#gpt-5.3

ChatGPT's Citation Pool Shrank by 21%. Your Baseline Data Is Now Wrong.

A 14-week study of 27,000 ChatGPT responses found that the shift to GPT-5.3 Instant as the default experience cut cited domains from 19 to 15 per response. Any GEO baseline captured before March 2026 reflects a citation pool that no longer exists. With GPT-5.5 detected in live API testing, a second compression is likely days away.