The Maze: People in a survey published by NP Digital in October 2026 most often picked discovery for TikTok or Instagram and comparison for ChatGPT or other AI assistants. The 983 respondents chose one main role for each platform or platform group when considering a purchase, with non-use as an option. Each percentage is a share of all 983 respondents, including those who did not use that group for purchase decisions. The publisher left the survey dates, geography, recruitment and weighting unspecified.

The details
🔍 Discovery and comparison favored different platforms. 43% chose discovery for TikTok or Instagram, versus 21% for Google, a 22-percentage-point gap. 39% chose comparison for ChatGPT or other AI assistants, versus 19% for Google, a 20-point gap. Each grouped figure covers the named platforms together. The survey asked about each group separately, so these shares describe stated roles across respondents. Tracing a shopper from one channel to the next would require separate evidence.
🧭 Google covered several parts of the decision. 24% chose learning more for Google, close to the 21% for discovery and 19% for comparison. A search visit might therefore start with very different questions. Someone searching a product name may need to understand how it works, while someone searching two model names may be deciding which to buy.
Our take
A retail team can use these roles to choose what each piece of content should help a shopper do, then check whether that matches how its own customers shop.
⚖️ Make product differences easy to verify. The 39% who chose comparison for AI assistants give retailers a reason to test which differences those tools can explain. For two similar products, publish prices, compatibility and trade-offs side by side. Then ask assistants the questions buyers ask and check their answers against the published specs. That test can reveal factual errors. Any effect on sales would need a separate test.
🎬 Give videos one question to answer. 28% chose learning more for YouTube, and 40% reported non-use. 73% reported no purchase-decision use for Reddit or online communities. A fit or installation demo could help people who seek it, even if the group it reaches is small. Test a demo against a question customers already ask, and measure how many relevant shoppers see it separately from how well it answers the question. The survey gives no category breakdown or basis for predicting paid-media returns.
Sources: NP Digital’s October 2026 survey


