The Maze: About 37% of US online shoppers in a 2026 ECDB survey found AI shopping assistants annoying. These are tools on a retailer’s site that answer questions or guide shoppers to products. They had the lowest share of shoppers rating them annoying among the six features, yet only about one in four shoppers called them helpful. For a retailer, the design choice is when to offer help and what job the assistant can do once the shopper opens it.
The details
🛒 AI assistants led the comparison on helpfulness. In ECDB’s survey of 3,042 US shoppers, about 25% rated assistants helpful, 38% were indifferent and 37% found them annoying. Discount campaigns and stock availability displays drew roughly 66% and 64% annoyance, respectively. Participants had bought online in the previous 24 months. They rated features in a survey, so the results capture sentiment about a feature rather than the effect of adding it to a store.
👥 The balance changed across age groups. In the group ECDB labeled 18–45, 44% found assistants annoying and 20% called them helpful. In the group labeled 45+, both shares were 30%. ECDB’s published labels overlap at age 45, and the available material leaves the assignment of 45-year-olds unclear.
📊 The finer age groups put the crossover in context. Subtract the share calling assistants annoying from the share calling them helpful. On that measure, the gap was about −34 points at ages 25–34, near zero at 55–64, and +16 above age 75. The 55–64 result is close enough to even that it gives retailers little reason to build around a particular birthday.
Our take
A retailer’s first bet is where the shopper encounters the assistant. Offer it where it can answer a specific question, then check how that placement affects the rest of the visit.
💬 Give it a concrete job. AI assistants had the lowest share of shoppers rating them annoying among the six features, but the overall split still favors annoyance over helpfulness. Make the assistant easy to find when a shopper asks for help. A product-page prompt such as “Will this fit my model?” gives the tool a clear job. Its answer needs to name the compatible variant and offer a path to a person if the catalog cannot confirm a match.
🧪 Measure what happens to the whole visit. 38% were indifferent to AI assistants. Count all shoppers who saw the new entry point in the test. Compare a quiet entry point and a more prominent placement against the existing page. Track how often shoppers get a useful answer and whether people who skip the tool can still finish their original task. Survey ratings alone cannot show which placement works.
🧭 Use behavior to decide when to offer help. The 45–54 group was about 18 points below zero while 55–64 was near zero. These are group estimates. Separately test a proactive offer of help when shoppers get stuck. A shopper who has filtered repeatedly without finding a compatible item gives the retailer a concrete signal that help could be useful, whatever their age.
Sources: ECDB’s 2026 US online shopper survey
