What can Gemini 3.5 Flash automate in Search?

Google is rolling out Gemini 3.5 Flash with an emphasis on “agentic” work—tasks that extend beyond simple Q&A into longer-horizon automation. Across its Gemini app and Search’s AI Mode, the model is positioned to handle instruction-following that can involve multiple steps, including coding-related activities.

In Search, this shows up as an AI-powered experience that can support longer queries and more complex goals, rather than limiting users to short prompts. The new Search box is described as letting people ask lengthier questions and also upload photos and videos as part of their requests. That multimodal input matters because it broadens what an “automated search task” can include: the system can use visual context in addition to text.

The agentic framing also suggests the system can take on work that typically requires users to do repeated searches and follow-ups. Instead of only returning links, Google’s AI Mode is designed to perform parts of the search process on the user’s behalf—making it more feasible for someone to describe a goal once and then receive a more complete outcome.

Google’s messaging around Gemini 3.5 Flash highlights two major areas:

• Agentic long-horizon tasks: the ability to handle work that spans multiple steps.
• Coding and automation: described as particularly strong for coding-related requests.

Google also characterizes Gemini 3.5 Flash as its “strongest agentic and coding model yet,” and it says the model can reduce enterprise AI costs dramatically, which is relevant because agent-style features may be more expensive to run than traditional chat.

Taken together, the changes indicate Google is trying to shift Search from a results browser to a workflow executor. The practical impact is that users can increasingly expect the system to take actions—potentially including creating outputs or drafts—rather than just surfacing information for them to assemble manually.