How do I stop people skipping past my thumbnail?

By making it look like somebody competent made it, and by making it legible at the size it is actually shown. In the one study that recorded people talking through their video choices, the two things they said most often about a thumbnail were that it made the video look interesting or that it made the video look boring and unprofessional, and both came up more than any comment about whether the video was on the right topic. So people judge the look of the picture before they judge the subject. Nobody has published what specifically makes a thumbnail read as unprofessional, so the only supported fix is the measured rules: one clear subject, legible at about 168 pixels wide, with moderate detail rather than an empty frame or a cluttered one. On emotion, the largest controlled test that exists ran on headlines rather than thumbnails and found fear lowered clicks by 0.7% and anger had no effect at all, while sadness raised them 0.7%.

Do fear and outrage make people click a thumbnail?

Not according to the largest controlled test of emotional wording that exists, though it measured headlines rather than thumbnails. Robertson and colleagues, in Nature Human Behaviour in 2023, analysed about 105,000 headline variations from 22,743 randomised experiments covering 5.7 million clicks across more than 370 million times shown. Each additional negative word raised the click-through rate by 2.3% and each additional positive word lowered it by 1.0%. The negative words that came up most often were plain words of judgement rather than dramatic ones: wrong, bad and awful, against love, pretty and beautiful on the positive side. But broken down by emotion, fear lowered clicks by 0.7%, anger had no measurable effect at all, joy lowered them by 0.9%, and sadness was the only one that helped, at 0.7%. Negative is not the same as alarming: bleak wording won while frightening and furious wording did not. Note these are changes to a click-through rate averaging 1.39%, not changes of 2.3 percentage points, and nobody has run the equivalent experiment on thumbnail images.

What is a good click-through rate on YouTube?

YouTube's own published answer is a range rather than a target: half of all channels and videos sit somewhere between 2% and 10%. YouTube's Creator Liaison Rene Ritchie has said the only honest answer is better than your last video, because the same video can show 20% when it is offered to a hundred loyal subscribers and 2% when it is offered to a million strangers. YouTube also teaches that a falling click rate is often a success signal, and gives its own worked example of a video going from 10,000 times shown at 9% to 100,000 times shown at 3.5%, which turns 900 views into 3,500 views.

Does YouTube's thumbnail A/B test pick the winner on clicks?

No, and YouTube says so plainly. Its test and compare feature never shows you a click rate at all. YouTube's help documentation states that to help your video get high quality engagement, they optimise tests for overall watch time over other metrics like click-through-rate. You can test up to three options, a test usually finishes within two weeks, and it is normal to get no winner at all. Shorts, premieres, scheduled live streams and made-for-kids videos cannot be tested.

Do faces make YouTube thumbnails perform better?

Only where the surrounding thumbnails do not already have one. An analysis of more than 300,000 popular videos across 52,000 channels found thumbnails with and without faces performed comparably overall, with faces hurting in gaming and helping in lifestyle, finance and beauty. The mechanism appears in a marketing study of social media images: adding an image lifted engagement by 119% to 213% on one platform and did nothing on another, because human images were already common there and so were not distinctive enough to help. The widely quoted figure that 69% of the thumbnails on videos that took off contain faces is a count of how common faces are, with no comparison group, and it actually argues that a face helps less rather than more.

What Makes a Good YouTube Thumbnail? Every Rule, Checked

Q: What makes a good YouTube thumbnail?

Being instantly legible at the size it is actually shown, saying something different from the thumbnails beside it, and promising only what the video delivers. Those are the parts with evidence behind them. When people choose a video, eye tracking shows the thumbnail takes about 35% of the time they spend looking against about 10% for the title, and they use it mainly to decide which videos to skip rather than which to watch. Most of the famous design rules were never tested: the three to five words rule has no traceable origin, arrows and circles have never been measured by anyone, and the well-known colour statistics turn out to be invented. The largest study ever run, across roughly 500,000 thumbnails, found that more text on a thumbnail went with fewer people watching.

Everybody treats a YouTube thumbnail as the thing that wins you the click. In a 2023 study led by Philipp Bitzenbauer, a physics education researcher, researchers put 24 people in front of an eye tracker and watched them choose from eight real YouTube results.

The thumbnail decided 26 of the choices those 24 people made. Only 5 of those 26 were somebody choosing to watch a video. The other 21 were somebody choosing to skip a video.

Illustration: a grid of nine dark video cards each holding a blank rounded figure, with one card lit up in bright blue and marked by three small coral squares

Build your thumbnail around one subject, a few large words, and a promise the video keeps

Build your thumbnail around one subject and at most a few large words, because a viewer's single look at a thumbnail lasts about a fifth of a second.
Make your thumbnail different from the thumbnails around it rather than louder than them.
Promise only what the video delivers, because YouTube judges a thumbnail by whether viewers keep watching after they click it.

What the video is about	The words on the thumbnail	How many words	What it promises
Camera review	AWFUL CAMERA	2, readable at a glance	A negative review, which the video then has to deliver
Repair guide	THIS SNAPPED	2, readable at a glance	Something broke and you fix it on camera
Price comparison	$400 VS $40	3, readable at a glance	A straight comparison of two prices
Recipe that failed	I BURNT IT	3, readable at a glance	The recipe went wrong and you show it going wrong
Long-running project	6 MONTHS LATER	3, readable at a glance	A before and after, with the after shown
Investing opinion	EVERYONE IS WRONG	3, readable at a glance	An argument against the common view
Beginner tutorial	YOU ARE DOING IT WRONG	5, near the limit	A correction to something you already do
Gaming build	HOW I BUILT MY GAMING PC FOR UNDER $800	9, unreadable at 168 pixels	Too many words to take in during one look

Illustration: a wide dark rectangle under a dense cloud of small blue dots, beside a thin flat bar under only a scattering of them

Spend four times longer on your thumbnail than on your title, because that is the attention split

The best measurement of where a viewer's attention goes on YouTube is the 24-person eye-tracking study, which used a real YouTube layout.

Part of the video card	Share of all the glances	How long one glance lasted
The thumbnail	45.6% nearly half of all glances	About a fifth of a second
The title	11.6%	About a fifth of a second
The channel name	7.1%	About a fifth of a second
View count	3.1%	About a fifth of a second
The video length	1.1%	About a fifth of a second

The thumbnail got 45.6% of all the glances people gave a YouTube video card, roughly four times the 11.6% the title got.
Viewers do not look at a thumbnail for longer than at the title, they just look at the thumbnail more often.
The caveat is that the eye-tracking study watched only 24 people, who chose from a fixed page of YouTube results rather than a live scrolling feed.

Illustration: a vertical stack of dark cards each marked with a coral cross and one marked with a blue tick, beside a large funnel narrowing downward

Design your thumbnail to avoid being skipped rather than to persuade anybody to click

In the eye-tracking study the thumbnail decided 26 choices, and 21 of those 26 were people deciding to skip a video rather than watch it.
When people in the eye-tracking study did choose a video, they mostly could not say what on the video card made them choose it.
A thumbnail that makes a claim invites the viewer to judge the claim, and judging the claim is how a video gets skipped.
A thumbnail that states a plain fact gives the viewer nothing to reject, which is all a thumbnail needs to do before the click.

What the video is about	Trying to persuade, so it gets skipped	Just stating what the video is, so it survives	Why the persuading version loses
Excel tutorial	THE BEST EXCEL TRICK EVER	PIVOT TABLES IN 3 MINUTES	Makes a claim the viewer cannot check, so it reads as bait
Camera review	YOU WON'T BELIEVE THIS CAMERA	THE R6 AFTER A YEAR	Makes a claim the viewer cannot check, so it reads as bait
Fitness progress	TRANSFORM YOUR BODY FAST	6 MONTHS LATER	Makes a claim the viewer cannot check, so it reads as bait
Cooking	THE MOST AMAZING PASTA	I BURNT IT	Makes a claim the viewer cannot check, so it reads as bait
Coding	LEARN PYTHON THE EASY WAY	WHY MY CODE BROKE	Says nothing specific, so there is nothing to recognise
Travel	THE TRIP OF A LIFETIME	3 DAYS IN TOKYO	Says nothing specific, so there is nothing to recognise
Product comparison	THIS CHANGED MY LIFE	$400 VS $40	Says nothing specific, so there is nothing to recognise

Make your thumbnail look professional first, because people called skipped videos boring and unprofessional

The 24-person eye-tracking study also recorded what people said out loud as they rejected each YouTube video.

The two things people in the eye-tracking study said most about a thumbnail were that it made the video look interesting, or that it made the video look boring and unprofessional.
People in the eye-tracking study judged the look of a thumbnail before they judged whether the video was on the right topic.
21 of the 24 people in the eye-tracking study said outright that the thumbnail decided their choice, ahead of every other part of the video card.
The caveat is that the 24 people in the eye-tracking study were trainee physics teachers choosing a video to teach with.

Use Camera to Clipboard to shoot thumbnail photos on your phone

A real photograph of a real subject is the cheapest way to stop a thumbnail looking amateur. The awkward part is getting a phone photo off the phone and into your design tool. Camera to Clipboard, one of my own apps, turns moving the photo from phone to computer into a single step.

Download Camera to Clipboard from the App Store

Download Camera to Clipboard from Google Play

A photo you take inside Camera to Clipboard is ready to paste on your computer the moment you take it.
Pasting a Camera to Clipboard photo into Photoshop drops it in as a new layer, which is where a thumbnail actually gets built.
Camera to Clipboard can send several photos at once from your phone's library to your computer, so every take of the same subject arrives together.

Illustration: two pale cards side by side, the left crowded with many stacked dark lines and a coral arrow pointing down, the right holding one short bar and a blue arrow pointing up

Use as few words as possible, because thumbnails with more text get watched less

The largest thumbnail study ever run, a 2024 study of almost 500,000 thumbnails by Loebbecke and colleagues, found that more text on a thumbnail meant fewer people watched the video or bought the item.
A separate study of 16,215 YouTube videos also found that more thumbnail text went with less watching.
Punctuation, capitals and emoji are the only kind of thumbnail text decoration with a measured benefit.

Illustration: a dark rounded bar on the left and a row of five blue squares on the right fading gradually to nothing

Keep it to about three words, because that is all a glance has time to read

Adults read about four words a second, and a single glance at a thumbnail lasts about a fifth of a second.
At four words a second, one glance at a thumbnail has time to read under one word, and three glances about three words.
Put about three words on a thumbnail, and never more than five.
Judge your thumbnail text by whether it is legible when the thumbnail is 168 pixels wide, not by counting words.

Use Repeat Recorder to rehearse the first line of your video

A thumbnail only has to get somebody to press play, and YouTube then judges the video on whether the viewer stayed. Repeat Recorder, another of my own apps, plays each recorded take of your intro straight back without you touching the phone.

Download Repeat Recorder from the App Store

Download Repeat Recorder from Google Play

Repeat Recorder replays each take and starts recording again on its own, so you can rehearse one line of your video twenty times.
Hearing a take played back is what catches a video intro that takes forty seconds to deliver what the thumbnail promised.
Repeat Recorder saves the takes worth keeping into folders you name yourself.

Illustration: a row of six pale cards each holding an identical blank dark head shape, and a seventh card holding a bright landscape instead

Add a face only when the thumbnails beside yours have none, because faces do not win on their own

"Put a face on it" is the most confident advice about YouTube thumbnails.

Across more than 300,000 popular videos, thumbnails with and without faces performed about the same.
Adding an image to a post lifted engagement 119% on Twitter but did nothing at all on Instagram, where every rival post already had an image.
Faces on a thumbnail reliably grab a viewer's eye, but grabbing the eye is not the same as getting the click.
The effect of an exaggerated shocked face on a thumbnail has never been measured.

Illustration: two groups of feed cards, the left group blurred and soft with one sharp card among them, the right group sharp with one blurred card lost inside it

Be specific only when the thumbnails beside yours are vague, and vague when they are specific

Holding something back in a thumbnail is the one piece of popular thumbnail advice with a good study behind it.

Making a headline more specific raised its clicks 5.5% when the headlines around it were vague, but cut its clicks 9.9% when they were specific.
Only 8.7% of the headlines tested gained clicks from becoming more concrete, while 50.9% lost clicks.
The caveat is that the specificity study measured headlines, not thumbnails.

Use negative words instead of positive ones, because each one is worth 2.3% more clicks

The effect of negative words was measured across 105,000 headline versions in 22,743 randomised tests.

Each extra negative word in a headline raised its click rate by 2.3%.
Each extra positive word in a headline lowered its click rate by 1.0%.
The three commonest negative words in the tested headlines were "wrong", "bad" and "awful".
The three commonest positive words in the tested headlines were "love", "pretty" and "beautiful".

What the video is about	The positive version that loses	The negative version that wins	What the swap is worth
Dividend news	GREAT NEWS	BAD NEWS	Drops a positive word and adds a negative one, so both effects apply
Index funds	WHY I LOVE INDEX FUNDS	EVERYONE IS WRONG	Drops the commonest positive word for the commonest negative one
Portfolio review	MY BEST FUND	MY WORST FUND	Drops a positive word and adds a negative one, so both effects apply
Camera review	A BEAUTIFUL CAMERA	AN AWFUL CAMERA	Drops the commonest positive word for the commonest negative one
Beginner tutorial	THE EASY WAY	YOU ARE DOING IT WRONG	Adds a negative word to text that was only neutral
Desk setup	MY SETUP TOUR	MY SETUP IS A MESS	Adds a negative word to text that was only neutral

Illustration: two blank sign boards side by side, one carrying a small steady mark and the other a jagged burst with a downward arrow

Do not frighten anybody, because fear lowered clicks by 0.7%

Negative wording and alarming wording are not the same thing.

Frightening wording in a headline lowered clicks by 0.7%, so a frightening thumbnail is worse than a neutral one.
Angry wording in a headline did nothing measurable to clicks, so outrage in a thumbnail is wasted effort rather than a mistake.
A negative word passes judgement on something while a frightening word predicts a disaster, which is why "the worst fund" beats "the coming collapse".

What the video is about	The alarming version that loses	The judging version that wins
Bond market	THE COMING BOND COLLAPSE	THE BOND MISTAKE
Market outlook	MARKETS WILL CRASH	MARKETS ARE OVERPRICED
Health advice	THIS WILL KILL YOU	THIS IS A BAD HABIT
Online security	YOU ARE BEING HACKED	YOUR PASSWORD IS WEAK

Illustration: three pale rounded cards in a row, the left completely empty, the middle holding one clear blue shape, the right crammed with dozens of small scattered shapes

Put one thing in the frame and then stop, because clutter loses to a clear frame

One thing in the frame means one subject shot close enough to read at a glance, plus at most a few words. One thing never means a face and an arrow and a circle and a caption together.

An empty thumbnail and a cluttered thumbnail both do worse than one with a moderate amount in the frame.
The finding that empty and cluttered thumbnails both do worse came from 22,958 thumbnails backed by two randomised experiments, the strongest evidence behind any thumbnail design rule.
In a study of 249 advertisements, a densely busy design cut viewers' attention to the brand and made viewers like the advert less.
YouTube's own thumbnail guidance also tells creators to keep the frame uncluttered.

What the video is about	The one thing to put in the frame	The words on it	What tips it into clutter
Camera review	The camera held up close, filling half the frame	AWFUL CAMERA	Adding your face, an arrow and a red circle as well
Recipe	The finished dish shot close up	I BURNT IT	Adding the raw ingredients and a logo beside it
Price comparison	The two prices side by side	$400 VS $40	Adding a face, arrows and a border
Repair guide	The broken part held in your hand	THIS SNAPPED	Adding the whole tool kit and three captions
Before and after	The two states split down the middle	6 MONTHS	Adding a face, a date and an arrow
Reaction video	Your face filling a third of the frame	I WAS WRONG	Adding the clip you reacted to plus a caption

Illustration: a solid navy arrow, a navy circle outline and a navy rectangular border in a row, each with an empty dashed box beneath it

Skip the arrows, circles and colour rules, because nobody measured them and they add clutter

Nobody has ever measured what an arrow, a red circle or a border does to clicks.
Every extra element in a thumbnail has been measured to make it do worse, so an arrow is a loss even though arrows themselves were never tested.
The two most quoted studies about thumbnail colour cannot be found at all.
The figures quoted online for how much a red thumbnail helps disagree, at 23%, 30% and 32%, and none of them says what it is a percentage of.
The popular claim that a consistent thumbnail style lifts clicks by up to 38% was traced back, and it has no source.
The only things genuinely measured about thumbnail colour are contrast and legibility, which are worth a few percent at most.

Illustration: a grid of small cells each holding either a solid filled circle or a hollow outlined circle, in navy and blue

Act on these six thumbnail rules and avoid these five, with the evidence for each

The six thumbnail rules to act on are all free, and every one has some measured evidence behind it.
The five thumbnail habits to avoid either cost you clicks or waste your time for nothing.

The rule	Do it or avoid it	What was actually found	How strong the evidence is
Make it readable at a glance	Do this first	A glance at a thumbnail lasts a fifth of a second	Eye-tracked on a real YouTube layout
Be different from the thumbnails beside you	Do	Clicks up 5.5% against vague rivals, down 9.9% against specific rivals	8,977 pre-registered tests
Use fewer words, about three	Do	More text went with less watching	500,000 thumbnails, peer reviewed
Put one thing in the frame and stop	Do	Empty and cluttered thumbnails both lose to a moderately filled one	22,958 thumbnails plus two randomised tests
Use negative words, not positive ones	Do	Clicks up 2.3% per negative word, down 1.0% per positive word	105,000 headline tests, never on thumbnails
Add a face only if the thumbnails beside yours have none	Do	No overall difference, so a face only helps where faces are rare	300,000 videos, and the result depends on the feed
Fear or alarm	Avoid, it costs you	Cut clicks by 0.7%	105,000 headline tests, never on thumbnails
The rule of thirds	Avoid, it failed	Good thumbnails put the subject in the middle instead	Tested once, and the result went the other way
Arrows, circles and borders	Avoid, adds clutter	Never measured, and clutter is measured to cost you	Nothing exists in either direction
Chasing a particular colour	Avoid, wastes time	Claims for red thumbnails disagree, at 23%, 30% or 32% of an unstated base	The two colour studies cited cannot be found
The exaggerated shocked face	Avoid, wastes time	Only 1 in 20 thumbnails on videos that took off used a shocked face	No measurement found

YouTube officially recommends the rule of thirds for thumbnails, but the only study that tested the rule of thirds found good thumbnails put the subject in the middle instead.
Read the rule-of-thirds study carefully before quoting it, because it never measured clicks.

Illustration: the same rounded rectangle shown twice, once dark on a white background and once bright blue on a near-black panel, with an empty dashed outline between them

Test your thumbnail against dark mode yourself, because half your viewers see it on near-black

Roughly half your viewers see your thumbnail against a near-black dark mode background rather than white, but nobody has ever tested how thumbnails perform in dark mode.
Every statistic in circulation about thumbnails in dark mode is invented.
You can test your thumbnail in dark mode for your own channel in about two weeks with YouTube's own thumbnail testing tool.
Bear in mind that YouTube's thumbnail testing tool judges a thumbnail on watch time, not clicks.

Illustration: a split scene, on the left a small blank figure pressed against a huge solid navy wall, on the right an identical figure standing in a row of similar figures with a coral dot above it

Stand out from the thumbnails next to you, which is the opposite of the advice for banner ads

Standing out is worth nothing on its own, because in a classic 1980 visual search study by the psychologists Anne Treisman and Garry Gelade a distinctive item alone on a screen was no faster to find.
In a 2024 eye-tracking study of 100 people by Aline Simonetti and Enrique Bigne, marketing researchers at the University of Valencia, banner ads took 5.5% of people's viewing time when they were reading a page, but 11.9% when they were choosing a link to click.
Make your thumbnail legible and relevant faster than the thumbnails next to it, rather than louder than them.

Use Camera to Clipboard to compare thumbnail sketches on paper

Deciding which thumbnail layout is least like the thumbnails next to it is quicker on paper than in an editor. Camera to Clipboard, one of my own apps, photographs a sheet of paper sketches and puts the photo straight onto your computer.