That means you can actually trust it. It's been shown to enough people where YouTube can say yes, this is the final result with 100 percent certainty. So don't end the test before it's actually finished.

[01:58] Otherwise, you'll be making decisions based off of false data. It hasn't collected enough data yet for this to actually be true. And I see this all the time, like an hour into the test, two will be crushing it and then I check back in and the test has been finished and it turns out thumbnail one is the winner

[02:12] and thumbnail two is actually the worst one of the batch. The second learning here is also another big mistake you should avoid doing. So it'll give you an example with a title test we did on KSI's channel.

[02:24] So for this video, about a day into the upload, we did two different titles because we wanted to test them out because the video wasn't doing so hot. The first title was this one right here. Parents pop the balloon, daughter's edition.

[02:37] Okay. So it was the original title. And then the second one that we did was a title like if her parents hate you, you lose the girl. Okay. So I'm just going to put like a new title over here. This was a day into the video being posted and lo and behold, the new title one, which

[02:52] made sense to me. It sounded like a title that would do way better, right? It's a lot clearer. Anybody can understand what the hell does pop the balloon me. But we didn't really see views go up after that title one.

[03:04] So a couple days later, I think about a week later, I ran the test again and lo and behold, they actually tied. I think the this title actually won by a little bit. And so this is actually a really important learning you should keep in mind that there are

[03:17] different stages of your videos life span. Okay. When you first upload the video, it's mostly being shown to your core audience to people who really like you, really like your content. And so any test you run during the first 48 to 72 hour period, even up to a week will be

[03:33] different results because it's people who love you. It will be different results than if you wait maybe after a week when it's going to be a broader group of people, maybe casuals, maybe the videos blown up and now it's hitting a lot of different people.

[03:45] If you want to run tests and get data for what works for the broadest amount of people possible, I would highly recommend you do these kind of tests here after the first few days. That being said, if the video is not doing well, if it's flopping, you should 100% be testing

[03:59] the title of the thumbnail, trying to get it up in views. Just realize that if you don't see a massive jump improvement, there's potential here that adjusted well with a certain segment of your audience and not with everybody on YouTube,

[04:11] if that makes sense. I have been running a lot of tests recently, especially trying to make old videos blow up. So we have old videos on some client channels that are just flat and we're trying to change a thumbnail title to make the views go boom, go kaboom and blow up and there are some things

[04:26] you should change in a thumbnail that will definitely get your videos to blow up and other things that don't matter at all. So what are the things that don't matter? Well, the things that don't matter is anything that's minor. So for example, let's say you a thumbnail in your shirt is red, okay?

[04:41] This is my crew to attempt at drawing a shirt. This shirt is red. If you run an AB test, changing it to blue, those don't make a difference. We've tested this extensively across like I think 10, 15 different videos in small variations

[04:56] like that rarely make it impact. Usually what ends up happening is that the result is just 50 to like 49 and it comes back as inconclusive. So that does not make a big difference. Anything minor doesn't make a big difference.

[05:09] This isn't going to be the same with expressions, you know, changing the level of zoom, with an important caveat I'll talk about and you know, changing random colors of stuff or the brightness or the saturation. What's way more important than all of this stuff is going to be these two things, C plus I.

[05:28] What are these two things? C stands for clarity, okay? How clear is your thumbnail and I'll tell you the second. For the clarity point, I'll show you this crazy example. So this video was uploaded to a client's channel and it's called I tested every band school

[05:42] gadget. And as you can see over here, we did a thumbnail test that blew it up, but what was it exactly? It was a clarity change. So the original thumbnail was this object over here. And if you're like me, you might think, what the hell is that?

[05:56] Oh, it's some sort of like device for cheating or whatever. And all Hafu did, the owner of the channel, all he did is he changed it from this red thing and he made it look more like a pencil.

[06:09] And like two days later, it just blew the hell up. And so that's what I mean by clarity things. When it comes to thumbnails, the most powerful thumbnails are extremely clear and they're very, very interesting.

[06:21] That's the second one right here. And so if you want to A, B test things, the highest leverage things are going to be making your thumbnail clearer, easy to understand, more logical and understandable in a glance and also changing it to be more interesting.

[06:35] The original title and thumbnail was this thumbnail right here, lava being poured into a pool. And the title was molten lava versus pool experiment. Now the video had completely died. And so I was looking around and I noticed two things.

[06:47] First off, I noticed this little molten object here was doing really well on Instagram and YouTube shorts. Like it was crushing. They would basically drop it through different materials and see how fast it would burn through things.

[06:59] And then second, I knew that watermelons just crushed in thumbnails. The biggest reason is that it's a globally known food and the bright red really catches your attention. So what I did is I tested this original thumbnail and this original title molten lava versus

[07:13] pool experiment to this thumbnail right here, a 4500 degree ball versus a watermelon. And this one won and it popped off the video as you can see here. The views in the last 90 days were average around 3 to 2,000.

[07:27] And now they climbed to a high of 18,000 views in a single day. And so all we did is we changed the thumbnail and title to something that's more interesting right now, more relevant to viewers, viewers right now are finding more interesting.

[07:40] And this old ass video has jumped up in daily views by like almost 10x. The fourth important thing I want to talk about here is to remember the end goal. A lot of the time you'll switch out thumbnails and a new one will win.

[07:53] Let's say the original one here was here and then you try it against a new thumbnail and then new one wins and you say, hooray, I'm so pumped, you're super excited. Let's go. You switched the thumbnail from this one to this one, because this one won, but views don't

[08:06] go up. And you have to remember the goal when you're switching thumbnails, the goal is to stimulate old videos and get more views or to save a dying one that you just posted and get more views. So just because a thumbnail won an AB test doesn't mean you should leave that thumbnail up because

[08:22] so many times a thumbnail will win the AB test and then views will stay the exact same. And that's not the goal, the goal is not to get a win, the goal is to get more views. So if you change a thumbnail and it doesn't get more views, then you should consider updating

[08:37] that video again. There is an important caveat here and that's point number five is that when these videos would blow up, when we would change these thumbnails and start seeing the growth, a lot of the time it happens up to seven days after we change from the old thumbnail to the new one.

[08:52] So just because a video doesn't pick up the immediate second that you change the thumbnail or that a thumbnail wins, that doesn't mean you should consider switching back or trying with new thumbnail yet again, sometimes you just need to wait a couple days and then you'll

[09:05] see it pick up over time. But if after seven days you don't see any growth, it's worth considering testing another thumbnail against this winner right here. The sixth thing I want to talk about here is that some videos should be prioritized for

[09:18] AB testing over others and if you know which videos to prioritize, they're way more likely to blow up and get crazy views. So which videos are those? Well for a little while, my entire job was just changing thumbnails on the back catalog

[09:33] of this channel right here. This is Nick D. Giovanni, he's the number one chef YouTuber and all I did was change thumbnails for old videos to try and stimulate new growth and I quickly saw a pattern. The videos that actually blew up when we changed the thumbnail weren't videos that had flocked

[09:49] but instead were videos that had already crushed, had gotten crazy views and had just died. And so if you have videos on your channel that have crazy views, the top videos on your channel but they've completely fallen off, those are the ones you want to prioritize first

[10:03] to change the thumbnails because they're great videos, they've probably hit millions of views already and so they're most likely to pop off again if you change the thumbnail. As opposed to like a video that is terrible, that barely got any views and you try to save

[10:17] it with a thumbnail change. Those ones could have flocked for many reasons, maybe the video was bad, the idea was bad, the editing was terrible, you just don't know what caused that video to flop. Versus a video that did well that got tens of millions of views already, that one's proven

[10:30] to have worked and so if you just make a slightly better thumbnail, you can see it blow up and get served to a lot of more people. Anyways, that's all I'm sharing for now, hope you guys enjoyed, peace.