Seer: ChatGPT names fewer than 1 in 10 brands before its first search, in 20 prompts
What this answers
- Does ChatGPT name brands before it searches?
- Rarely, in Seer Interactive's test. Across 20 prompts run through OpenAI's API on November 4, 2026, fewer than 1 in 10 brands were named in the model's reasoning before its first search, and 59% appeared right after the first round of broad searches.
- Does ChatGPT search brand websites directly?
- Mostly in later rounds, per Seer's 20-prompt test: 13% of 144 first-round queries used the site: operator to target one website, against 75% of 223 queries in later rounds.
Seer Interactive ran 20 prompts through OpenAI’s API on November 4, 2026, using a model it calls GPT-6 Luna at medium reasoning, and read every reasoning summary. Fewer than 1 in 10 brands were named before the first search, and 59% appeared right after the first round of broad searches, per Seer, which does not give the number of brands behind those shares. Of 144 first-round queries, 13% used the site: operator (limiting a search to one website); of 223 in later rounds, 75% did.
Seer calls the API “directional” for what users see in ChatGPT, not 1:1. It says 38% of the model’s final reasoning notes gave a reason to drop an option. Its five reasons for brand exclusion are Seer’s reading of those summaries from 20 prompts, not a measured ranking.
Why it matters: If it holds beyond 20 prompts, brands may get onto ChatGPT's shortlist through what broad searches retrieve, with their own sites mattering mainly at the later checking stage, though this is a small vendor-run API sample.
Posted to the wire November 18, 2026. Edited by Joe Balewski.