ChatGPT routes web searches through four separate pipelines, Konitzny says
The standard search_query sits alongside image, business and product fan-outs, each running on its own backend.
David Konitzny, writing on LinkedIn on 24 September 2026, describes four fan-out types that ChatGPT can trigger during a web search.
The default is search_query, the core web search used whenever browsing is active. Beyond that sit three specialised fan-outs. image_query fires when the system decides images belong in the answer. business_query powers local and map results for places like hotels and restaurants. product_query handles dedicated product searches that surface items directly in the ChatGPT interface.
Each runs on a different search engine pipeline behind the scenes, Konitzny writes. Improving one does not improve the others.
We think the practical point is that a single "ChatGPT visibility" score hides more than it shows. A brand can be cited well in search_query and absent from product_query. The pipelines are separate, so the diagnoses are separate too.
Before acting on this, check which fan-out your queries actually trigger. Run the prompts your buyers would use. Watch what ChatGPT shows: a map module, a product card, an image strip, or plain text with links. Each surface points to a different pipeline, and each needs its own read.
Konitzny's post names the four fan-outs but does not publish ranking tests behind each one. Treat the pipeline split as the finding. Treat any single tactic for "winning" one of them as unverified until you measure it on your own queries.