Google Gemini's answers are shaped by raters' work: Here's how

Technology Sep 12, 2025

Google's Gemini chatbot isn't just powered by high-tech code—it also relies on thousands of contract "AI raters" who check and fix its answers.
These workers, hired through firms like GlobalLogic, help keep the chatbot's replies accurate and safe.
But their work often comes with tough conditions and exposure to some pretty disturbing stuff.

US raters earn $16-$21 an hour

US-based raters earn $16-$21 an hour—less than AI researchers but more than some overseas colleagues.
Their tasks can get intense: reviewing responses in tricky areas like medicine or astrophysics, all while juggling tight deadlines and shifting rules.
Many weren't warned about the upsetting content they'd see.

Raters are key to keeping AI helpful

These behind-the-scenes workers are key to making sure AI stays helpful (and not harmful).

Relying on underpaid people raises questions

Google insists these raters only give outside feedback and don't directly shape how Gemini learns.
Even so, relying on underpaid people for such important work raises big questions about fairness and responsibility as AI keeps growing.