For a long time, using an AI meant opening an app.
ChatGPT. Claude. Gemini.
Then asking a question.
Google now seems to be aiming at something else.
Gemini Spark: an agent that works even once you have walked away
At Google I/O 2026 on 19 May, Google unveiled Gemini Spark.
Spark is described as an agent able to:
- run in the cloud;
- act in the background;
- reach into Google’s products;
- prepare tasks;
- organise information;
- keep working even when the user is no longer in front of their device.
Google plans a beta rollout for some US subscribers.
Spark is not an isolated move. It is the same shift we described when AI tools stopped answering and started acting.
Gemini is outgrowing the chatbot
Google is now pushing Gemini into:
- Gmail;
- Docs;
- Search;
- Android;
- Chrome;
- Workspace;
- YouTube.
The goal is no longer only to answer.
The goal is to become a permanent layer.
More than 900 million users
Google states that Gemini now passes:
900 million users across more than 230 countries.
The company is also pushing:
- Gemini 3.5 Flash;
- Gemini Omni;
- Daily Brief;
- MCP integrations;
- tools able to act across several services.
An announcement is one thing, a standing is another: our rankings track where Gemini actually sits against ChatGPT and Claude and which agents hold up on long tasks.
What actually changes
The main shift is simple.
Yesterday:
you opened an AI.
Tomorrow:
the AI may already be there, everywhere you work.
In your emails.
In your browser.
In your documents.
In your searches.
What to take away
Google no longer presents Gemini as a simple chatbot.
The company is gradually building a permanent AI layer able to assist, organise and carry out tasks across its whole ecosystem.
The subject is no longer only how intelligent the model is.
The subject is becoming how constantly it is there.