ChatGPT Thinking mode changes which brands get cited

ChatGPT cites largely different sources when its Thinking mode is switched on, according to a Semrush study published on 30 June 2026 with growth advisor Kevin Indig. Only 25.6 percent of cited domains overlapped between minimal and high reasoning across 200 test responses. For brands, that means AI visibility is now two separate lists, and ranking in one does not secure a place in the other.

What happened

Semrush tested 100 prompts twice, once in each mode, across 20 buyer journeys in four sectors: B2B SaaS, finance, consumer tech, and health and lifestyle. According to the study, high reasoning ran 1,130 web searches against 245 for minimal reasoning, citation rates rose from 50 to 68 percent of responses, and the average number of sources per response nearly doubled, from 2.6 to 4.5.

The mix of source types changed as much as the volume. Semrush reports that Reddit citations fell from 15 to 7 percent when reasoning was turned on, while government and academic sources climbed from 1.9 to 8.8 percent and official documentation also gained share.

Why it matters

Thinking mode behaves like a different search engine, not a more careful version of the same one. When ChatGPT reasons, it searches more, cites more, and reaches past community chatter toward authoritative, structured sources. A brand that shows up in quick conversational answers can be absent from the deeper research answers that shape considered purchases, and the reverse is equally true.

“Only 25.6% of cited domains overlap between minimal and high reasoning.”

Semrush, ChatGPT reasoning and AI visibility study

What this means for multi-location brands

A central team cannot treat AI visibility as one number any more. If reasoning modes reward authoritative and first-party sources, the priority for a chain with hundreds of locations is making its own data the easiest thing for an AI to trust: consistent business listings across every platform, accurate structured data, and location pages that answer questions directly. The shift away from forum content also cuts both ways: less risk from a stray Reddit thread, more weight on the sources you control.

Measuring it needs the same split view. Track how your locations and pages surface in AI answers with Places AI, and treat ranking in AI search results as work that has to hold up in both a quick answer and a reasoned one.

The bottom line

AI answers are diverging by mode, and brand visibility is diverging with them. Semrush’s numbers suggest the winners under quick answers and the winners under reasoning are mostly different names, so multi-location brands should verify their presence in both before assuming either one is safe.