OpenAI shut down the Sora app for a two-sentence social media post in late March, but the reason for its downfall had nothing to do with video quality. The company said it was “saying goodbye to the Sora app,” according to the Associated Press, after months of pressure over intensive use by Michael Jackson, Martin Luther King Jr., and Mr. Rogers forced OpenAI to be scaled down before it interfered with family areas and the actors’ union. Sora didn’t fail because the model couldn’t produce a convincing video. It failed because no one built a trust infrastructure around it before releasing the notification box to the public.
This week, YouTube enforcement data told a related story from the other side. In January, the platform permanently removed 16 channels under what we now call a fake content policy, a July 2025 rebranding of the old duplicate content policy. Those channels had 35 million subscribers and 4.7 billion views, and were producing generic, mass-produced video with no human input behind it, according to the Hollywood Reporter.
Both stories are about the same failure. It’s not really about video AI getting better or worse. They’re about what happens when the scale exceeds human judgment to sit on, and that gap is exactly what I created my 5-Pillar Framework for AI content to close back in April. Four months later, it doesn’t need much updating.
Average Costs Have Just Dropped Again, Boosting The Numbers
Two days before Sora’s shutdown made headlines, Google published a blog post announcing the Veo 3.1 Lite, its low-cost video production model, priced at less than half the Veo 3.1 Fast with the same speed. Developers can now produce four-, six-, or eight-second clips in landscape or portrait at up to 1080p, designed specifically for high-volume applications.
That’s not a criticism of the tool. It is the truth that deserves the basic truth. The cost of producing video at scale keeps falling, which means the pressure Pillar 1 of my framework is built to manage – the temptation to treat AI as a shortcut rather than an infrastructure – will only increase. Cheaper production makes initial discipline more necessary, not less.
The Human Face Has Become a Signal of Trust, Not Just a Style Choice
Craig Billings runs a science channel called Doctor NOS with 1.7 million subscribers, and he told The Hollywood Reporter that the no-nonsense channels that cover his site are being hit hard by the crackdown. Most of them are monetized, he said, while creators who have never touched AI but never shown their faces are caught in the same net.
That’s the real cost of imperfect implementation, and it’s worth mentioning in good faith, but it also confirms something that my framework has already argued for in Pillar 5. YouTube’s policy page, “How Creators Use AI to Create Content,” clearly states that the platform requires creators to disclose when AI was used or to produce real content, and that song can edit short videos embedded in long videos. If a creator skips disclosure and YouTube’s systems detect AI anyway, the label is automatically applied, and creators can’t remove it once there’s a high level of confidence that it’s AI-generated.
Four months ago, I wrote that hiding the use of AI is read as weakness to a sophisticated audience and that disclosure is read as skill. That is no longer just a trust strategy. It’s now embedded in the actual infrastructure of the arena, and treating it as the PR polish of choice is a strategic mistake, not just a missed opportunity.
What Effective AI Video Really Looks Like
Compare slop channels with Think and the new Google Creativity Edition guidance documents. Matthew Carey of Google Creative Lab described the creation of a short film assisted by AI ANCESTRA by deliberately avoiding general information, encouraging the capture of the cosmos using the descriptions of specific microscopes and lights rather than the word cosmos itself, because the transparent information produces a visual measure of every automatic model. Monks co-founder Wesley Haar, later told the same publication that companies that succeed with AI have done a poor job of integrating exactly what their brand is before they produce a framework.
There is no example that treats AI as a volume machine. Both take it as a capacity to act sitting under a certain human judgment about what belongs on the screen and what doesn’t. That’s Pillar 1 and Pillar 5 working together, and it’s the difference between YouTube’s cut channels and the case studies Google now touts as the industry standard.
The Trust Gap Is Wider in the USA than in the UAE
There is a market sector in this that American advertisers tend to weigh less. In a YouGov survey of 19 markets I covered in July, the US had the lowest rate of AI-assisted searches of any country surveyed, at 48%, compared to 89% in India, Indonesia, and the UAE, and only 28% of US searchers said they trusted an AI assistant’s answer at all.
I teach a module called “Engaging Audiences with Content in the AI Era” at the New Media Academy in the UAE, in an environment where AI-assisted adoption has become the norm instead of the exception. The lesson is not that Americans are wrong to be skeptical. That the disclosure requirements and personal judgment built into Pillar 5 are not good for state property. They are the basics that a skeptical American audience needs and a receptive Emirati audience will expect anyway once law enforcement reaches acceptance.
3 Updates You Can Make Before Your Next AI Video Goes Live
First, check whether your disclosure practices meet the platform’s policy languagenot your level of inner comfort. YouTube’s guidance says the labels apply to content that is photoshopped or purposefully altered, and creators lose the ability to remove that label if the system flags it with high confidence.
Second, set a price for your production plan against any tools like Veo 3.1 Lite now make it possible at scale, and deliberately choose to produce below what the ceiling allows. The technical ability to produce a thousand varieties does not oblige you to publish thousands of varieties.
Third, name the decision maker for every piece assisted by AI before it is shipped, the way Carey’s team is doing it at Ancestra and ter Haar’s team is doing it with Monks product knowledge bases. If no one can answer who decided this was the right cut, the piece is wrong.
My Take
The AI discussion in this field keeps being framed as a content quality issue, and I think framing it is wrong. It’s an infrastructure trust issue, and Sora, YouTube’s cleanup, and Veo’s falling costs all point to the same space from three different angles.
What I objected to in April is valid. The only thing that has changed is that the platforms have stopped arguing. That said, it can’t be done automatically, and the rapidly growing tools are what make crossing the human checkpoint all the more tempting. My draft didn’t need a rewrite this fall. It just needed the industry to reach Pillar 5.
Additional resources:
Featured image: Gorodenkoff/Shutterstock
