Elijah Westbrook Subscribe

Home / Studies

YouTube · Data study

What Makes a YouTube Thumbnail Work? 2,038 Thumbnails From the Same Channels

33%

of viral thumbnails had a bigger face. Most didn’t.

The short answer

Viral thumbnails looked basically the same as normal ones from the same channel. Brightness was 51 vs 52. Faces appeared in 55% vs 58%. The face was bigger in only a third of the viral thumbnails. What separated the best ones was that they showed the promise instead of repeating the title.

Viral vs normal thumbnails from the same channels
Shows a face (viral)55%
Shows a face (normal)58%
Viral face is bigger33%

1,019 pairs

01

Bigger faces and louder colors weren’t the difference

If the advice you’ve heard is “make your face bigger” or “use brighter colors”, that isn’t what separated these videos. Same channel, same style, same face. The promise changed.

What to do

Spend your thumbnail time on the idea it shows, not on the saturation slider.

02

Show the promise in one picture

Two-thirds of MrBeast’s top 100 thumbnails have no text. When there is text, it’s a counter or a price that shows progress. You understand the video without reading a word.

What to do

Decide what your promise looks like as one picture. If it needs words, use a number that shows stakes.

03

Or add a second hook with a few words

Alex Hormozi does the opposite. His most common layout (38 of his top 100) is his face next to a short statement, about three words, and it never repeats the title.

What to do

If you use text, make it add something the title doesn’t say. Then test it, because creators are bad judges of which thumbnail wins.

From my channel

Alex Hormozi’s YouTube Growth Formula Explained in 269 Seconds
Alex Hormozi’s YouTube Growth Formula Explained in 269 Seconds

How I did it

The thumbnails of 1,019 viral/normal pairs from the same channels (2,038 images), measured for brightness, faces and face size; plus the top 100 videos from MrBeast and Alex Hormozi, tagged by layout.

The limits

Measured features can’t capture everything a person sees in a thumbnail. Always test your own with YouTube’s Test & Compare.

Questions

Do faces in thumbnails get more views?

Not in this data. Viral thumbnails showed a face 55% of the time and normal ones 58%, from the same channels.

Should thumbnails have text?

Either works. MrBeast mostly uses none; Hormozi uses about three words that never repeat the title. Text should add a second hook.