Google seems to have altered the way free users access the ‘Thinking’ mode of its Gemini 3 Pro AI model. According to a new report, the firm has edited the rate limit description on its support page to specify that daily limits for users without a Google AI Pro plan "may change frequently". The update does not indicate in which direction the limits will shift – tighten or expand – but suggests dynamic changes pegged to system load.
The revised wording suggests that Google is preparing for fluctuating AI demand that has surged after the launch of its Gemini 3 lineup.
Changes may be limited to the US
It was 9to5 who first reported about Google’s latest update. It is interesting to report that the new wording has appeared on the US variant of Google's support page but it is yet to reach the Indian version- suggesting that the changes may currently apply to only certain regions.
Earlier, free access to Gemini 2.5 Pro and early Gemini 3 testing allowed as many as five complex prompts per day. The report now speculates that Google might further reduce these limits for "Thinking With 3 Pro" as more users test advanced features.
Token-based limits, not prompt-based
Google's limits are not on how many individual questions can be asked, but on token usage that represents how much language the AI processes. A short, simple query takes up fewer tokens, while complex requests, such as long analysis or multi-step reasoning, use significantly more.
This means that a user can send 10 fast prompts in a day and only a few detailed ones before reaching the limit. The wording by Google says that it will keep adjusting the limits based on query complexity and system capacity.
NotebookLM Scales back features too
Meanwhile, Google has confirmed with X that NotebookLM is temporarily rolling back two major features for free users. Even Google AI Pro subscribers will face additional usage limits due to "overwhelming demand" and capacity strains. The company says the features will return once infrastructure is expanded.