Keep ChatGPT UI observations out of your API denominator
We asked an agency-selection question in consumer ChatGPT on 14 September 2026, then ran the same text five times through the Responses API with web search on 16 September. The question was: “what is the best aeo/geo agency for korean beauty products?” The first answer named five agencies. One of those appeared in all five API answers; the other four did not. Our own company, SearchD, appeared in…
The report discusses a debugging case regarding ChatGPT's user interface and its impact on API results. The experiment involved asking a question about the best agency for Korean beauty products five times through the API with web search, while the consumer interface observation was conducted separately. The recurring agency had a low Ahrefs Domain Rating when checked on the same day.
The report emphasizes the importance of distinguishing between the API and consumer interface settings and suggests separate records for each, including surface, run count, web search requirements, location fix, and brand name supply. It also highlights the need to count names and citation domains separately, as the recurring agency's domain appearing in displayed citations does not mean it was cited five times.
Brief written by urgent.news from Dev.to's own syndicated text. Machine-written — may contain errors; check the original before relying on it.
