Back to all
At a glance:
- Shoppers who get a clear explanation of why an AI recommended a product are nearly 2x more likely to buy without verifying anywhere else.
- Only 32% of shoppers named accuracy as the top trust-builder in AI recommendations. 49% chose a clear explanation of why a product was selected.
- 47% of 28-to-43-year-olds say AI makes them faster decision-makers. Shoppers “very familiar” with AI tools are 3x more likely to purchase without verification.
- When product data is wrong or incomplete, 80% of shoppers stay in the AI channel and ask again.
A shopper asks ChatGPT for noise-canceling headphones for an open office, under $300. One result explains why it fits: isolates low-frequency hum, weighs less than comparable models, includes a transparency mode for conversations. Another lists a name, a rating, and a price.
In a Rithum and Retail Dive survey of 1,046 U.S. and U.K. online shoppers, 49% named a clear explanation of why a product was chosen as the top trust-builder in an AI recommendation. Always-accurate information came in at 32%. And shoppers who get that explanation are nearly twice as likely to buy without checking anywhere else.
Why shoppers value AI explanation over accuracy in product recommendations
Shoppers expect AI to get the basics right. 67% named price as the top detail AI needs to be accurate on, followed by reviews and availability. But when asked what would most increase their trust, they reached past accuracy. 49% chose a clear explanation of why a product was selected. Always-accurate information came in at 32%.
Any ecommerce team has seen this on a product detail page. Accurate price and clean specs keep a listing live. Rich attributes are what make it sell. The same applies to AI. An LLM builds its explanation from whatever product data it can find. If your listing includes driver size, noise cancellation type, and a note about comfort for all-day wear, the AI has something specific to say. If it doesn’t, the AI defaults to price.
A jacket listed with fabric composition, weight, care instructions, and a note that it runs slim through the shoulders gives AI something to work with. A jacket listed as “men’s jacket, blue, available in S-XL” gives AI a price to compare.
Newer brands with complete, attribute-rich product data already use this to their advantage, earning more persuasive recommendations than established names running on thin listings. When product data is wrong or incomplete, 80% of shoppers stay in the AI channel and ask again. The next answer is built on whatever data is available at that point.
AI-powered shoppers buy faster and verify less
47% of 28-to-43-year-olds say AI makes them faster decision-makers, compared to 21% of shoppers 60 and older. Shoppers who are “very familiar” with AI tools are 3x more likely to purchase without verification.
For these shoppers, the explanation in the recommendation has to do the work that a product page, a review site, or a friend’s opinion used to handle. When the explanation falls short, the shopper moves to the next option in the response. There is no second visit, no follow-up search. The sale goes to whichever product explained itself best.
How to optimize product content for AI recommendations
Product content built for explanation earns stronger AI recommendations than content built only for visibility.
- Enrich product attributes beyond the minimum required fields. Include use cases, compatibility notes, and sizing context.
- Keep pricing and availability current across every channel where AI pulls data.
- Monitor where shoppers verify AI recommendations and ensure your data holds up in those places too. Shoppers turn to search engines first at 28%, followed by online reviews at 19% and friends and family at 17%.
- Test your own visibility: ask an LLM about your product category and evaluate whether your products appear with a clear, specific reason attached.
- Prioritize data hygiene: validate and standardize titles, attributes, categories, and inventory/pricing sync so AI doesn’t amplify broken inputs across channels.
Prioritize data hygiene: validate and standardize titles, attributes, categories, and inventory/pricing sync. AI can’t fix bad data. It can only move faster with whatever you give it, and when the inputs are off, that speed works against you.
For a full breakdown of the data, download The New Discovery Engine report.
Talk to our team
