Google released Gemini 2.5 “Computer Use”, a version of their AI model that can interact with websites by opening pages, submitting forms, clicking buttons, and generally surfing the web like you would.
It operates in a browser sandbox (without full system access), with 13 predefined core activities (open, type, drag, scroll, etc.). Developers can test the new Gemini model using Google AI Studio or Vertex AI.
What this means for creators
Tools can automate operations such as autofilling forms, troubleshooting online flows, and scraping structured data, with fewer APIs.
Gemini could help you automate elements of your process, such as posting content, managing web dashboards, and performing repetitive web tasks.
But there is risk. If a model misclicks or submits incorrect information, you bear the expense of the error. Always maintain oversight.
Also read: Fabulate launches SparQ 2.0 to AI‑power influencer discovery
What this mean for entrepreneurs
This shifts one aspect of the infrastructure challenge: instead of developing custom scraping or UI integrations, startups may rely on “agentic UI models” that manage web functionality.
Tool suppliers must think more about safety, sandboxing, and permissions. Include guardrails such as rate limitations, confirmation signals, and rollback features.
Monetisation options include charging for “web action credits,” task bundles, and managed agent services. These models become actual revenue streams.
