Gemini Spark fumbles Chrome control, nails file sorting

Gemini Spark, Google's latest AI agent in beta, jumps into digital task automation with a split skill set. It blitzes through file organization, yet can't touch browser automation or product research. The oddest twist: Spark, built by the team behind Chrome, can't even steer Chrome.

That's straight from the source. When prompted to handle browser tasks, Spark replies, "I cannot directly access or interact with your open Google Chrome window." For a Google-built AI, this is a head-scratcher. ChatGPT Work and Claude already automate browsers. Chrome integration should be Spark's home field. Instead, users hit a wall and walk away with unfinished work.

Speed versus Substance

Gemini Spark moves fast but leaves gaps. In Amazon product research, Spark spits out results in under 30 seconds. ChatGPT Work drags its feet for 10 minutes. But Spark's answers miss the mark. Instead of Amazon listings, it serves up random web results, mislabels them as Amazon Prime, and skips opening browser tabs for comparison. The output misleads, showing Spark leans on Google's search index instead of Amazon's actual site.

Sentiment analysis lands better. Spark pulls together Reddit and Substack reactions to Claude's Opus 5.5 update, surfacing both upgrades and sticking points. But it ignores X and Facebook, even when told to include them. The summary is sharp but incomplete, pointing to either data access limits or missing social integrations.

File Management: Fast, Smart, and Risky

Spark's real muscle shows up in local file sorting. Drop a messy folder of PDFs, and Spark sorts, analyzes, and renames files with speed that matches or beats rivals. It even creates a backup of every file it touches-a safety net missing from ChatGPT Work and Claude. But there's a catch: Spark asks for sweeping permissions, including the right to "share the content of this folder with third parties" before any job starts. For anyone guarding sensitive data, that's a dealbreaker. The lack of task-specific permissions erodes trust and will keep privacy-minded users away.

Still, Spark's workflow is smooth. It renames blandly titled files and sorts them into clear categories. The backup feature is a rare plus, but no one should trust an AI agent alone with critical files. The privacy trade-off overshadows the technical win.

Usage Limits and Google's Relentless Iteration

Gemini Spark doesn't nickel-and-dime users. Four heavy tests burned through just 20% of the five-hour monthly cap on the $20 Google AI Pro plan. That makes Spark a budget-friendly option for those willing to put up with its current flaws.

But Google's AI launches rarely go smoothly. As previous outages show, reliability is still a moving target. Spark's rocky start fits the pattern. Google's history, though, is all about relentless upgrades-Android started rough and now rules the market. The Nano Banana image tool is already top-tier. Spark is likely to close the browser automation gap and tighten up permissions, but there's little sign privacy will ever top the agenda.

Right now, Spark isn't the AI agent to beat. Its file sorting is unmatched, but browser and research features lag behind Claude and ChatGPT Work. The privacy model blocks it from sensitive jobs. Google's pace of improvement could change that, but for now, Spark's strengths and flaws are both on full display.

According to an official Google Cloud announcement, Gemini agent remains in closed preview for a select group of enterprise users. No public launch date or pricing has been set. Google says specialized versions for legal and financial workflows are in early testing, with plans to expand into government, healthcare, and retail.