You’ve uploaded a video, written what feels like a decent title, and waited. A week later: 47 views, 38 of which are probably you refreshing the page. Meanwhile, a channel that started two months after yours has thousands of views on every upload. The frustrating part isn’t the gap — it’s that you don’t know why the gap exists. Understanding what signals YouTube uses to recommend videos is the difference between posting into the void and actually getting found.
Here’s the scale of the problem: YouTube processes over 500 hours of video uploaded every single minute, according to YouTube’s own press data. The algorithm isn’t watching your video and deciding if it’s good. It’s reading specific, measurable signals your viewers generate — and then deciding whether to push your video to more people or quietly stop showing it. Most beginners optimize for things that don’t move those signals at all. This article breaks down the five that actually matter.
What Signals YouTube Uses to Recommend Videos (And Why Views Aren’t One of Them)
Raw view count tells YouTube almost nothing useful. A video can have 10,000 views and still be a signal that the algorithm treats as a dead end — if those views came from people who clicked away after 8 seconds. What YouTube is actually trying to figure out, according to its own research paper published by engineers Covington, Adams, and Sargin, is two things: satisfaction and relevance. Every metric below is just a different way YouTube measures those two things.
The recommendation system works in two stages. First, YouTube narrows down the entire library to a few hundred candidates that might fit a specific viewer. Then it ranks those candidates against each other using engagement signals to decide which one actually gets shown. If your video doesn’t generate the right signals in Stage 1, it never even makes it to Stage 2. That’s why a technically well-made video with bad retention data gets buried — it doesn’t pass the first filter.
Takeaway: Stop chasing view count. Start tracking the five signals below, because those are what actually move your video through the recommendation system.
Signal #1: Click-Through Rate — The First Test Your Video Has to Pass
CTR — that’s the percentage of people who see your thumbnail and title and actually click on your video — is the first signal YouTube reads. When YouTube shows your video as a suggestion and someone clicks it, that’s a positive vote. When they scroll past it, that’s a negative one. According to YouTube Creator Academy data, most videos on the platform sit at a CTR between 2% and 10%, with the average for small channels landing around 2–4%. Anything above 6% on a channel with under 10,000 subscribers is genuinely strong.
Here’s what most beginners miss: CTR isn’t fixed. It changes based on where your video is being shown. A thumbnail that performs at 7% CTR on your subscriber feed might only pull 2% on Browse (that’s the homepage YouTube shows to people who aren’t subscribed to you). YouTube tests your video in small batches first. If the CTR is strong in that test group, it gets shown to bigger audiences. If it’s weak, distribution slows down or stops.
To check your CTR: Go to YouTube Studio → Analytics → Reach tab. You’ll see your impressions (the number of times your thumbnail was shown to someone) and your CTR next to it.
If your CTR is below 2%, your thumbnail is almost certainly the problem. Test a new thumbnail — change the facial expression, increase contrast, reduce the amount of text on the image to five words or fewer — before you touch anything else.
Takeaway: Check your CTR in YouTube Studio right now. If it’s below 2%, your thumbnail needs to change before your next upload.
Signal #2: Average View Duration — How Long People Actually Stay
AVD — Average View Duration — is exactly what it sounds like: the average number of minutes and seconds viewers actually watch before clicking away. YouTube uses this to measure whether your video delivered on the promise your thumbnail and title made. A 10-minute video where people stay for 7 minutes sends a very different signal than one where they leave after 90 seconds.
According to data from vidIQ’s 2023 channel study, videos with an AVD above 50% of their total length are significantly more likely to be recommended in the “Up Next” sidebar and on the homepage. For a 10-minute video, that means keeping people watching for at least 5 minutes on average. For shorter content, the standard is tighter — a 3-minute video where people drop off after 45 seconds is a strong negative signal.
The place most channels bleed AVD is the first 30 seconds. That’s where the average viewer decides whether to keep watching. Aim to have at least 70% of your viewers still watching at the 30-second mark — you can check this in YouTube Studio → Analytics → Engagement tab → Audience Retention. If you’re losing more than 30% of viewers before the 30-second point, your intro is too slow.
Cut the intro music. Cut the “don’t forget to subscribe.” Start the video with the most interesting thing you’re going to say. The first sentence should either answer a question, create one, or show something unexpected.
Takeaway: Open your last video’s retention graph. Find the biggest drop-off point in the first 60 seconds and identify exactly what you said right before it. That’s what to cut next time.
Signal #3: Likes, Comments, and Shares — Engagement That Actually Moves the Needle
Engagement signals — likes, comments, shares, and saves — tell YouTube that your video created a reaction strong enough for someone to act on it. Of these, shares are weighted most heavily, according to statements made by YouTube’s Chief Product Officer Neal Mohan. When someone shares your video outside of YouTube — via text, WhatsApp, Reddit, Twitter — YouTube registers that as a strong satisfaction signal because it required effort from the viewer.
Comments are the second strongest signal, specifically because they indicate that someone felt something or thought something worth typing out. A video with 200 views and 40 comments sends a better signal than a video with 2,000 views and 3 comments. Likes matter, but less than most creators think — they’re a low-effort action, which is why YouTube weights them lower than shares or comments.
One thing that’s often overlooked: how quickly engagement accumulates in the first 24–48 hours matters more than the total. A video that gets 50 comments in its first two days signals to YouTube that it’s generating strong early reactions. That triggers broader distribution. A video that gets the same 50 comments spread over three weeks doesn’t create the same spike.
This is why sending your video to your most engaged viewers first — via community posts, email, or social — can have a real algorithmic effect. It’s not just about the numbers. It’s about the speed.
Takeaway: On your next upload, ask one specific question in your video and repeat it at the end. A direct, easy-to-answer question in the comments section can double your comment count, which sends a stronger engagement signal in the critical first 48 hours.
Signal #4: Post-Video Behavior — What Viewers Do After They Watch
This one surprises most beginners. YouTube doesn’t just track what happens during your video. It tracks what viewers do after it ends. Did they watch another video on YouTube? Did they close the app? Did they click to a completely different channel? This is sometimes called “session watch time” — the idea that YouTube rewards videos that keep viewers on the platform, not those that send them away.
According to YouTube’s own Creator Academy documentation, videos that lead viewers into longer YouTube sessions are actively preferred by the recommendation system. This is why end screens — those clickable links to other videos that appear in the last 20 seconds of your video — aren’t a nice-to-have. They’re a signal. If viewers click your end screen and keep watching YouTube, that’s a positive post-video signal attached to your content.
To add end screens: Go to YouTube Studio → Content → click your video → Editor → End Screen. Add at least two video links in the last 20 seconds. Point viewers to your most-watched video or the video most likely to be relevant to what they just watched.
What you want to avoid: a video ending so abruptly that viewers close YouTube entirely. Even saying “here’s another video you might like” on camera while your end screen appears has been shown anecdotally to increase end screen CTR — because you’re giving viewers a reason to click before they decide to leave.
Takeaway: Add end screens to your last five videos if you haven’t already. Two linked videos in the final 20 seconds is the minimum. Verbally direct viewers to them on camera.
Signal #5: Survey Data and Viewer Satisfaction — The Signal You Can’t Game
YouTube periodically shows viewers a short survey after watching a video: “Did you enjoy this video?” or “Do you want to see more content like this?” Most creators don’t know these surveys exist, let alone that they’re feeding directly into what signals YouTube uses to recommend videos. According to YouTube’s research team, these satisfaction surveys were introduced specifically because high watch time alone wasn’t a reliable indicator of whether viewers actually enjoyed what they watched.
This means you can technically hold someone’s attention with clickbait — fast cuts, misleading titles, artificial tension — and still get punished by the algorithm if the survey data shows viewers felt unsatisfied afterward. YouTube calls this “optimizing for long-term viewer happiness,” and it’s one reason why the platform has explicitly stated it penalizes misleading thumbnails and titles even when they initially drive high CTR.
You can’t directly access your survey scores in YouTube Studio — YouTube doesn’t surface this data to creators. What you can do is monitor your “likes to views” ratio as a rough proxy. A likes-to-views ratio above 4% generally indicates strong satisfaction. Below 1% on a video with decent view count is a warning sign that viewers felt the content didn’t deliver.
The only real strategy here is an honest one: your thumbnail and title should accurately represent the video. Under-promise and over-deliver. If your title says “How I got 10,000 subscribers in 60 days,” the video had better explain exactly how you did that — not give a vague five-minute pep talk.
Takeaway: Compare your last three videos’ likes-to-views ratios in YouTube Studio. If any are below 2%, look at the title and thumbnail — there’s likely a mismatch between what you promised and what you delivered.
What to Do When You’ve Optimized the Signals But Still Can’t Get Traction
Here’s the honest reality: even a well-optimized video needs an initial audience to generate signals. YouTube’s algorithm can’t amplify engagement it can’t measure — and if a new video has only been seen by 20 people, the signal sample is too small to trigger broader distribution. This is the catch-22 that kills a lot of genuinely good channels early on. If you’re in that position, it’s worth looking at Flintzy’s YouTube promotion service, which helps creators get a real first wave of views on new content — not bots or inflated numbers, but actual exposure that gives the algorithm enough signal data to work with. Sometimes the content is ready; it just needs the initial push to get the feedback loop started.
The One Thing to Check in YouTube Studio Today
Understanding what signals YouTube uses to recommend videos only matters if you act on it. So here’s your one specific task: open YouTube Studio right now, go to Analytics → Reach tab, and look at the CTR on your last three videos. If any are below 2%, that’s your first problem to fix — before you worry about retention, before you worry about engagement. A low CTR means the algorithm is testing your video, viewers are seeing it, and they’re choosing not to click. Fix that thumbnail first. Everything else you’ve learned today — the retention curves, the end screens, the engagement timing — those compound on top of a CTR that’s actually working. Start there.