What is a good CTR on YouTube, and why YouTube publishes no benchmark for it anywhere in its own help pages

YouTube does not publish a target click through rate, and the help article that defines the metric gives no range, no average and no figure of any kind. Read on 31 October 2026, it defines impressions click through rate only as "How often viewers watched a video after seeing a thumbnail", and it gives no formula. So any specific percentage you have been told is a good YouTube CTR came from somebody's sample and not from YouTube. What YouTube does publish, in detail, is the definition of the denominator, and that is far more useful: it lists exactly where an impression is counted, exactly where it is not, and the two physical conditions a thumbnail has to meet. Once you know that, you can build the only benchmark that applies to you, which is your own channel's distribution by traffic source.

What counts as an impression, in YouTube's own words

Two conditions have to be met, both quoted from YouTube's help documentation as read on 31 October 2026. "Impressions are counted if the thumbnail is shown for more than 1 second" and "at least 50% of the thumbnail is visible on the screen." A thumbnail that scrolls past in half a second is not an impression, and neither is one half hidden behind the edge of the viewport.

The surfaces where impressions are counted, as listed:

• "YouTube on computers, TVs, game consoles, Android, iPhone, and iPad"
• "YouTube Search"
• "YouTube homepage (includes auto-play)"
• "YouTube feeds (subscriptions, trending, history, Watch Later)"
• The "Up Next" recommendations beside the player, including autoplay
• "Playlists"

The nine places that do not count, which is where the misreadings come from

This is the list that reframes the whole metric. As published, impressions are not counted on:

• "External websites and apps (for example, links and embeds outside of the YouTube website)"
• "YouTube mobile website"
• "YouTube Kids app"
• "YouTube Music app"
• "Content within the video player (for example, in cards or end screens)"
• "Email or notifications"
• "Videos that play in the background tab (no visible impression)"
• Thumbnails less than half visible, or visible for under one second
• "TrueView video discovery ads"

Three consequences follow immediately, and each one explains a figure people find alarming.

An embedded video has views with no impressions. If a large share of your watch time comes from your own website or a partner's, those views arrive with no thumbnail impression attached, so they are not in the denominator at all. A channel whose traffic is mostly embeds will show a CTR that describes only its YouTube surface traffic.

Notifications do not dilute it. A video pushed to subscribers by notification gets clicks that do not carry a counted impression, which is a relief rather than a problem: your CTR is not being punished for people who found you by alert.

End screen and card clicks are invisible to this metric. Which means the work you do inside the player to move viewers to the next video is not measured here at all, and has to be judged on its own report.

Why one number cannot be a benchmark

Click through rate on YouTube is an average across surfaces that behave nothing like each other, and the mix is different for every channel. A search impression comes from somebody who typed a specific intent, so it converts at a far higher rate than a homepage impression shown to a passive browser. A suggested video impression beside a related video sits somewhere between. So two channels with identical thumbnails and identical content will report different CTRs if one is discovered mostly through search and the other mostly through the homepage, and neither number tells you anything about thumbnail quality.

That is why the comparison that matters is internal. The same logic applies to click through rate on a search results page, where borrowed average curves mislead for exactly the same reason, and the method for building your own is in build your own curve instead of borrowing one.

Building your own benchmark in half an hour

1. Open the traffic source breakdown in YouTube Studio and read impressions and click through rate separately for YouTube search, browse features, suggested videos and playlists. You now have four benchmarks instead of one, and they are yours.
2. Write down the median for each source across your last twenty videos. The median rather than the mean, because one breakout video distorts an average badly at low video counts.
3. Note the impression volume beside each rate. A high rate on four hundred impressions is noise. The rate is only interesting where the volume is large enough that it could not have come out differently by chance.
4. Judge every new video against its own source mix, not against the channel average. A video that goes out to a browse heavy audience will look worse and may be performing better.
5. Re read the numbers after a week, not after a day. YouTube's own documentation notes that the key metrics card updates more often than the impressions report, so the two can disagree, and the detailed report is the one to trust for a considered reading.

The levers that actually move it, in order of effect

• The title and thumbnail as a pair. They are read together and they should not say the same thing twice. The thumbnail carries the situation and the title carries the specific, or the other way round, but not both the same.
• Legibility at small size. Most impressions are small. Three or four words maximum on the image, high contrast, one focal point. Shrink the thumbnail to the size of a fingernail and see whether it still reads.
• Subject first in the title. The recognisable noun at the front, because titles truncate and because recognition is what gets the click.
• Consistency across the set. A viewer who liked one video should be able to recognise the next as yours from the thumbnail alone. This raises the rate on browse and suggested surfaces more than any single clever image does.
• Retiring your own worst performers. A back catalogue of poor thumbnails drags the channel average and quietly teaches the recommender that your impressions do not convert. Re cutting the thumbnails on ten old videos is usually the cheapest available improvement.

What not to do, even though it raises the number

A thumbnail that promises something the video does not contain will lift click through rate and lower average view duration in the same week, and the second number is the one that decides whether YouTube shows the video to anybody else. The honest version of this advice is positive rather than cautionary: pick the most interesting thing that is genuinely in the video and put that on the thumbnail. There is almost always one, and finding it is the actual creative work.

Questions about click through rate on YouTube

What is a good CTR on YouTube?

YouTube publishes no figure, and the help article that defines the metric gives no range. The usable answer is your own median by traffic source, measured across your last twenty videos, because a channel discovered through search and one discovered through the homepage cannot be compared on a single number.

How does YouTube define impressions click through rate?

As "How often viewers watched a video after seeing a thumbnail", read on 31 October 2026. The documentation defines the impression conditions in detail and does not publish a formula for the rate itself.

Why is my CTR low when my views are high?

Usually because a large share of those views arrive from places that count no impression, such as embeds on external sites, notifications, or the YouTube Music app. Views from those surfaces are in the numerator of nothing and absent from the denominator, so a well watched video can show a modest rate.

Does a one second rule mean scrolling past counts?

Only if the thumbnail was on screen for more than one second with at least half of it visible. A fast scroll produces no impression, which is why a channel's impression count is lower than its apparent exposure.

Do end screens and cards affect CTR?

No. Content within the video player, including cards and end screens, is explicitly excluded from impressions, so in player promotion has to be judged on its own reports rather than on this metric.

Should I change thumbnails on old videos?

Yes, on the worst performers, and it is usually the fastest improvement available to an established channel. Change one variable, give it two weeks of comparable impression volume, and read the source level rate rather than the channel average.

Where a thumbnail has no job at all

Every number above describes a surface where somebody sees an image and decides. There is a growing surface where nobody sees one. When a viewer asks an assistant which channel to watch to learn a particular thing, or which creator covers a subject best, the answer comes back as two or three names with a sentence each, assembled from text: your video descriptions, your website, articles about you, and transcripts. No thumbnail is involved and no impression is counted anywhere. For a channel built as a business rather than as a hobby, that is a second discovery route with its own rules, and tags and metadata do less for it than people expect, as the reading of YouTube's own position on tags in what YouTube says tags do sets out. AI Knows Us takes the reading of that second route, asking the questions a viewer or a buyer would ask and recording which names come back and what each answer was built from.