2025 Year in Review: Gen AI Trends & Transformations

2024 was the year of chat.

2025 was the year of image and video models.

Where is 2026 heading?

By Andreea Bucescu

When new features and models are released every week, it’s hard to keep up with the latest AI advancements and really feel like you can understand them. But looking back on a whole year is pretty productive for understanding patterns. Let’s take a look back at what happened in the previous year, models that broke through the noise, and what could happen next.

The Year of Chat

ChatGPT first made its debut to the public on November 30, 2022. Candidly, I only started using it in early 2023, and by the end of the year, I was consulting with a global marketing agency headquartered in Buenos Aires on how AI integrates into their departments and where to invest next.

While leaders began to explore AI in 2023, the technology was still too early to have a significant impact. AI courses were popping up, and people started selling their “Top 50 Prompts for X” – a symptom of a small group trying to dumb down concepts to the public, who haven’t yet caught up.

In 2024, however, ChatGPT became a household name. Enterprises were buying AI subscriptions, universities started implementing AI policies, and, most importantly, you could look up “ChatGPT” on TikTok or YouTube and be faced with a ton of content. People were finally talking about it everywhere.

The Year of Video

It is no surprise that, in the visual world we live in, we needed something more than text.

While image generators were available in GPT early on, the early versions felt AI-generated. People looked weird, overpolished, and no text or small pattern was accurate. The public had speculated about Google’s image model for a long time before it was finally released. And when they did release it, Nano Banana broke the internet.

Waves of AI-generated photography and videography took off as they became more available and accessible (read: affordable) to the public. Businesses like Fal offered an all-in-one marketplace where users could easily switch between models and experiment with them. And this sparked several trends across social media. Queuing up the Visual Honour Roll.

The Visual Honor Roll

The Action Figure Trend

It started off as individuals showing off their personalities and hobbies by simplifying them into an action figure: a photographer came with a camera, a cook came with a pan and a recipe book, and an entrepreneur came with a phone that was always on. It was so easy to recreate with a simple prompt that even brands like the British Royal Mail recreated it – and it was so well received by the public because it was a playful experimentation with AI.

The Cinematic Video Trend

This took over X, and made appearances on other short-form platforms, too. Whether it was a cinematic video of a girl running through the mountains in Vietnam, or a cyberpunk ninja jumping between skyscrapers in NYC, it was fascinating to see consistent environments and human-like movements in action for the first time, without looking unnatural.

Sora: Absurd Videos Engineered for Virality

Not too long ago, Sora was released by OpenAI. It was groundbreaking because (1) the technology was genius, and (2) everything about the content was highly shareable. 

The technology was genius because you no longer needed to write a long, detailed prompt, as video models like VEO expected you to. Rather, you could simply write “horse blown away by a tornado, other farm animals scared” and you’d get an incredibly realistic video, that is in a 9:16 ratio (perfect for short-form content!), with dialogue or plot twists that are excellent for keeping a scroller engaged every second.

The biggest key to going viral, however, was shareability. And Sora engineered this perfectly: first, the app was invite-only. You had to know someone who had already joined to receive one of their five codes. Second, the content shared around wasn’t a ninja running – it was bizarre scenarios like the horse blown away by the tornado, or Sam Altman sneaking up on you in the locker room, telling you to try Sora. You could add yourself and your friends to the videos, and make it feel a lot more personal. Third, you could remix other people’s work with one tap (...let’s just say I saw plenty of animals being blown away by a tornado…). This allowed for quick compounding of creativity and the emergence of ridiculously funny content in seconds.

What’s Next for 2026?

The Sora experience is a great segue into my AI prediction for 2026. Next year, we can expect to see a significant increase in collaboration in AI. Integrations from text-to-photo-to-video (-to-game?) and beyond. 

I’d expect apps to emerge where users know that everything is AI-generated, and yet they still love to continue playing with them. 

Because, after all, creation is play.

---

Andi Bucescu is a marketer and thought leader exploring how creativity and technology intersect. She works at the edge of AI and storytelling, building digital experiences that make innovation feel human, approachable, and fun.

At Rosebud AI, Andi leads marketing for a next-gen creative platform, crafting campaigns that position generative technology as playful, creative, and accessible to everyone. Her perspective is informed by undergraduate research in consumer psychology and social media communication, as well as hands-on experience in crafting campaigns that make complex technology accessible to everyday audiences.

Before joining Rosebud, Andi built her career with EdTech and FinTech startups across the U.S. and Europe, while working on meaningful projects in South and East Asia, Latin America, and Eastern Europe, focused on using marketing for social impact. She graduated from Minerva University, one of the world’s most innovative universities, with a degree in Business and Brand Management.