Is AI the Future of Retail Checkouts?

Retailers spent a decade removing humans from checkout lanes. The data that came back forced them to reconsider.

Self-checkout was sold to the industry as a straightforward efficiency play with fewer cashiers, faster throughput, and lower labor costs.

What it produced instead was a theft environment that grew steadily worse and is now generating a regulatory response. Walmart, Dollar General, and others are pulling the machines. But bringing humans back is expensive and, as one retailer is demonstrating, not the only option.

The more forward-looking response is AI, deployed not to replace the cashier, but to do what the cashier was supposed to prevent.

The Theft Data That Changed the Calculation

A December 2025 LendingTree survey of 2,050 US consumers put specific numbers to what retailers had been observing on the floor.

27% of self-checkout users admitted to intentionally stealing at least one item without scanning it, up from 15% in 2023, a 12-point increase in two years.

Another 36% said they had accidentally left with an unscanned item, and of those, 61% kept it rather than returning it. In total, 69% of self-checkout users agreed the machines make stealing easier, and 55% of those who had stolen said they would do it again.

Food price inflation adds context to the motivation data. Food prices were 2.7% higher in March 2026 than a year earlier, and the USDA projects prices will rise 2.9% across 2026.

LendingTree's chief consumer finance analyst framed the theft increase as driven by economic pressure as much as opportunity.

The 12-point increase in admitted theft between 2023 and 2025 tracks directly with the period when self-checkout became the default lane at most major retailers.

The machines did not create dishonest shoppers. They created an environment where the likelihood of getting caught dropped significantly.

The results in Shrewsbury, Missouri offer the clearest store-level evidence of what removing the machines actually produces. From January through May 2024, the Shrewsbury Police Department responded to 509 calls from that Walmart location.

During the same period in 2025, after self-checkout was removed, the number dropped to 183. Arrests fell from 108 to 49. The local police chief attributed the reduction directly to the removal of self-checkout.

The Shrewsbury data proves that human presence at checkout reduces theft measurably and rapidly. It also proves that self-checkout at that location was generating roughly one police call per day. The question it raises is how many other Walmart locations look like Shrewsbury and have not yet been identified.

What Retailers Are Actually Doing

The industry response has diverged into two directions, and the difference between them is the story.

Walmart has removed self-checkout at specific locations including South Philadelphia in March 2026, Shrewsbury, Cleveland, and New Mexico before that, and is planning to overhaul more than 650 existing stores this year.

The official explanation cites customer feedback and personalized service. The theft data is the operational driver. But Walmart is not simply reverting to staffed lanes.

It has simultaneously deployed AI-powered "Missed Scan Detection" cameras across more than 1,000 stores that identify items passing through without being scanned.

By mid-2025, the company had eliminated 60% of live guard camera monitoring across US regions and replaced it with AI-powered alerting and forensic search tools. The human is coming back to the register. The AI is watching everything else.

Walmart's dual approach is a recognition that the two problems are separate. Self-checkout created a theft opportunity at the point of transaction.

AI loss prevention addresses theft across the entire store. Walmart is solving both problems simultaneously with different tools

Dollar General removed self-checkout from 12,000 stores nationwide in 2024, the largest single rollback in retail history, and returned to staffed lanes.

Costco has deployed staff members to scan cart contents before customers reach the register, a human solution to a human-created problem.

Sam's Club is taking the opposite direction. In 2026, the Walmart subsidiary is remodeling all 600 US locations to become checkout-free.

Physical checkout stands are being phased out in favor of Scan & Go, a system where members scan items using the Sam's Club app as they shop and pay digitally before leaving.

At the exit, AI scanning technology verifies the cart without a human receipt checker. The Grapevine, Texas location, which opened in 2025, has no registers at all.

It is the furthest expression of a model that treats AI not as a supplement to the checkout process but as its replacement.

Walmart serves the general population across 4,600 US stores. Sam's Club serves members who pay $50 to $110 annually to shop there. That self-selection produces a customer who is on average more digitally comfortable and more willing to manage their own checkout experience. Sam's Club's AI bet works mainly because of who its members are and not just because of what the technology can do.

The Legislative Pressure Making AI More Attractive

The industry's internal reckoning is being matched by a growing legislative response that changes the economics of both options.

Lawmakers in California, Connecticut, Massachusetts, New York, Ohio, Rhode Island, and Washington are all pursuing or considering laws that would regulate self-checkout deployment.

In New York City, a proposed amendment would cap self-checkout purchases at 15 items in supermarkets and pharmacies and require at least one employee for every three kiosks.

If those laws pass, retailers operating self-checkout at scale will face mandatory human staffing ratios that erase much of the labor cost advantage that justified the machines in the first place.

AI-powered loss prevention becomes the cost-efficient answer to a regulatory environment that requires human presence without specifying what those humans have to do.

Requiring one employee per three kiosks is not just a labor protection measure. It reflects a recognition that when automated systems flag a missed scan or generate a theft alert, someone has to be responsible for what happens next. AI can watch. It cannot be held accountable. That gap is what the legislation is trying to close

That is the underlying dynamic behind Walmart's parallel tracks: removing self-checkout from locations where theft is worst, while deploying AI monitoring across the stores that keep it. The machines may be going. The AI watching them is not.