YouTube End Screens and Cards: How to Engineer the Second Click Now That Viewers Can Hide Them

Almost every packaging conversation ends at the first click. Thumbnail, title, impressions, click-through rate — the whole vocabulary describes a stranger deciding to start your video. Then the video ends and a second decision happens that nobody planned for: the viewer watches something else of yours, watches something else of somebody else's, or leaves. That decision is worth more than the first, because the person making it has already proved they like your work.

YouTube gives every creator two free tools to influence it. End screens are the interactive panel over the last few seconds of a video; cards are the small clickable elements you can drop at any timestamp. They cost nothing, take a couple of minutes in Studio, and are the only in-player promotional surfaces the platform hands out for free.

They are also getting weaker, and it is worth being honest about why. In September 2025 YouTube added a Hide button to end screens after viewers asked for fewer distractions, and killed the desktop hover-to-subscribe button outright. What follows is what these features actually do in 2026, what the platform's own documentation and analytics will and will not tell you about them, and how to build the last twenty seconds of a video so the next click is a decision rather than a default.

What the two features actually are

The specifications matter here more than usual, because most bad end screens are the result of someone discovering the constraints after the video was already edited. YouTube's end screens documentation and its cards documentation set the boundaries.

	End screens	Cards
Where it appears	The last 5 to 20 seconds of the video	Any timestamp you choose
How many	Up to 4 elements on a standard 16:9 video; other aspect ratios may allow fewer	Up to 5 cards per video
Minimum video length	25 seconds	None stated
Element types	Video, playlist, subscribe, channel, link	Video, playlist, channel, link
Extra requirements	The link element needs YouTube Partner Programme membership; the channel element requires a custom message	Message and teaser text are optional, except on channel cards where they are required
Not available on	Videos set as made for kids, even if an end screen was added earlier	Videos set as made for kids
How easily it is ignored	A Hide control has cleared it in one tap since September 2025	The teaser shows briefly, then collapses into an icon the viewer has to tap

Two of those rows are the ones creators trip over. The 25-second minimum quietly excludes a lot of short-form-adjacent content, and the 5-to-20-second window means the end screen sits on top of footage you shot, not on a black card YouTube appends afterwards. If your video ends on a hard cut, the panel still appears — over whatever happened to be on screen while you said goodbye.

The change that matters: viewers can now hide your end screen

In September 2025 YouTube announced two changes to the watch page, both framed explicitly as responses to viewers asking for fewer distractions. The first: a Hide button now appears in the corner of the player when an end screen pops up. Tap it and the elements disappear for that video, leaving the footage clean. A Show control brings them back. The dismissal is per-video rather than a permanent setting, so it is not a global opt-out — but it means any given viewer can remove your four elements from their screen in one tap, at the exact moment they were meant to work.

The second change is the more instructive one. YouTube also removed the desktop hover-to-subscribe button — the Subscribe prompt that appeared when a mouse passed over the channel watermark — and gave a reason with a number attached: internal data from June 2025 showed that fewer than 0.05% of all channel subscriptions came from it. The subscribe control remains directly below the player, where it always was.

None of this makes end screens worthless. It makes them a closing argument rather than a storefront. The viewer who reaches the last twenty seconds of a fifteen-minute video is the most qualified audience you will ever have. The mistake is treating the panel as if it were doing the persuading. It is not. The video is.

An end screen is twenty seconds of your video, not a slot after it

This is the most common structural failure, and it is invisible in Studio, which shows you the editor over a paused frame. In the real player three things happen at once in those final seconds: your conclusion is still running, the elements slide in over the top, and attention starts migrating to the sidebar. If your closing 20 seconds are doing real work — a summary, a punchline, the payoff you promised in the title — the elements compete with your own content. If they are dead air, the viewer left before the elements rendered.

The fix is boring and effective: build a deliberate end screen bed. Finish the argument, then hold a designed 15-to-20-second outro that has almost nothing in it — a static or slowly moving background, no dense on-screen text, no new information, and open space in the regions where the elements will land. Say one sentence over it naming the specific thing you want the viewer to do next, and stop talking. The panel then reads as part of the video instead of a layer over it.

Two constraints on that bed. Keep it under 20 seconds, because anything longer is a retention cliff you built by hand; the flat tail it produces is easy to spot in your audience retention graph, where the drop usually starts the instant the talking stops. And keep your own graphics clear of where the elements sit, since YouTube draws them wherever you placed them and will happily cover your logo or your final call to action with a thumbnail.

The four elements, and why most channels pick wrong

You get up to four elements on a 16:9 video, and the temptation is to use all four. Resist it. A grid of four small thumbnails in the last seconds of a video is a menu, and menus are work. The channels that get end screens to do anything usually run one or two elements with an obvious hierarchy.

The element types, and what each one is genuinely for:

Element	What it does	When it is the right pick
Specific video	Features one video you choose	Almost always. It is the only option where you control the argument for the next click.
Most recent upload	Always shows your newest video	Back catalogue you will never revisit, on a channel where every video suits every viewer. Rare.
Best for viewer	YouTube picks a video from your channel to suit that viewer	As a second element, or on old videos you do not want to maintain. It is the only element that uses YouTube's personalisation on your behalf.
Playlist	Sends the viewer into a public playlist	Series content, tutorials, anything with an order. The highest-leverage element on most channels.
Subscribe	Channel avatar with a subscribe control	Cheap to include and easy to over-rate. It converts people already convinced, not people deciding.
Channel	Promotes another channel with a required custom message	Collaborations and credits. The message field is the whole value; "check them out" wastes it.
Link	Sends the viewer to an approved external site, for Partner Programme members	When the video's purpose is off-platform. Note the trade: it ends the YouTube session.

The pairing that works on most channels is one specific video plus one playlist. The video is your argued recommendation — the thing you say out loud in the outro, chosen because it follows from what the viewer just watched. The playlist is the escape hatch for the viewer who wants more but not that one in particular. A subscribe element is fine alongside them, as a convenience for people who already decided.

The case for "best for viewer" being your second element

Creators tend to distrust the "best for viewer" option because it hands the choice to YouTube. That instinct is worth reconsidering. YouTube's own description of its recommendation system is of a model learning from an enormous volume of behavioural signals, including whether people were satisfied rather than merely whether they watched. It knows this specific viewer's history; you know only that they finished one video. On a back catalogue of a hundred videos, letting the system pick from your own library is often a better bet than your guess from two years ago — and it never goes stale.

Cards: five allowed, one or two useful

Cards are the mid-video version of the same idea, and quieter than most creators expect. A teaser appears at the timestamp you set, shows your message for a few seconds, then collapses into a small icon the viewer has to tap to browse the video's cards. You get up to five per video, in four types — video, playlist, channel and link — and channel cards require both a message and teaser text.

The honest framing: a card is an interruption, and interruptions cost retention. That makes them right for a narrow set of jobs and wrong for general promotion.

• The reference. You mention a previous video, and a card appears at that exact sentence. This is the only card use that is unambiguously good, because the viewer's question and the answer arrive at the same second.
• The correction. You got something wrong and a card points at the fix. Cheaper than re-uploading, and it can be added to a video that already has views.
• The prerequisite. Early in a tutorial, a card to the setup video.
• The off-ramp you actually want. A link card on a video whose purpose is a conversion, not a view.

What cards are not good for is scattering five promotions across a video in the hope that one lands. Every card is a chance for a viewer who was watching to start browsing instead, and a click that leaves early trades a lot of watch time for one click you cannot measure well. Place cards where a viewer would naturally want them, not at the arbitrary midpoint some checklist recommended.

The metrics that exist, and the one that does not

Here is the part almost nobody knows, and it changes how you evaluate all of this. End screen thumbnails do not generate impressions. YouTube's impressions and CTR FAQ states plainly that not all impressions are counted — those on external sites or end screens are excluded — and that one only registers when a thumbnail is shown for more than a second with at least half of it visible.

So the second click is entirely absent from your click-through rate. A video that sends thousands of viewers into a playlist looks identical, in the CTR column, to one that sends nobody — judge packaging by CTR alone and you are grading half the job. It is the same measurement gap that makes the traffic sources report easy to misread, which we worked through in what browse, suggested and search are actually telling you.

What you do get sits under Engagement in YouTube Analytics, per YouTube's engagement documentation:

• End screen element click rate, available once you have clicked into an individual video. This is the closest thing to a CTR for the second click.
• Top end screen element types, across the whole channel — the report that tells you whether your audience takes videos, playlists or the subscribe control.
• Top videos by end screen, which shows which end screens are earning the clicks.
• Top cards, showing which cards viewers clicked most across your videos.

A caveat on where to find these. YouTube has been staging a redesigned Studio through 2026 in which the Analytics tab is renamed Insights and the layout regroups metrics, so two creators comparing screenshots this month may be looking at different interfaces. The reports themselves have not changed underneath the rename.

Autoplay is what you are actually competing with

The end screen is not competing with nothing. It is competing with the up-next slot and with autoplay, which plays a related video the moment yours ends and shows the viewer what is coming. YouTube's autoplay documentation notes a few things worth knowing: the setting is per device, it is turned off by default for signed-in users aged 13 to 17, and a viewer can hit Cancel at the end of a video to stop the next one.

The thing being autoplayed is chosen well. YouTube's own description of the recommendation system says it uses the video you are currently watching as the main signal when suggesting what to play next, while the homepage leans mainly on watch history. That is a personalised recommendation, and it is often a better fit than whatever you hard-coded three months ago.

Which tells you exactly when your element wins. It wins when it offers something the recommendation system structurally cannot: continuity. Part two of the thing they just watched. The build in the series. The specific answer to the question your video just raised. It loses when it offers "another video by me", because on that comparison the algorithm is better informed than you are.

The playlist is the highest-leverage element you are not using

A video element buys you one more view. A playlist element buys you a queue. Once a viewer is inside a playlist the next video is not a decision at all — it plays, and the sidebar competition is replaced by an ordered list you designed.

Playlists have also become a real packaging surface rather than a filing system. YouTube now supports custom playlist thumbnails — an image from your device, a photo you take, or one generated with AI — so a playlist no longer has to inherit the thumbnail of whichever video sits first. Titles run to 150 characters and descriptions to 5,000, which is more room than most creators use for something that appears in search results and on your channel page.

Three rules make playlists work as an end screen destination:

1. Name the playlist like a product, not a folder. "Beginner Guitar in 30 Days" earns a click. "Guitar videos" does not. The title is doing the same job a video title does.
2. Put the strongest video first. The playlist is judged on its opener, and the opener is what plays when someone clicks the element.
3. Give it a thumbnail that reads as a set. A playlist thumbnail sharing the channel's colour and type system, with a title treatment that signals a series rather than an episode, tells the viewer they are entering a collection.

Designing for a thumbnail that appears at end screen size

The thumbnail on an end screen element is doing a different job from the thumbnail in a search result, and it fails in different ways. It renders small, over moving footage, in a panel the viewer did not ask for, next to up to three other elements. Everything that makes a thumbnail work at 168 by 94 pixels in a feed matters more here, not less — the rules in our size and safe-zone guide apply directly.

A few things are specific to this surface. Contrast has to work against your own outro rather than a white or dark interface, so a pale thumbnail over a pale outro loses its edges entirely. Text has to be shorter than usual. And the second thumbnail should look clearly different from the one they clicked twenty minutes ago — if your thumbnails are heavily templated, the next video can read as the same video, and the viewer skips it as a repeat.

The practical test is to look at the panel as a viewer sees it, not as the editor shows it. Export a frame from your outro, drop the candidate thumbnails on top at the size they will render, and check that each is legible and distinct. Our thumbnail preview tool does the feed version of that check for free, and the same instinct — judge it at the size it will be seen — is what stops end screen thumbnails from disappearing into the footage.

Where none of this exists

Much of the confusion around these features is that they are simply absent in several common situations, and Studio does not always explain why.

• Videos set as made for kids. Cards and end screens are both unavailable, and if you added an end screen before setting the audience, viewers will not see it. The channel watermark goes too, along with a long list of other features, to comply with children's privacy law.
• Videos under 25 seconds. No end screen at all — there is nowhere to put it.
• Non-16:9 aspect ratios. The four-element allowance is specified for standard 16:9; other ratios may permit fewer.
• Shorts. The end screen and card tooling is built around the long-form watch page, and the swipe feed has no equivalent surface. Continuity on Shorts is a packaging and pinned-comment problem instead.

The channel watermark survived the September 2025 change even though its hover behaviour did not. It is still configured under branding in Studio — a square image of at least 150 by 150 pixels, under 1 MB, showing for the last 15 seconds, from a custom start time, or throughout. Without the hover-to-subscribe control it is a branding mark rather than a conversion tool, which is roughly what it always was in practice.

A twelve-minute routine per upload

None of this deserves an hour. It deserves a repeatable pass at the end of the edit.

1. In the edit, cut a 15-second end bed. Low information, open space where the elements will land, one spoken sentence naming the next video by what it gives the viewer.
2. Pick the next video before you pick the elements. Ask what question this video leaves the viewer holding, and answer it with a specific video from your library.
3. Place two elements. The specific video, and a playlist that matches the topic. Add "best for viewer" as a third only if you have a deep catalogue.
4. Add one card, or none. Only at a moment where you reference something the viewer might want right then.
5. Check the thumbnails at render size against your outro frame. Legible, distinct from each other, distinct from the thumbnail they already clicked.
6. Two weeks later, open Engagement. Read end screen element click rate on the video, and top end screen element types across the channel. Change one variable at a time.

Do that for ten uploads and you will know something specific about your own audience that no benchmark can tell you: whether they take playlists or single videos, whether the subscribe element earns its slot, and whether the outro bed length you chose is costing you retention. That is a better foundation than any published average.

Frequently asked questions

How long can a YouTube end screen be?

End screens can be placed in the last 5 to 20 seconds of a video, and the video must be at least 25 seconds long to have one. Twenty seconds is a maximum, not a target — the right length is however long your closing sentence takes plus a beat.

Can viewers really remove my end screen?

Yes. Since September 2025 a Hide control appears when the end screen pops up, and tapping it clears the elements for that video. It is not a permanent setting — a Show control restores them, and the next video starts with the end screen enabled again.

Do end screen clicks show up in my click-through rate?

No. YouTube's impressions FAQ excludes end screen impressions from the counted total, so clicks from end screens never enter your CTR. Use the end screen element click rate report under Engagement instead, and read your CTR as a measure of first clicks only.

How many cards should I put on a video?

YouTube permits five. One is usually correct, placed where you reference something specific. Cards interrupt, and an interruption you did not need costs watch time to buy a click you cannot measure well.

Why can't I add cards or end screens to my video?

The most common reason is that the video is set as made for kids, which disables both features along with the watermark, comments, notifications and several others. The second most common is that the video is shorter than 25 seconds, which rules out an end screen entirely.

Is the subscribe element worth one of my four slots?

Sometimes, but hold it loosely. YouTube's stated reason for retiring the hover-to-subscribe button was that it drove fewer than 0.05% of subscriptions — a useful reminder of how weakly ambient subscribe prompts convert. The video convinces people; the button collects the convinced.

What the last twenty seconds are for

End screens and cards are amplifiers, not engines. A strong closing argument with one well-chosen element beats a four-element grid over an outro nobody is watching, and no amount of overlay configuration rescues a video that ended without giving the viewer a reason to want more. They are worth twelve minutes an upload because they are cheap, not because they are powerful.

What is powerful is the sequencing decision underneath them: knowing which video should follow this one, and being able to say so in a sentence. That is a packaging skill, not a Studio skill, and it is the same one that decides whether a playlist earns clicks or sits unopened — roughly the argument of the complete thumbnail playbook.

If the constraint you keep hitting is production rather than judgement — you know which video should come next, but making a second thumbnail that reads clearly at element size is another hour you do not have — that is the specific problem Thumblore exists to remove. Generate the thumbnail, check it small, place it, and get back to the part of this that actually requires you.