According to CNET, a new plus button has appeared for some users in Google Search, allowing the upload of images and documents directly for analysis by Google’s Gemini AI. This feature, part of Google AI Mode, lets users ask questions about the uploaded content, like identifying parts in a model kit manual or components in a guitar photo. The company has not made an official announcement, suggesting this is a limited experiment. This enhancement arrives as Google makes a full pivot to AI amidst fierce competition from OpenAI, which recently entered a “code red” state after the release of Google’s Gemini 3 model. In response, OpenAI launched its GPT-5.2 model to stay competitive. Google leverages its scale to offer more features per dollar, including massive Google Drive storage and context windows for Gemini users.
The Quiet Feature War
Here’s the thing: Google isn’t just launching a new AI chatbot. It’s weaving AI directly into the fabric of the internet’s front door—Search. That’s a huge advantage. OpenAI’s ChatGPT is a destination. Google’s AI is becoming part of the journey. You’re not going to a separate app; the capability is right there where you were already typing. This is classic Google ecosystem leverage. They can just… add a button. And for users, it feels seamless. Need to understand that confusing diagram? Upload and ask. It’s practical utility, not just a demo.
The Infrastructure Advantage
But the real story might be under the hood. The article points out that Gemini 3 was trained on Google’s own Tensor Processing Units (TPUs). That’s a big deal. It means Google isn’t waiting in line for expensive Nvidia H100s like seemingly everyone else. They control their own stack from silicon to software. That gives them cost control, scheduling certainty, and the ability to optimize in ways others can’t. When you combine that with their vast data centers and the sheer integration with products like Drive, you see a moat that’s incredibly hard to cross. It’s not just about having a smart model; it’s about having a smart, efficient, and deeply embedded model.
So, Has Google Won?
Some observers say Google has pulled ahead. I think it’s more accurate to say they’ve leveraged their home-field advantage. OpenAI sparked the race with a stunning product, but Google is playing a different game—one of attrition and integration. They’re not just competing on model benchmarks; they’re competing on making AI disappear into everything you already do. The plus button is a perfect symbol of that. It’s small, unassuming, but it unlocks a powerful capability right where you are. The question is whether users will adopt these woven-in features as eagerly as they flocked to the novelty of ChatGPT. The battle isn’t just about whose AI is smarter; it’s about whose AI is more useful in the daily grind. And Google is betting everything on that latter point.
