How the TikTok algorithm hands out views, hour by hour: 38 AI films measured

When you post a TikTok, how long until you know how it did? We measured it on our own films. Theia, our AI content studio, runs several TikTok accounts on auto-post: she makes each film, posts it, and reads its views every 20 minutes for the first hours. This post looks at 38 of those films, posted from 6 October to 4 November 2026, on two accounts.

We cannot see inside the TikTok algorithm. What we can see is when the views arrive, and that is what most people mean when they ask how it works. The short answer: it depends on the account, and on one of ours the answer changed how Theia scores her films.

At a glance

	Main account	Small Problems
Films measured	31	7
Format	AI films in several formats	Clay animation about daily annoyances
Usual posting time	About 12:00 UTC	About 13:30 UTC
Share of 48-hour views at 1 hour (median)	39%	8%
At 3 hours	89%	35%
At 6 hours	97%	65%
At 12 hours	98%	97%
At 24 hours	99%	99%
Lowest share at 6 hours	86%	23%

"Share of 48-hour views" means the views a film had at that hour, divided by the views it had at 48 hours.

Main account: one burst, done in 3 hours

On our main account the pattern barely changes from film to film. About a third of the views come in the first 30 minutes. Then there is a pause, then a burst between 1 and 3 hours after posting. By 3 hours the median film has 89% of its 48-hour views, and by 6 hours 97%.

Before6 hoursAfter

30 minutes

34.3%

1 hour

38.6%

2 hours

57.1%

3 hours

88.7%

6 hours

96.8%

12 hours

98.1%

24 hours

99.3%

Median of 31 films, posted 6 October to 4 November 2026.

The films that did better than usual followed the same clock. Two films from 11 and 12 October reached about 700 views, three times the account's usual first batch of about 230. They had 80% and 82% of their 48-hour views at 3 hours and 99% at 6 hours. The bigger audience came inside the same burst, not after it.

After 48 hours almost nothing changes. For the 25 films we could follow for a full week, views at 7 days were a median of 1.2% higher than at 48 hours, and at most 6.6% higher.

Small Problems: two waves, the second one late

Small Problems is a second account that Theia runs, with clay animation about small daily annoyances. Its films post at about the same time of day, but the views arrive very differently. There is no early burst. Views climb slowly for the first 6 hours, to a median of 65% of the 48-hour total. Then most films get a second rise between 10 and 15 hours after posting.

Before6 hoursAfter

30 minutes

3.8%

1 hour

7.8%

2 hours

24.8%

3 hours

34.7%

6 hours

64.6%

12 hours

96.7%

24 hours

99.3%

Median of 7 films, posted 27 October to 4 November 2026.

The second wave matters most for the best films. The film posted on 29 October had 79 views at 6 hours, 23% of its 48-hour total. Between 11 and 15 hours it went from 133 to 331. The film posted on 6 November did the same: 88 views at 6 hours, 336 by 21 hours. In both cases the fastest growth was at about 03:00 UTC, late evening in New York.

Small Problems film	Views at 6 hours	Views at 24 hours	Share at 6 hours
29 October	79	339	23%
6 November	88	336 (at 21 hours)	26%
Median of the other 6	77	111	65%

The 6 November film is not in the 38, because it is not 48 hours old yet, so its share at 6 hours is measured against its views at 21 hours.

What this means for scoring a film

Theia scores each film on its views at 6 hours, so she can learn from one film before she makes the next. She has scored at 6 hours since 20 October, based on the main account's data, as we wrote in How Theia learned to make TikToks. On the main account that choice holds: a 6-hour score is within a few percent of the final number.

On Small Problems a 6-hour score alone would be wrong for the films that matter most. That is why the rule has a second part: a film that is still growing (more than 5% from 3 to 6 hours, or 400 or more views at 6 hours) is read again at 24 hours, 48 hours and 7 days. Every Small Problems film grew more than 5% between 3 and 6 hours, so the rule sends every one of them to the later reads.

The lesson for anyone scoring their own TikToks: measure your own account's curve before you pick a cut-off. The same posting time and the same studio gave two accounts curves that look nothing alike.

Likes arrive with the views

Likes follow the same clock as views. On the main account, the median film had all of its 48-hour likes by 6 hours. On Small Problems the median film had two thirds of its likes by 6 hours, the same share as its views. So on an account like Small Problems, a like rate read at 6 hours describes only the first part of the audience.

How we measured

• Theia's backend reads each film's public view, like, share, save and comment counts about every 20 minutes for the first 6 to 8 hours, then less often.
• The posting time comes from the TikTok video id, which carries the time the video was created.
• We kept films with a first reading within 30 minutes of posting and readings past 48 hours: 31 on the main account and 7 on Small Problems.
• Between readings we drew a straight line to estimate the views at an exact hour.
• 60 of 1,463 readings were lower than an earlier reading for the same film by more than 5%, some of them zero. Views do not go down, so we treated these as failed reads and left them out.
• Data read on 7 November 2026.

What this does not tell you

• It is two accounts and 38 films. The main account usually stops at about 230 views, Small Problems at about 100 to 120. Accounts with bigger audiences may get views over a longer period.
• It does not say why the two accounts differ. The format, the audience, the follower count and the account's history all differ, and we have not tested any of them on its own.
• It covers TikTok only. On Instagram Reels we have seen a slower curve: a spoken scene that got 1,507 views by about 23 hours got most of them after 6 hours.
• It does not tell you how the TikTok algorithm ranks videos, only when the views arrive.

More

• How Theia learned to make TikToks: the whole story, from slideshows to spoken scenes.
• What Theia learned, week 3: the films from 31 October to 5 November, including the first one past the main account's usual ceiling.
• Theia's first 100,000 views: four accounts, one week, and what broke out on Instagram.
• Recursive self-improvement: how Theia changes her own code from these numbers.
• Do TikTok hooks predict views?: an open AI model scored 2,881 public TikToks from 119 creators, and what went with more views.