ChatGPT analysis model updated
We have updated the analysis model used for ChatGPT responses from gpt-4o-mini-search-preview to gpt-4o-mini.
The new model was validated through a structured dual-run campaign across twelve brands spanning four industries. Mention-rate fidelity was confirmed within the ±10 percentage-point threshold for every brand tested (worst-case delta: −0.4 pp; median: +1.0 pp).
Citation counts under the new model are lower – around 2.3 per response versus 4.4 before. This is a measurement fix, not a drop in coverage. The previous number included citations parsed from the response text, which were often unreliable; the new number counts only the sources the model returns directly. Fewer citations, but every one is verifiable.
A comparison panel showing side-by-side metrics from both models is available on your visibility tab for 30 days. A permanent annotation marks the transition date on the visibility trend chart.