A study result, not a permanent ranking
A study reported by Search Engine Land, conducted by Semrush and Kevin Indig across 100 queries in 20 buying journeys, found 25.6 percent overlap between brands cited in the lightest and deepest reasoning modes. The result shows substantial variation in that sample, but it does not describe every ChatGPT answer or every industry.
Responses can also change with the model, prompt wording, language and date of the test.
Measure several scenarios
An ecommerce brand should not treat one answer as a stable position. A better approach is to define a limited set of representative questions and repeat them across several modes at regular intervals.
Precise product pages and verifiable evidence reduce ambiguity but cannot guarantee a citation. The useful signal is the trend across repeated tests, complemented by referral traffic and conversions that can actually be measured.