Ask ChatGPT the Same Question Twice and It Names a Different Brand First Three Times in Four, Study Finds
The brand ChatGPT recommends first for a buying question usually changes when the same question is asked again, according to a study by the New York marketing agency BusySeed.
The study compared 172,624 pairs of answers to the same question, collected between April 27 and October 16, 2026, across 172 product and service categories. The brand named first stayed the same in only 25.4% of pairs. In the other 74.6%, ChatGPT put a different brand on top.
How much time passed between the two answers made almost no difference. For answers less than a week apart, the top brand held 23.9% of the time. For answers more than four weeks apart, it held 25.9% of the time.
The effect showed up across the market. In 143 categories with enough data, the share of repeat answers with a different leader ranged from 34.3% to 92.8%, with a median of 77.3%. No single category made up more than 1.7% of the data.
Every comparison matched answers from the same model version, the same web search setting and the same prompt wording, so the changes were not caused by OpenAI updating its model.
A separate 15-week BusySeed study of ChatGPT, Claude and Gemini found the same thing across all three assistants. The company named first changed in 56.4% of week-to-week comparisons, and only 85 of 480 tracked question series kept the same leader for the whole period.
“If you check where you rank in ChatGPT on a Monday, you’ve learned something about Monday,” said Omar Jenblat, BusySeed’s founder and chief executive. “What holds up is how often you show up across many asks, not where you placed once.”
The finding matters for a growing market of tools that report a single AI ranking for a company. The research suggests that number may describe one answer rather than a lasting position.
BusySeed, founded in 2013 and based in New York City, provides search, advertising and lead generation services. It publishes its studies, including the underlying data, on its website.
The study used OpenAI’s developer interface rather than the logged-in ChatGPT app, which can personalize answers. It also cannot separate randomness inside the model from real changes in the web pages the model reads. Methodology and data: https://busyseed.com/research/ask-chatgpt-the-same-question-twice.
The post Ask ChatGPT the Same Question Twice and It Names a Different Brand First Three Times in Four, Study Finds appeared first on BNO News.
Ella Rae Greene, Editor In Chief
