Snapchat taps Google Gemini to supercharge My AI

Snapchat's in-app chatbot is getting a brain upgrade, and it is looking way past basic text.

Snap announced an expanded partnership with Google Cloud to integrate Google's Gemini models into its My AI chatbot. The upgrade brings Gemini's multimodal capabilities directly to users, allowing the assistant to process and understand text, audio, images, and video.

Moving beyond OpenAI to visual search

When Snap originally launched My AI in February 2023, the bot ran on OpenAI's ChatGPT. Now, the company is leaning on Google's generative models to power camera-first features.

At the annual Snap Partner Summit last week, Snap revealed new Lens-like capabilities that run entirely on Gemini. Users can now snap a photo of a street sign in a foreign country to get an instant translation, or take a picture of a restaurant menu and ask the bot to pick the healthiest dish.

According to Google, integrating Gemini into My AI pushed chatbot engagement in the United States up by more than 2.5 times.

The deal deepens a relationship between Snap and Google Cloud that spans over a decade.

However, the rollout comes with familiar baggage. My AI has faced scrutiny since day one over user safety risks, especially regarding younger audiences and children. Giving the bot eyes and ears makes it smarter, but it also raises the stakes on keeping guardrails intact.

Why it matters

For everyday Snapchatters, My AI is shifting from an annoying text buddy pinned to the top of their feed into a legitimately useful camera tool. Snapping a menu for diet advice or translating a foreign sign in real time gives people a practical reason to actually open the chat.

We think tapping Google instead of doubling down on OpenAI was a sharp move by Snap. Snapchat is fundamentally a camera company, and Google's multimodal strengths fit visual, real-world utility far better than standard text generation.

This also shows that consumer apps will not stay loyal to a single AI vendor. Moving forward, platforms will mix and match models based on raw capability rather than brand loyalty.

The downside is content moderation. Multimodal AI opens up far trickier safety vectors, and because Snapchat caters heavily to teens, Snap will need to prove its guardrails can handle images and video without slipping up.