A new Google Research study analysing 15 million anonymised interactions with its Gemini AI across the Gemini App, AI Mode, and the Gemini API has found little evidence that AI is driving mass automation or displacement of white-collar work, despite widespread predictions to the contrary. The research, dubbed the "AI & Economy ATLAS", concludes that current use of the tool "remains shallow and overwhelmingly collaborative in nature", with AI functioning mainly as a complement to human effort rather than a replacement for it, a finding relevant to ongoing debates about AI's impact on jobs and the wider economy.

Researchers classified interactions using official US occupational and task databases and found that only 21 percent of all tracked work tasks met the threshold to count as meaningfully assisted by Gemini. For 29 percent of occupations, not a single task reached this threshold, while a further 30 percent saw fewer than a quarter of their tasks significantly touched by AI. Only 3 percent of occupations, including software quality assurance testers, human resources specialists and document management specialists, had at least three-quarters of their tasks regularly supported by Gemini. Heaviest use came from fields such as finance, computing, and arts and entertainment, while sales, transport and food service workers were notably underrepresented.