The Problem With Auto Model Selection
Gemini free accounts now pick the model for you. Convenient, until the hard question quietly gets the cheap answer.
Google's move to Auto on free Gemini accounts is reasonable for most prompts, but it hides a choice the user used to make. When the system routes a deep-reasoning question to Flash-Lite to save cost, you don't get a warning, you get a flimsier answer.
The fix exists: turn Auto off and pick Flash or Pro yourself. But most people won't, because the whole point of free tier is not thinking about it. So the model that decides your model becomes the one making a judgment call you never see.
I'm not against routing. I'm against routing without a tell. A one-line 'answered by Flash-Lite' would let me upgrade the moment the topic got serious. Without it, the cheap path is the silent default.
The broader trend is the same everywhere: the assistant decides how hard to think. Useful until it guesses wrong about what's hard, and the user has no lever. Give me the label, even if I ignore it.
Updated October 28, 2026: the routing behavior below still applies to free accounts.
Frequently asked questions:
Q: What changed with Gemini's free tier?
A: Google moved it to automatic model selection on October 9 - most prompts route to Flash-Lite, harder ones to Flash or Pro, with no indication of which you got.
Q: What's the actual downside?
A: When the system routes a deep-reasoning question to Flash-Lite to save cost, you get a flimsier answer with no warning. The model that decides your model makes a judgment call you never see.
Q: Can you turn Auto off?
A: Yes - pick Flash or Pro yourself in settings. Most people won't, which is exactly why a one-line 'answered by Flash-Lite' label would fix the trust gap.