Back to all
TL;DR
- Product journeys, from discovery to decision, are shifting to AI. Among AI-active shoppers, 90%+ use LLMs to research products and compare options, and 53% use them to decide where to buy.
- AI is diminishing buyer loyalty. 19% say they now buy from brands or products they had not heard about before, and 13% say they are more likely to switch retailers or products after using an LLM.
- Brand and retailer sites have less time to influence the decision. 32% of shoppers spend less time browsing other sites after using LLM tools, and only 5% verify AI shopping information on a retailer or brand site.
- Product information now plays a bigger role in trust. 49% say a clear explanation would do the most to increase trust in an AI recommendation, 67% say price is the most important detail to get right, and 58% say trust in the brand drops when an LLM provides incorrect product information.
Rithum’s new report, The new discovery engine: How consumers are using AI to find, trust, and choose brands, and what’s at risk for those they never see, has a clear message for retailers and brands: the shopping journey is no longer confined to shelves, search results, category pages, or product detail pages.
Based on a survey of 1,046 online shoppers in the U.S. and U.K., the report shows how LLMs have become the entire shopping journey and a one-stop shop where products get researched, compared, narrowed down, and chosen. Among AI-active shoppers, more than 90% use LLMs to research product information and compare options, while 53% use them to decide where to buy. By the time a shopper lands on a product page, the filtering may already be over.
That shift to AI is creating room for brands that were not previously in the mix. In the survey, 19% of shoppers say they now buy from brands or products they had not heard about before. Established brands are facing a tougher version of the same market. Recognition still helps, but it has less force when AI is doing more of the sorting, ranking, and explaining before the click.
The shortlist is forming earlier
AI shopping adoption is the starting point for 80% of shoppers ages 18-to-27 and 80% among shoppers ages 28-to-43. Among households earning $100,000 to $150,000, it reached 84% adoption. These are commercially important shoppers, and they are already weaving LLMs into their purchase journey.
The filtering that used to happen across tabs, retailer sites, and review pages is now happening inside LLMs. A product that appears high in the response moves ahead. One that doesn’t can disappear before the shopper has seriously considered it.
ChatGPT’s product comparison feature adds to that shift. Shoppers can compare products side by side inside the chat, with price, features, reviews, and other details presented in one place instead of scattered across multiple retailer tabs.
More than half of shoppers already trust AI tools as much as brand websites, and among high-income households, confidence in AI accuracy climbs as high as 80%. That trust gives LLM recommendations real weight early in the decision process, according to the report.
Retailers now have more riding on how products are represented on their page, even before a shopper lands on the site. Brands face the same pressure, but with fewer natural intervention points. A retailer may still appear as the place to buy, even if a brand is filtered out earlier. If a brand’s product information is incomplete, inconsistent, or hard for AI to explain, it can be dropped from consideration before its own product page or brand story has a chance to influence the decision.
New brands are finding room where familiar brands once had an easier ride
You can see the LLM effect far beyond just initial research, with ripples into what shoppers buy. Nineteen percent of shoppers say they are more likely to buy from brands or products they had not heard about before if an LLM suggests it. Another 13% say they are more likely to switch retailers or products after using an LLM. Together, those numbers create a shopping environment where familiar brands have less room to rely on recognition alone.
That creates an opening for challenger brands. A newer brand does not need years of broad recognition to get in front of a shopper. It needs usable, consistentproduct information and enough context for AI to present it clearly and convincingly.
Established brands have less room to lean on familiarity alone. Customer loyalty still helps, but it no longer ensures that they go to your site first. Nearly half of shoppers say a clear explanation of why a product or brand was chosen would do the most to increase trust in an LLM recommendation. What carries weight here in an LLM recommendation is not name recognition but whether the recommendation feels specific, informed, and ready to act on.
Brand-owned sites get fewer chances to influence the outcome
The shopping journey used to leave more room for second thoughts. A shopper could open a few tabs, compare prices, read reviews, leave, come back, then change course. LLMs have shortened that process.
In the survey, 32% of consumers say they spend less time browsing other sites when using LLM tools to shop. Another 36% say they make faster decisions, while 34% say they feel more confident about their purchases.
These three stats don’t live in a vacuum. They indicate a continual trust-building experience for the shopper: they’re saving time, they’re finding what they need faster, and they feel better about their purchases. Why would they leave that experience to go back to a retailers website?
The same pattern appears in how people verify what they see. Shoppers who double-check an LLM recommendation rarely begin with looking for confirmation on a brand or retailer site. Twenty-eight percent turn to search engines (which is likely also relying on AI tools), 19% specifically look for online reviews, 17% ask friends and family, and only 5% go to a retailer or brand website. A beautiful, brand-forward website won’t convince them to buy your product. They won’t even see it. But a PDP with in-depth specifications, GEO-optimized keywords, and highly relevant descriptions will impact consumers’ decisions, even if they don’t see the page.
The recommendation is only as strong as the product story behind it
Ask an LLM why it chooses one product over another, and it has to build that answer from the product facts it can find: materials, dimensions, compatibility, intended use, etc. The recommendation that an LLM givesis assembled from those pieces in real time.
For brands, that raises the standard for product content. Copy, attributes, use cases, and supporting details are no longer sitting off to the side as content maintenance. They are becoming part of the recommendation itself. When the product story is thin, generic, or inconsistent, the answer reads that way too.
Retailers feel the same pressure across the assortment. Pricing, inventory, attribute completeness, and feed quality all shape how products are represented before a shopper ever reaches the site. Anyone who has spent time inside a catalog has seen how quickly that can start to fray. A bad price, a missing dimension, or stale availability can make a solid product look less reliable than it is.
The harder question is whether the product story still holds together everywhere that LLMs are pulling from. This includes product content, syndication, pricing, availability, and the systems that keep those details aligned. It also includes sources brands and retailers cannot fully control, such as reviews, forums, and social discussion. When those external signals surface alongside structured product data, inconsistencies become more visible. That makes it even more important for the information you do control to be accurate, complete, and easy for AI to explain. LLMs only give recommendations they can trust, based on the information that holds it together.
Trust is moving closer to the data itself
The survey leaves little ambiguity on price. In an AI shopping recommendation, 67% say it is the most important detail to get right. Reviews, availability, where to buy, and technical specifications all come after it in the list of prioritization
That order will feel familiar to anyone who has watched shoppers abandon a cart over a mismatch or lose confidence over a number that does not look right. A wrong price or stale detail does not stay in the background. It becomes part of the recommendation, which means it becomes part of the shopper’s impression of your brand.
The report puts numbers behind that. When an LLM provides incorrect product information, 58% say trust in the product or brand decreases, and 16% say they leave the purchase altogether.
At that point, the issue is no longer confined to data quality. The recommendation may come from the model, but shoppers are not spending time sorting out where the error began. They decide whether the information feels reliable, and the brand lives with the result.
The next phase is close enough to shape decisions now
The report also looks ahead to a shopping flow where the model takes on more of the decision itself. More than 25% of AI-active shoppers say they are already very likely to hand purchasing decisions to AI, and another 39% say they are somewhat likely to consider doing this, if and when it’s available. Among the most AI-active shoppers, 65% say they are very or somewhat likely to use an AI agent that would buy for them.
What AI sees already shapes what shoppers buy. Thin product content, stale pricing, patchy attributes, and inconsistent availability all weaken the recommendation before the shopper has done anything beyond type in a prompt.
The priorities are clear. Keep your product story consistent. Keep pricing accurate. Keep availability current. Make products easier to compare, easier to explain, and less likely to be misread. New brands already have more room to enter the conversation. Established brands have less room for weak information, stale details, or missing context.
For more details on the survey and a full breakdown of the results, download the report here.
Methodology
Rithum’s 2026 report is based on a survey of 1,046 online shoppers in the U.S. and U.K. Some questions look at behavior in the last 3 months, some category questions use the last 6 months, and some trust and behavior questions are broader and are not tied to a single recall window.
