Third-party AI tools are deciding which products end up in grocery carts and how much shoppers spend. Few retailers can see it happening.
An estimated 38.9% of US consumers are already using AI to shop for groceries, and among that group, 65% engaged it at first touch, before they had even decided what to buy.
NPS Prism, in partnership with Dynata, surveyed nearly 2,000 US grocery shoppers and found that for many, the grocery list is being built in a chat window, before your site, your store, or your promotions enter the picture. Two important implications stand out for grocery leaders:
They are the same problem viewed from different angles: an influential channel that is both untracked and working against margin. Managing it will require an understanding of how and why your customers use AI to shop.
In the AI-assisted grocery shopping journey, engagement peaks during research and comparison, then falls away sharply as the transaction nears. The sharpest single-stage drop, 14 percentage points, comes between comparing options and deciding where to buy, indicating that retailer choice is still largely habitual. Another 12-point drop follows between building the final shopping list and completing the purchase.
What makes this even harder to manage: 52% of AI-using grocery shoppers relied exclusively on third-party tools like ChatGPT and Gemini. The retailer only captures the sale, but the channel doing the most commercial work is one you do not operate, do not have visibility into, and cannot directly influence.
This influence and attribution gap matters even more because AI is not a neutral planning tool. During research and comparison, it is running a value comparison, and shoppers are acting on it.
Among AI users, 32% report they would have spent more without it. Only 9% say they would have spent less. For every shopper AI nudged toward spending more, it moved more than three the other way.
The use-case data shows why. Meal planning leads at 46%, followed closely by substitutions and alternatives at 44%. That second use case is where the commercial risk sits. When a shopper asks AI for a cheaper alternative or a substitute for something out of stock, that is a brand-switch decision. It used to belong to shelf placement and promotions. It now belongs to ChatGPT.
The influence gap, measurement blind spots, and the value steer all point toward the same set of responses.
Build measurement infrastructure that connects pre-shopping AI activity to basket outcomes. Retailers who can’t tie AI influence to revenue will eventually defund the programs that could help them manage it.
Treat AI presence management as a marketing discipline. Structured product data, AI-readable pricing, and recipe content that AI can cite are no longer optional investments.
Audit how your private label and value ranges appear in AI outputs. Substitution is AI's second-most-used grocery function and a silent share-taker. Run periodic tests of how your own-brand alternatives appear in AI substitution outputs.
For warehouse clubs and membership retailers: make member pricing AI-readable now. Per-unit member value is the single strongest argument your format can make in a value comparison, and general-purpose AI can’t see it today. It sits behind login walls and dynamic pricing engines. Close that gap before your competitors do, and AI's value steer turns from a structural threat into a routing advantage.
Retailers who invest in AI presence and measurement will be positioned to gain. Those who don’t will lose retailer choice to AI before they have built any way to respond.
NPS Prism tracks how your brand shows up across the moments AI now influences, giving retailers visibility into how you’re stacking up to competitors as AI’s impact on grocery carts grow. Request a free competitive intelligence report to start strategically managing this channel instead of guessing at where to invest.