For most of YouTube's life, a family of four watching a video on the living room TV counted exactly the same as one guy watching it on his phone in a parking lot. One playback, one view. Everyone in the measurement business knew that was wrong. We just did not have anything better from the platform itself.
On September 5, YouTube quietly changed that. There is a new metric in YouTube Analytics called Views (co-viewed), and I think it is one of the more important things to happen to creator and publisher measurement this year. I also think a lot of people are going to misuse it. So let me lay out what it is, what it is not, and how I would actually put it to work.
What changed
YouTube says more than a billion hours of YouTube get watched on TV screens every day, and a big chunk of that happens in groups. The standard Views number has never accounted for that. It counts devices, not people.
Views (co-viewed) is YouTube's estimate of how many people were actually in front of the screen. If three people in a household watch your video together, the standard view count still says one. The co-viewed count says three. If that same household watches it twice, standard views says two and co-viewed says six.
That sits next to Unique reach, which arrived back in May and answers a different question: how many distinct people did this video reach, counting each person once no matter how many times they came back. Same family, same two viewings, unique reach says three.
So one video now has a small stack of audience numbers attached to it:
- Views - the public number on the watch page, one playback equals one view
- Views (co-viewed) - estimated total viewing occasions including everyone sharing the screen
- Unique reach - estimated distinct people, counted once each
- Engaged views and qualified views - the numbers YouTube actually pays on
Each one answers a different question. The trouble starts when someone grabs the biggest one and calls it "views."
How YouTube gets the number
There is no camera in your TV. Nobody is counting heads in your living room. YouTube is running a statistical model that looks at demographic patterns, video genre, and time of day, and predicts how many people were likely watching when a video played on a connected TV.
If that sounds familiar, it should. Television has been doing exactly this for decades. Nielsen's household ratings have always leaned on models and panels to translate a set that was turned on into a number of people who were watching. Every upfront that has ever been negotiated ran on modeled audience. So the idea that YouTube is now estimating co-viewing is not radical. It is YouTube catching up to how TV has been measured since before most of its creators were born.
What is different is who gets to see the number and who gets to check it.
What did not change, and why that matters
Three things stayed exactly where they were, and I would tattoo them on the inside of every media kit if I could.
First, the public view count is untouched. The number your audience sees on the watch page is the same device-level count it always was.
Second, the money is untouched. Revenue sharing, AdSense payouts, and Partner Program eligibility still run on engaged views and qualified views. Your co-viewed number can triple and your check does not move.
Third, this is a private metric. It lives inside YouTube Studio, visible to the channel owner. A brand cannot log in and verify it. A buyer at an agency cannot pull it from Google Ads. If a creator shows a sponsor a co-viewed figure, the sponsor is taking the creator's screenshot on faith.
That last point is the one I keep coming back to. YouTube has positioned this as a number creators can bring to brand deal conversations, and it is. But we just handed one side of the negotiation a bigger audience figure that the other side has no way to audit. Anyone who has sat on the buying side of a sponsorship knows how that goes over. The first few times a creator walks in with a co-viewed number, the reaction is going to be "says who."
How I would actually use it
I spent the better part of a decade building measurement for Snapchat, Instagram Stories, Facebook Live, and TikTok at Delmondo, and the pattern with every new platform metric was the same. Early on, people either ignore it or overclaim on it. The ones who win are the ones who figure out the honest use case fast. Here is mine.
- Treat co-viewed as a TV audience multiplier, not a replacement for views.** The right way to talk about it is "our standard views were X, and YouTube estimates the TV audience for that content at Y once you account for shared screens." Label it as estimated. Show both numbers. If you only show the bigger one, you are going to get caught, and you deserve to.
- **Pair it with unique reach every single time.** The co-viewed metric answers "how many viewing occasions did we generate." Unique reach answers "how many actual people did we get in front of." A sponsor asking about reach wants the second number. A sponsor asking about frequency and total exposure wants the first. If you hand over one without the other, you are answering a question they did not ask.
- Watch the date range.** Co-viewed data takes up to 48 hours to process. If you pull the last 28 days or this month, the most recent two days are going to be short, and the number will look lower than it really is. Set a custom range that ends at least two days back. This is a small thing that is going to produce a lot of wrong numbers in a lot of decks.
- For sports and live-adjacent content, this is a big deal.** If you run social for a team, a league, a network, or a show that people gather around, connected TV is where your audience actually watches. Postgame shows, highlight compilations, documentaries, watch-along content. These are living room formats. Your standard view count has probably been undercounting your real audience for years, and now you have a platform-backed way to size the gap. Just be honest about how you got there.
- Understand what the model can and cannot see.** YouTube has not published an error range, a retraining schedule, or any third-party validation. The model is built on genre, demographics, and timing. That means a video in a genre YouTube associates with family viewing, watched at 8pm on a Saturday, is going to get a healthy multiplier whether or not anyone was actually on the couch. Treat the number as directionally useful and internally consistent, not as a headcount.
The bigger picture
Step back and look at what YouTube has done in the last few months. On August 24 they changed the public view count to start from the first frame across every format. In May they added unique reach. Now co-viewed. Meanwhile the numbers they pay on have not moved.
YouTube is building two separate measurement systems. One is the audience story, and it keeps getting bigger and more TV-like. The other is the monetization ledger, and it stays conservative. There is nothing wrong with that. Television has always had a gap between the audience it claims and the impressions it sells. But it means the people reading these numbers need to know which system they are looking at, and right now most of them do not.
That is where the measurement job actually lives. Not in the metric itself, but in translating it. Views (co-viewed) is a good number. It is also a number that arrives with no audit trail, in a private dashboard, at the exact moment a creator is trying to justify a rate. If you are on the creator or publisher side, use it carefully and show your work. If you are on the buying side, ask for it alongside unique reach and standard views, and ask how the range was pulled.
The couch was always full. YouTube just started counting the people on it. The rest of us have to figure out what to do with that honestly.
*Nick Cicero is the CEO of Mondo Metrics, a measurement company for social, video, podcast, and newsletter audiences in sports, media, and entertainment. He previously founded Delmondo, acquired by Conviva in 2018, and writes All Signal No Noise on Substack.*