Last month a client asked me for a thumbnail for a Short. I said Shorts do not really have thumbnails. He sent me a screenshot of his own channel page where every Short had one, hand picked, sitting in a neat grid. He was right and I was working off a rule that quietly stopped being true.
That is the honest summary of where we are. Shorts thumbnails went from "a frame YouTube picks for you" to "a real design surface with real rules", and most of the advice floating around is still describing the old version. So here is what actually changed, what it means for how we design, and where I think people are wasting money.
What actually changed
For years the Shorts thumbnail was a non decision. YouTube grabbed a frame, you got a cover image for the channel grid, and that was the end of it. The feed itself never showed a thumbnail, because the feed shows the video playing. So the logic went: the first frame is your thumbnail, design the first frame, forget the rest.
Then Shorts got a proper cover image picker. Then covers started showing up in more places: the channel grid, search results, the Shorts shelf on the home page, subscriptions, and the recommended rail after a Short ends. On the TV app the Shorts surface behaves differently again, which I wrote about in more detail in how Shorts and long form behave on a living room screen.
None of that makes a Shorts cover as important as a long form thumbnail. I want to be clear about that up front, because the loudest take right now is that Shorts thumbnails are the new frontier and you should be paying for them the way you pay for long form. That is not what I see in the data.
What I see is narrower and more useful: the Shorts cover barely affects the feed, and heavily affects everything downstream of the feed. Two different jobs, two different design briefs.
The feed does not care about your cover
In the vertical feed nobody sees a cover image. They see frame one, playing, with audio. So the thing that decides whether someone swipes is the first half second of motion, the first three words spoken, and whether the framing reads at arm's length on a phone.
We tested this on our own channels because I did not want to burn a client's inventory on a hunch. We uploaded the same Short twice on two of our channels, identical video, different covers, one plain and one heavily designed with text and a face crop. Feed retention and swipe away rate were basically identical. Channel page click through was not. The designed one pulled roughly twice the clicks from the grid on a channel in the 40k range.
That is one test on our own small channels. I am not going to dress it up as a law. But the direction matched what I expected from how the surfaces work, and it has held on every channel we have tried it on since.
The Shorts cover is not a feed asset. It is a back catalogue asset. It sells the video to people who already found your channel.
Where the cover actually earns money
Once you accept that split, the decision about where to spend design effort gets easy.
A Shorts cover matters when the Short is:
- Sitting on your channel page, where a new visitor decides whether you make things worth watching
- Showing up in search, where it competes against long form thumbnails in the same list
- In a playlist you actually promote
- Old and still pulling views, which is where Shorts quietly make most of their money
It does not matter much when the Short is:
- One of forty posted this month with no shelf life
- A clip of a longer video where the point is the clip, not the packaging
- Trend bait that will be dead in nine days
Most channels post far more of the second kind than the first. So the correct answer for most channels is: design covers for the ten Shorts you want people to find in six months, and let YouTube pick the frame for the rest. That is an unpopular answer for a studio that sells thumbnails, and I still think it is the right one.
Vertical breaks everything you know about composition
Here is where the design side gets interesting, and where most people get it wrong.
A long form thumbnail is 16:9 and gets rendered small. You have room to put a face on one side and a subject on the other, and the eye reads left to right. That composition does not survive a 9:16 crop. The face and the subject stack vertically instead, and the grid on a channel page crops the top and bottom on some surfaces anyway.
So the rules invert:
Text goes shorter, not smaller. In 16:9 we can usually get four words to read at small size. In vertical, three is the ceiling and two is better, because vertical text blocks eat the frame that the subject needs.
The face gets bigger and moves up. Center of the frame is where the play affordance and the duration badge tend to land depending on surface. Pushing the subject into the upper third keeps it clear.
One idea, no second layer. Long form thumbnails can carry a primary hook and a secondary detail. Vertical cannot. Anything you add as a second layer is noise at 200 pixels wide.
Contrast beats color theory. In a grid of Shorts the covers sit shoulder to shoulder in a dense block. Saturation does not separate you, value does. A dark cover in a wall of bright ones wins more often than a brighter bright one.
This is why the copy paste approach fails. Taking the long form thumbnail and cropping it to vertical produces a technically valid image that reads as mush. Every cover we make is drawn for the aspect ratio it will live in, which is the same reason we never work from templates on the long form side either. If you want to see how that plays out across formats, most of our work shows both.
The MrBeast problem, vertical edition
The dominant visual language of long form thumbnails, big face, big reaction, blown out saturation, heavy outline, got copied into Shorts wholesale. It works worse there, and it works worse for a specific reason: it was designed to win a fight for attention in a mixed grid where your competitor is a different creator. On a channel page you are not competing with other creators. You are competing with yourself, forty times over.
Forty maximum intensity covers in a grid cancel each other out. Nothing stands out because everything is shouting. I have argued before that the big face style is a tool and not a personality, and vertical is where that becomes obvious fast.
The channels whose Shorts grids actually look good tend to do the opposite. They pick a repeatable frame, a consistent color, a consistent text position, and let variation happen inside that system. The grid reads as a body of work instead of a pile of screaming.
How to test this without lying to yourself
The measurement trap on Shorts is brutal and I want to name it clearly.
Shorts get enormous impression counts from the feed, and feed impressions swamp everything else. If you swap a cover and look at overall click through rate, you will see almost no movement, because 95 percent of your impressions came from a surface where the cover was never displayed. You will conclude that covers do not matter. You will be measuring the wrong thing.
The number that moves is channel page and search click through, and you have to go looking for it by traffic source. This is the same class of mistake as reading raw impressions on long form, which I went into properly in why high impressions can mean nothing at all.
The second trap is sample size. A Short that got two million feed views might have had four thousand channel page impressions. Four thousand impressions is not a test. It is a rumor. Most of the "Shorts thumbnail tests" I get sent are decided on differences that would evaporate with another week of data, which is a whole subject on its own and one I covered in the point where a test is actually finished.
If you want a clean read, run the variant, then wait for the Short to age out of its feed push and look at what it does over the following month. The back catalogue period is where the cover does its work anyway, so that is the period worth measuring.
What this means for what you should buy
Our pack is $250 for 8 custom thumbnails, one per video, with unlimited revisions and a dedicated designer, and each one takes 48 to 72 hours. At about $31 a thumbnail, spending one of those eight slots on a Short is a real decision, not a rounding error. The full breakdown of that math lives on what thumbnails actually cost.
My honest recommendation:
- If you post long form and Shorts, spend the pack on long form first. The math is not close.
- If a Short is a pillar piece, an intro to your channel, a promo for something you sell, it earns a slot.
- If you want a variant to compare, the A/B variant add on is $15 and that is a cheaper experiment than a second full thumbnail.
And if your Shorts come out of a podcast or a long form episode, the packaging conversation starts one step earlier, at how the clip was cut. We handle that side through Pixel Samy's podcast editing, where clips are $25 each, and a well chosen clip with a plain cover beats a badly chosen clip with a beautiful one every time.
The part I am still unsure about
I do not know how long the current surface mix lasts. YouTube has moved Shorts around the interface repeatedly, and every move changes how much a cover is worth. If covers start showing in the feed tomorrow, everything in this post about the feed becomes wrong and the budget answer flips.
So I am holding this loosely. What I am confident about is the underlying principle, which does not depend on the layout: design for the surface where the image is actually seen, and measure only on that surface. Everything else is decoration and guesswork.
If you are trying to work out whether Shorts covers deserve a slot in your next batch, send me the channel and I will tell you honestly. Sometimes the answer is no.