New research on AI assisted shopping has produced a finding that should reset how brands think about discovery. AI tools are measurably boosting how often products get found, but customer reviews remain the factor that actually closes the sale. according to Modern Retail In other words, artificial intelligence widens the top of the funnel while human opinion still controls the bottom.
What the research actually shows
The pattern across the data is a division of labor between machine and crowd. AI assistants excel at translating vague intent into candidate lists, a shopper describing a problem rather than a product gets useful options within seconds. according to Modern Retail But when the assistant presents three or four plausible choices, the deciding input is overwhelmingly what other buyers wrote. Review volume, recency and specificity correlate with conversion far more strongly than any attribute of the AI summary itself.
Why reviews beat summaries
The reason is credibility asymmetry. Shoppers treat an AI summary as a compressed restatement of marketing claims, while reviews are perceived as unfiltered experience, even when many are incentivized. according to Modern Retail Kroger, Giant Eagle and DoorDash are all building agentic AI assistants for grocery shopping, and each is discovering that the assistant answer is only as persuasive as the review corpus sitting behind it.
The category exceptions
The review effect is not uniform. In commoditized categories with low risk, shoppers lean on the assistant summary and price; in categories where fit, durability or compatibility matter, reviews dominate because they carry the specifics a summary strips out. according to Modern Retail Grocery sits at one extreme and consumer electronics at the other, which is why assistant led grocery ordering has scaled faster than assistant led electronics purchasing.
The new discovery stack
The second engine is easier to neglect because it feels like legacy work. Review programs lack the novelty of AI deployments, yet they are the layer that converts the traffic AI delivers. according to Modern Retail The research finding is blunt on this point: when discovery improves but conversion does not, the missing piece is almost always review quality rather than assistant accuracy.
For merchants, the implication is that discovery now runs on two engines at once. The first is AI visibility, being present and accurately described in assistant answers. The second is review health, a continuous supply of specific, recent, credible feedback. according to Retail Dive Walmart opened a next generation California fulfillment center partly to speed delivery, and faster delivery is itself a review driver, since shipping experience dominates negative feedback in many categories.
Feeding the funnel
Operationally this changes the post purchase moment. Requesting a review is no longer a courtesy email but a supply chain input for discovery. Brands that systematically route satisfied customers into detailed reviews, with prompts about fit, durability or use case, are building the exact corpus that AI assistants cite when shoppers ask for recommendations. according to Modern Retail The same discipline Walmart is applying to training associates on AI applies upstream: the data has to be structured to be useful.
Measuring the two engines
Attribution is the hard part. AI assisted journeys rarely end in a single session, so last click models will undercount the channel entirely. Teams that rely on platform reported conversions alone will conclude AI does nothing, when in reality it is seeding demand that surfaces days later through branded search. according to Modern Retail Building a simple survey based touchpoint capture closes most of that gap without a data science project.
A workable scorecard tracks both layers separately. For AI visibility, monitor how often the brand appears in assistant answers to category questions and whether the description is accurate. For review health, track review velocity, average specificity and the ratio of reviews mentioning the attributes that matter in the category. according to Modern Retail When the two lines move together, discovery investment is working; when AI presence rises but conversion stalls, the review layer is the bottleneck.
