AI movies, ranked by what they actually did

What counts as an AI movie on this page

A film qualifies here when two things are true: it is a film-like entry (it sits in one of the two format buckets on this site — Feature Films & Series / Short Films & Stories — or its title itself is about a film, series, episode or trailer), and its own title declares AI-generated visuals. Across the whole 3536-video dataset, 694 entries carry that declaration; 198 of them are film-like. Most of those sit outside the two film buckets — sci-fi, fantasy, romance, gaming and music content that is a film in everything but the label.

One correction this page has to publish, because the site used to get it wrong in its own navigation: the two format buckets were once named "AI Feature Films" and "AI Short Films". They are not AI buckets. They are runtime buckets — 155 entries for Feature Films & Series, 28 of which declare AI; 218 entries for Short Films & Stories, 81 of which declare AI. Calling 155 films an "AI movie" set would have been a claim the data does not support, so the names now describe the format and the AI subset lives here, with both numbers on the page.

"Declares AI-generated visuals" is a title-level claim, not a verdict: it is set when a video's own title says so (AI-generated visuals declared in the title), and unset when the title says the opposite or is about AI rather than made with it — tutorials, reviews and behind-the-scenes entries are tagged separately as Talking about AI, not an AI-made film and stay off this list. No frame is inspected, and a real AI film that never says so in its title will not appear on this page.

Four kinds of entry are deliberately left out, because each would make the number on this page bigger without making it truer. First, titles that say the opposite: a trailer marked "No AI" is a human-made video, and one of them used to sit at number 17 on this list until the rule was fixed. Second, entries whose only AI evidence was a label this site itself put on them rather than anything the video says — counting those is circular, so the evidence is now the title and nothing else. Third, the 12 titles that talk about AI rather than being made with it and would otherwise have been film-like enough for this list: tutorials, reviews, tool launches and behind-the-scenes breakdowns. They are tagged Talking about AI, not an AI-made film and stay off the AI boards, though they remain on their own niche boards and searchable on the radar. Across the whole dataset there are 100 entries with that marker; most were never film-like, which is why the two numbers differ. Fourth, productions built on someone else's IP — official trailers, fan-made and concept trailers for films that do not exist, remasters and recuts. 107 such entries are excluded from every board on this site: they are not a format you can reuse, and remastering another studio's footage carries a rights problem on top of the relevance one. A few small originals that wrap themselves in the same "official" and "teaser" wording land on the wrong side of that line; they are flagged for manual review rather than silently kept. Anything else stays in the dataset too, but cannot count as an AI film.

The AI films with the fastest view velocity

Ranked by breakout score — views per day relative to the channel's own subscriber base, weighted by whether engagement looks natural for the niche. Rank is not a quality judgement; it is a measure of which of these films a small channel managed to push out to a much bigger audience than it had. Every row opens the film on YouTube, and the channel name goes to its page on this site.

Top 30 of 198. Median film on this list: 11 minutes long, 119.5 days old, 3.1K views per day, 1.3× the dataset median score. Page rebuilt 2026-10-11.

AI short films: what the short-form end looks like

If you are searching for AI short films rather than features, the honest answer is that short-form is where almost all of this activity is. 92 of the 198 tracked AI films come in under 10 minutes and 58 sit between 10 and 30 minutes; only 48 run 30 minutes or longer. The shorts are also the cheapest place to study structure — the median short here is 3 minutes, which is long enough for a hook, an escalation and a payoff and short enough that you can watch a dozen in one sitting.

The short-form end of this list is disproportionately serialised rather than one-off. That is a pattern in the data, not an impression: 24% of these films carry the "Episodic series / numbered parts" marker, against 5% across the whole dataset — numbered episodes and parts are the format you can actually ship on a schedule. Ending on a hook, by contrast, is no more common here than anywhere else in the library ( 11% against 12%), so this page does not treat "cliffhanger" as an AI-film trait. The concept-trailer and franchise-reinterpretation end of this space used to be part of this list; it is now kept off every board (see the exclusions below), because a trailer for a film that does not exist is not a format a small channel can reuse.

Read the age column before the velocity column: longer films in this dataset are also older, so part of the gap between bands is time on the platform rather than format. This is one capture snapshot per video, so velocity is a lifetime average, not a reading of whether something is still climbing.

Who is actually making AI films

125 channels have at least one tracked AI film. These are the twenty with the strongest single release — the list is dominated by small operations, not studios, and several publish serialised work rather than one-off films.

A channel with one tracked film is a topic signal, not a repeatable format — the channel pages state each channel's tracked-video count for exactly that reason.

How these films say they were made

Most AI films are published without saying what made them. Across the 198 tracked films, 13% name a specific generation tool in the title, and the rest give you nothing beyond the AI declaration itself. That is worth knowing before you read any "these are the tools behind AI movies" list: on this dataset, the tool is usually invisible.

These are mentions, not production credits — a film naming a tool was not verified to have used it, and the counts are small enough that they should be read as "these names show up", not as market share.

Markers come from the same rule-based pass as the rest of the site (title patterns, runtime and market; source: rule-based inference from title + duration + market (hookType / openingSeconds / structure / repeatableChoices); provenance: api metrics + metadata only, no frame-level video analysis). A film can carry several markers, so the column does not sum to 100%.

AI movies are not only an English-language thing

156 of the 198 tracked AI films are English-language, and the remaining 42split across 8 other markets — Hindi film and short-story channels, Arabic comedy and drama shorts, Spanish and Portuguese genre pieces, and single tracked examples in Urdu, Thai, Chinese and Indonesian. That distribution is the argument itself: the format travels, and in almost every non-English market it is still thinly covered.

Market is the video's content language, taken from the YouTube API audio-language field where it exists and from title script and keyword evidence otherwise — it is an inference, not a setting the uploader chose.

Method and limits

"Declares AI-generated visuals" is a title-level claim, not a verdict: it is set when a video's own title says so (AI-generated visuals declared in the title), and unset when the title says the opposite or is about AI rather than made with it — tutorials, reviews and behind-the-scenes entries are tagged separately as Talking about AI, not an AI-made film and stay off this list. No frame is inspected, and a real AI film that never says so in its title will not appear on this page.

Three limits on top of that. This is a filtered dataset, not a census: every entry cleared the tracker's entry conditions (at least 20,000 views, at most 100,000 subscribers at capture, reach of 3× or more), so an AI film that did modestly, or from a channel that already got big, is absent. There is one capture snapshot per video, so velocity is a lifetime average and cannot tell you whether a film is still climbing or already peaked. And the ranking metric normalises against this dataset's own median, so the scores here are not comparable with any other tool's numbers.

Data: YouTube Data API v3 public metrics · 3536 tracked videos · 125 channels with an AI film · page rebuilt 2026-10-11 · ranking method documented on the breakout board.

Keep going

• All Feature Films & Series breakouts →
• All Short Films & Stories breakouts →
• Niches ranked by velocity →
• AI short films →
• How faceless niches are built →
• Compare short-form markets →