A rewatch is not a new viewer
A rewatch is the same person looping the file. Minutes rise. Unique viewers do not. Read a curve past 100% as replays.
A rewatch is not a new viewer. Rewatch is a second or later play of the same clip — a loop, a replay, the rise on a retention curve past 100%. The person did not multiply. The minutes did. Short-form players loop by default. A file that ends and starts again is a stay so complete the player ran it again. Treating that stay as reach is how a six-second joke looks like an audience, and how a unique-viewer target gets "hit" without a second human entering the room.
Loops are the player, not a crowd
TikTok, Reels, and Shorts return to frame one when the file ends. That is the product. A tight clip will write more plays than people. Instagram and TikTok both treat replays as a quality signal. A retention graph that climbs after the last frame is the signature: people watched twice, not that the runtime grew.
Watch time is the minutes. A rewatch is why those minutes can exceed the length of the file. Two hundred people who each loop a twelve-second clip once write as much watch time as four hundred who leave at the end. The first group is a smaller room that liked the join. The second is a larger room that did not come back. Unique viewers, unique accounts, unique impressions — those are the crowd. Plays are not.
YouTube's public view after 24 August 2026 is a first-frame start. A loop can add another start. It still is not another person. Engaged views stay on the stay side of that split. Do not paste a looped play count into a sponsor deck as "viewers." Name plays, unique people, and minutes as three columns.
Completion rate can look perfect on a six-second clip that nobody wanted a second time. The last frame was easy to reach because there was almost no file. Rewatch rate is how you tell a tight joke from a clip that merely ended before the thumb moved. Compare inside one length band. A six-second loop and a forty-second story are not on the same scale.
You earn a rewatch with a join, not a plea
First frame and last frame have to meet. Same motion, same caption position, a punchline that recasts the open. The second play should feel like a new read of the first second, not a restart after a logo. A three-second sting at the end is a loop-killer: the replay begins on a card the first play already finished with.
First-last-frame generation is the model version of that join. Veo 3.1 and the first-last rows are for a planned pair of stills, not for a plea in the voiceover. "Rewatch this" spoken over a hold does not stitch the frames. The picture has to continue.
Captions are part of the join. If the last cue is a different block than the first, the loop pops. Keep the caption style and the caption slot stable. Burned-in captions are pixels; they will be there on the second pass. A sidecar will not help a Reel. The feed will not draw it.
Do not put an end screen on a looping Short and expect a mosaic. Short-form often loops instead of showing next-up. The last beat is a CTA card in pixels, or a join back to frame one. A fake subscribe pack in the last second is a covered loop.
Padding to chase watch time without a join just delays the skip. Unused seconds at the end of a curve are not hours. They are a slope. Hold rate is that slope after the hook. A rewatch is a bonus on top of a hold that already worked.
Where unique-viewer math goes wrong
Reach briefs quote unique people. Looping files inflate plays. A seeding drop into a group chat inflates plays again, and those sends are not organic rewatches. Keep seeded plays out of the rewatch rate the same way you keep them out of shares-per-view.
A campaign that pays on unique reach but optimises the cut for loops will look strong in the platform graph and weak in incrementality. The same two thousand accounts watched twice. The unique column did not move. If the insertion order says unique, report unique. If it says completed views, say whether a loop counts as a second completion on that buy. Many buys do not say. Write it down.
YouTube long-form is less loop-shaped. Rewatches there are replays from the watch page, suggested, or a return visit. Average view duration can exceed runtime when replays sit in the mean. That is still not new unique viewers. Studio's audience reports separate returning and new when you ask them to. The public view counter will not.
X Original Content Rewards, from 8 August 2026, pays unique Premium Home Timeline impressions, not loops of the same account on the same post. A person refreshing is not the payout object. Do not analogise a TikTok rewatch rate onto that meter.
Design the loop, then count people separately
Cut for a join when the format loops. Cold open on an action that can restart. End on the motion that began. Burn the first line where it will still be legal on pass two. Watch the file twice on mute at phone size. If the second pass feels like a restart after a brand card, you killed the loop.
Count people in a different column. Unique accounts, unique impressions, new versus returning. Put rewatch rate next to them, labelled as plays per person or as the area of the curve above 100%. Do not add the two.
The AI video generator will give you another take. It will not tell you whether the last frame meets the first. That is an edit pass. Add captions to a video after the join is locked. Recaptioning a loop because the last cue ran long is cheaper than regenerating the shot, and it is the usual miss.
A rewatch is not a new viewer. Same person, more minutes. Keep the crowd count honest.
Lock the join, then caption it, in Versely.
FAQ
Does a retention curve over 100% mean more viewers?
It means some of the people who were already there watched past the end, usually because the player looped. Unique viewers are a different report. Read both.
Should I lengthen a clip to reduce loops?
Only if the extra seconds have a job. A longer file can raise unique minutes and destroy rewatch rate. Completion rate will also move. Decide which scoreboard the slot is on before you pad.
Do YouTube public views count each loop?
The public counter after 24 August 2026 registers a view when playback starts. A replay can write another start. It is still not another person. Use unique viewers or engaged views when the question is audience size.