Gemini Spark's always-on agent expands to more countries, languages
By Steven Van ·
Google AI Ultra subscribers outside the original rollout regions may now see Spark in the Gemini app.
Gemini Spark is going global. Google announced that Spark, the personal AI agent that works in the background 24/7 to get things done under your direction, is now rolling out to Google AI Ultra subscribers in more countries and languages, moving it from a US-flavored preview toward a worldwide default for Google's highest tier.
What Spark actually is
Spark is Google's entry in the always-on agent category: not a chat you open, but a worker that keeps going after you close the tab. You give it direction, research this, watch that, handle the recurring thing, and it runs in the background around the clock, surfacing results as it produces them. Recent updates added real-time tracking, and the agent has been the centerpiece of the AI Ultra subscription since its debut.
Languages are the harder half
The expansion's language dimension matters more than the geography. A background agent is only trustworthy if it understands instructions with native fluency, an ambiguity that's a minor annoyance in chat becomes a wrong action repeated on your behalf in an agent. Shipping Spark in more languages is Google asserting that confidence, and it opens the always-on agent experience to creators and professionals working outside English for the first time.
The ambient agent race
Spark's expansion lands in the same month Higgsfield made its Supercomputer LLM free, OpenAI rebuilt its desktop app around background-capable Work and Codex modes, and Google's own Gemini API added scheduled triggers for developer-built agents. The consumer, prosumer, and developer versions of the same idea are all shipping at once: AI that runs while you don't. Spark is the most consumer-shaped of them, bundled with a subscription rather than built from parts.
Check your access
If you're an AI Ultra subscriber outside the original rollout regions, Spark may now be waiting in your Gemini app. The learning curve is less about features than delegation: the users who get value from Spark are the ones who hand it recurring work, not one-off questions.
Sources: Gemini on X, Google Gemini.