You get 1 million views on YouTube by getting recommended, because YouTube says recommendations drive more viewing than subscriptions or search. That means a title and thumbnail people click and a video they rate highly once they've watched, so test both with Test and compare and build every video around satisfaction, not just the click.
Most advice on hitting a million views is a list of tricks. I'd rather start with what YouTube itself has said about how videos travel, because that tells you where to put your effort. The short version: a million views almost never comes from your subscribers. It comes from strangers, sent to you by the homepage and the Up Next panel.
Where a million views comes from
YouTube's own explainer on its recommendation system says recommendations drive a significant amount of overall viewership, more than channel subscriptions or search. The system learns from over 80 billion signals a day: clicks, watchtime, survey responses, shares, likes and dislikes. So your job is to give it reasons to show your video to people who've never heard of you.
The signal people forget: did viewers value it?
YouTube found long ago that a click doesn't mean someone watched. When it first added watchtime to recommendations, views dropped by 20% overnight, and it accepted that because it wanted viewers satisfied, not just clicking. Today it goes further: it surveys viewers, asks them to rate a video from one to five stars, and only counts watchtime from four- and five-star ratings as valued watchtime.
What that means for you: a misleading thumbnail might win the click and lose the recommendation. Deliver exactly what the title promises, early.
A seven-step plan I'd follow
- Pick a topic strangers already want. Recommendations reach people who don't follow you, so the subject has to interest someone with no idea who you are.
- Write three titles and make three thumbnails before you film, and pick the idea where all three are easy to make. If it's hard to package, it's hard to recommend.
- Use Test and compare in YouTube Studio on a desktop. You can test up to 3 titles and thumbnails, and YouTube shows everyone the version with the highest watch time.
- Export thumbnails at 1280 x 720 or better. If any test thumbnail is below 720p, YouTube downscales all of them to 480p.
- Give the payoff the title promised in the opening, then keep earning the next minute. Valued watchtime is the goal.
- Stay well clear of anything borderline. YouTube demotes it, and after it started doing so, watchtime on recommended borderline videos in the US fell by 70%.
- Look at your best-performing video and make the next one for the same viewer. A million-view video often follows a few that reached a smaller version of that audience.
Reading the signs in YouTube Studio
You can't see YouTube's surveys, but you can see their shadow. In YouTube Studio, the Reach tab shows how often your thumbnail was shown and clicked, and the Engagement tab shows how long people stayed. A video that gets plenty of clicks but weak watch time has a packaging promise the content didn't keep. A video with modest clicks and long watch time usually has a packaging problem, and that's the one to put into Test and compare.
- Strong clicks, weak watch time: fix the opening and make sure the title tells the truth.
- Weak clicks, strong watch time: test new titles and thumbnails.
- Both weak: the topic may not interest strangers. Go back to step one.
Give each change time to show up. Recommendations build over days, not hours, so I'd judge a new title or thumbnail on a full week rather than the first afternoon.
What Test and compare can and can't do
| Detail | What YouTube says |
|---|---|
| Options per test | Up to 3 titles, thumbnails or both |
| Winner chosen by | Watch time, not clicks |
| Where it works | Desktop YouTube Studio, with advanced features on |
| Not available for | Shorts, Scheduled Lives and Premieres (until they end) |
| Time to finish | Within two weeks |
What a million views is worth when you get there
If you're monetized, YouTube pays 55% of net ad revenue on long videos. Run YouTube's own hypothetical CPM example through that split and a million long-form views comes out at roughly 770 US dollars. That's an estimate, and your viewers' location changes it a lot. It's also worth knowing whether YouTube Live has ads if streams are part of your plan, and whether video length affects monetization before you decide how long to make the video. For a country-level view, here's what 1,000 views pays in Canada.
My verdict
You don't need luck so much as a video strangers want to click and are glad they watched. Package first, test with the tool YouTube gives you, and judge success by watch time. I'd rather publish one video I'd bet on than ten I'm hoping about.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Frequently asked questions
What gets a YouTube video to 1 million views?
Recommendations. YouTube says they drive more viewing than subscriptions or search, and they're shaped by clicks, watchtime, surveys, shares, likes and dislikes.
Does clickbait help get more views?
It can win a click and still lose. YouTube counts only watchtime rated four or five stars in its surveys as valued watchtime, so disappointed viewers work against you.
Can I A/B test thumbnails on YouTube?
Yes, with Test and compare on desktop YouTube Studio. You can test up to 3 titles and thumbnails and YouTube picks the winner by watch time.
Can I use Test and compare on Shorts?
No. YouTube says it isn't available for Shorts, Scheduled Lives or Premieres until the Premiere ends.
Do subscribers help a video reach 1 million views?
They help it start, but YouTube says recommendations drive more viewing than subscriptions or search, so most of a million views usually comes from people who don't subscribe.
How much is 1 million views on YouTube worth?
Using YouTube's own CPM example and the 55% long-form share, roughly 770 US dollars. Real earnings vary with where viewers live.
