YouTube cuts reach of reuploaded Shorts as cheap AI video reshapes creator payouts

YouTube said on 2 October that it has adjusted the recommendation systems behind Shorts. Content reuploaded from other creators, with nothing added, now gets less reach. 9to5Google reported the change the same day. It noted that YouTube has not published a rollout date, though the change appears to be live already.

The company's wording, as quoted by 9to5Google, is narrow. The update aims to "further prioritize original content and reduce the reach of content re-uploaded from other creators without adding anything of your own." What counts as adding something is not defined.

That gap matters for every channel that treats Shorts as a distribution channel for clips it did not shoot. There is a second ambiguity. YouTube began testing a feature in March that lets users reimagine other channels' videos with AI. 9to5Google says it is unclear whether those videos will still be surfaced in recommendations. The same platform is therefore tightening one door on reused footage while opening another to AI-assisted reuse. The rules for the second door have not been spelled out.

Cheap generation, expensive direction

The economics of that second door are getting harder to ignore. Sweat AI, an AI-native BPO for banks and fintechs, published an account on 3 October. It made twelve short AI clay films in five days for $184 in cash, using Claude Code to run the production pipeline from script to upload. The first post went out on 27 September. By 2 October, at 14:57 UTC, the films had drawn 18,971 views and 25 followers across Instagram, YouTube Shorts and TikTok.

The company's numbers are specific enough to be useful. A finished film cost $7 to $11 to generate on fal.ai, including rejected takes and experiments. An ElevenLabs startup grant covered narration, music and the first three films' images and video, worth $112 at list price. The coding agents ran on subscriptions Sweat AI already pays for engineering: a Claude Max 20x plan at $200 a month and a ChatGPT Pro plan. Priced at API rates, the agent tokens would have added about $268. Paying list price for the ElevenLabs work would have added another $112. That takes the total to roughly three cents per view instead of about one cent.

Sweat AI's own writeup flags the awkward part. One long conversation kept re-reading its own history, pushing some films as high as $47 at API rates. The agent directing the film could cost more than the model animating it.

That is a cost line most creator monetisation calculators do not have a box for. It sits on the production side, before a platform's recommendation system decides whether anyone sees the result. The view counts in the same post are a caution against reading too much into a single launch. Five of the first seven films posted after the afternoon of 30 September had between 740 and 1,398 YouTube views as of 2 October. Sweat AI says it cannot explain the spread. The company also reports that Instagram viewers stayed past three seconds at higher rates when the first line named one strange, physical thing. Three office comedies held 49 to 53 percent. Those are one publisher's numbers, not platform data.

What the platform pays for, and what it won't show

YouTube's recommendation change is not a monetisation policy. It does not touch payouts, ad rates or the eligibility thresholds that decide whether a channel earns anything at all. It decides who gets shown. For Shorts, that is the input to every revenue calculation that follows. A channel that loses recommendation reach loses the views that any revenue share would be applied to.

The timing is uncomfortable for creators building on reuse. Cutting portions of long-form videos has been one of the easiest ways to produce Shorts with high view potential, as 9to5Google notes. The same outlet points out that YouTube has not drawn a clear line between a clip that adds commentary and one that adds nothing. Enforcement will be a judgement call made by systems the company has described in a single paragraph.

For context on how much money is at stake on the other side of the recommendation decision, recent headlines point to a creator monetisation market estimated at $29.07bn in a 2026 analysis report. They also point to YouTube monetisation rules that may get tougher on earning requirements, reported in August. Neither figure comes from YouTube. The company has not published Shorts revenue data in public material. Treat the market sizing as an analyst estimate, not a platform disclosure.

YouTube has been shipping creator-facing features steadily. A 9to5Google report lists Custom Feeds, GIF comments and a "Your podcast lineup" feature among recent announcements. None of those change the payout maths. They show a platform adding surfaces faster than it publishes rules for them, which is the pattern the Shorts update fits.

The legal bill for synthetic video

Outside the recommendation system, synthetic video is already producing court outcomes. The BBC reported on 2 October that the Arizona Court of Appeals quashed the sentence of Gabriel Paul Horcasitas. He was convicted of shooting and killing Christopher Pelkey, 37, at a red light in 2021 and sentenced to 10 years. The appeals court ruled that an AI-generated video message from the victim should not have been shown at the 2025 sentencing. Pelkey's family made it from voice recordings, videos and pictures.

"Rather than document an event or recording a particular moment, the AI video presents a depiction of the victim and his thoughts created from the imaginings of the victim's sister," the court of appeals wrote on Wednesday, according to the BBC.

The trial judge, Todd Lang, had welcomed the clip at the time. "I loved that AI, thank you for that," he said, according to the BBC. "As angry as you are, as justifiably angry as the family is, I heard the forgiveness."

Horcasitas, 55, must now be resentenced. The case is a criminal one, not a platform dispute. It establishes that a generated video of a real person can be ruled inadmissible. The reasoning turns on who authored the depicted thoughts, not on how convincing the output looked. That distinction is the one platforms are now being asked to automate. YouTube's Shorts update asks its systems to tell original from reuploaded without a published test. Sweat AI's post argues the opposite direction for production. Its agents turned each problem found into a check the next film had to pass, but every call of taste came from a person. The company labels every post as AI-made and renders real people as clay caricatures.

What no public account shows is a platform paying creators more for that labelling or for the human review time. Sweat AI explicitly excludes its own review hours from the $184 bill. YouTube's change may reward originality in reach, but reach is not a payout. No figure connects the two.

Who ends up holding the risk

The practical effect of the 2 October change depends on details YouTube has not supplied. Does the system look at visual matching, audio matching, or account behaviour? How does it treat a clip that is reuploaded with a new voiceover? 9to5Google's report leaves those questions open. Creators who built channels on compilation and reaction formats now face a recommendation penalty whose trigger they cannot inspect.

Meanwhile the tools that make original-looking video cheap are improving on a weekly cadence. Sweat AI's $184 week produced twelve films and 18,971 views. That ratio would have been implausible for a crew of humans with no budget. If generation keeps getting cheaper, the scarce input becomes the judgement about what to make, and the platform's judgement about what to show. Neither is priced in the $7 to $11 per film figure.

YouTube has said it wants to "focus on originality," in 9to5Google's phrasing of the company's release. The company has not said how it will measure that, when the change fully takes effect, or whether AI-reimagined videos count as original. Until it does, creators are making production and payout decisions against a rule they can read in one sentence and cannot test.