GoKawiilA hill I'm willing to die on: I don't consider content created entirely by an AI image or video generator "art."
This rule -- made by me, for me -- has occupied a lot of my brain space in 2025. Over the course of the year, we've gone from clunky, hallucination-ridden AI videos to clips that can be virtually indistinguishable from real videos. This year might seem like it's gone on forever, but the pace at which AI video has improved over the past seven months is truly frenetic. The same is true for image generation -- Google's Nano Banana and OpenAI's first image model are also only a few months old, as hard as that is to believe.
It's more than just the addition of audio to videos, although that was a big leap forward this year. Veo 3 proved cinematic AI video isn't an oxymoron, and Sora, the app and second-generation model powering it, showed us a terrifying glimpse into a future where your likeness is game for every internet weirdo's imagination. But if you can get over that queasy feeling (I still can't), it's also the best AI video model I've tested, with an undeniably impressive technical prowess that sheds itself of common AI errors.
Also this year, we heard more than ever from artists, creators and copyright owners that generative AI models are being created and deployed irresponsibly. Disney and Warner Bros. filed hotly worded copyright infringement lawsuits against Google and Midjourney, calling the latter "a bottomless pit of plagiarism." Anthropic announced a $1.5 billion settlement with the authors who accused it of piracy. And AI energy demands, which are especially high for video, have AI companies racing to build massive data centers across the US, despite concerns from the local communities and environmental experts.
I spend more time than most using these generative AI tools. These companies brand themselves as "democratizing creation" or "making it easier than ever to create art." That rhetoric was dialed up this year, as big tech companies, not exactly known for their creativity and compassion for creators, try to convince potential customers they know ball. The technical improvements we've seen with the new 2025 models, along with their viral popularity, have meant our online lives are filled with AI at an alarming rate. Undoubtedly, nothing AI makes is art. Period.
I expect we'll see a lot more of creative AI in 2026. It feels like a tide that won't slow down. So it's more important than ever to make a clear distinction between AI-generated content and truly human art. It will also be more important than ever to call out so-called AI "art" for what it is: pathetic, boring and unoriginal. While I'm still hopeful we'll get better AI labels, we need to rethink how we approach creative AI and the content (and slop) that it creates as it fills our online lives.
Don't miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source.
AI versus art
AI-generated content is a mimicry of human art. That's by design. These creative AI models are designed and refined using large swaths of human-generated data. For image and video models, that data includes photographs, designs and social media posts. The wider a model's training data, the more capable it is. For example, you can ask ChatGPT to create images in the style of Studio Ghibli (which many people did in March 2025). The model knew that the film studio created a specific cartoon/anime aesthetic and was able to apply that style to its own AI images.
Because of that process, AI rarely makes something new. In one of my favorite quotes about AI from this year, film writer (and former Meta AI data trainer) Nora Garrett told reporters while promoting her movie After the Hunt, "AI is sold to us like it's the future, but it's a regurgitation of our collective past, remarketed as the future."
... continue reading