Theft at self-checkout points is a growing challenge, causing significant inventory losses in the retail landscape. Visual AI technology ensures profitable and customer friendly-checkout points, without compromising on the convenience these systems provide.
The rise of self-checkout (SCO) systems is transforming the retail landscape, with two-thirds of consumers under 35 now preferring this method. However, as usage increases, so does theft, leading to significant loss and inventory discrepancies for retailers. Studies indicate that even generally honest customers are prone to exploiting SCOs. According to research published in The Grocer, only 63% of surveyed shoppers claimed they had never cheated at an SCO—a statistic that underscores the urgency for retailers to act.
Visual AI technology: a modern approach to fraud prevention
Retailers are exploring various strategies to combat SCO fraud, from increasing staff presence to reducing the number of SCOs. However, artificial intelligence (AI) and visual detection technologies are emerging as cost-effective and efficient solutions. These innovations offer real-time fraud interception capabilities and help prevent theft without disrupting the convenience that customers value.
The role of visual deep learning in fraud detection
At the heart of AI-driven fraud prevention lies visual deep learning (VDL), a transformative technology that powers next-generation security solutions. Compared to earlier AI models, VDL offers enhanced cost efficiency, energy savings, and stronger data protection. In an increasingly competitive retail environment, such efficiencies are crucial for maintaining healthy margins while delivering exceptional customer service. Particularly promising is Edge Deep Learning AI, which allows for real-time processing with minimal hardware requirements. Companies developing these solutions must possess specialised expertise in AI software development, model compression, embedded systems and edge hardware integration.
How AI detects fraud in real time
VDL-based SCO security systems address multiple fraud tactics:
• Hand movement analysis: Leading AI solutions monitor hand movements in real time, detecting whether an item is correctly scanned or deliberately bypassed.
• Barcode validation: If a barcode is obscured or a fraudulent barcode is used, the system identifies the mismatch and triggers an alert.
• High detection accuracy: The most advanced systems now boast fraud detection rates as high as 99%, significantly reducing false alarms.
Enhancing staff oversight with video snippets
Once an alert is triggered, the consumer can be prompted to self-correct the issue. If necessary, a staff member is made aware of the situation and receives a short video snippet showing exactly why the alarm was raised. Both the consumer and staff member can view the video to identify the specific item in question, allowing the issue to be solved immediately. This approach reduces the need to send video snippets to staff every time, making the process less staff-intensive and more efficient by giving the consumer the chance to self-correct before staff intervention is required.
Automating age verification for restricted sales
Another challenge at SCOs is the sale of age-restricted items. Traditionally, staff have had to verify a customer’s age manually, causing delays and cashier interventions. However, AI-driven age verification technology can pre-screen customers, flagging only those who may require ID checks. This reduces staff workload and speeds up transactions, enhancing both security and customer convenience.
Preventing mislabelling of non-barcoded items such as fresh produce and bakery
A common challenge at SCOs involves fraudulent selection of fresh produce, such as selecting a cheaper item for a more expensive fruit or vegetable. Advanced AI solutions can now visually recognise produce, even when packaged in transparent bags, and eliminate non-applicable options. This helps customers quickly and accurately select the right produce, speeding up the selection process and enhancing customer satisfaction by reducing confusion and errors during checkout. Advanced AI solutions can now visually recognise produce, even when packaged in transparent bags, eliminating opportunities for intentional or accidental incorrect entries.
The deterrent effect: preventing theft before it happens
Beyond immediate fraud detection, AI-powered SCO systems serve as a deterrent. When customers see that theft attempts are swiftly flagged, the risk of getting caught discourages further misuse. This ‘psychological barrier’ can significantly reduce theft incidents over time.
Practical considerations for retailers
Retailers considering AI-based SCO fraud prevention should evaluate factors beyond detection rates:
• Cost and hardware requirements: Modern solutions require only a small commercial camera and a compact edge computer, eliminating the need for cloud hosting, wi-fi, or expensive additional infrastructure.
• Ease of integration: Leading systems can be installed within days, seamlessly integrating with existing SCO setups.
• Protecting existing SCO investments: To maximise ROI, retailers should prioritise retrofit-friendly solutions that enhance security without requiring entirely new SCO hardware.
A profitable and secure future for SCOs
With visual deep learning now at market maturity, retailers have access to powerful tools for fraud prevention, operational efficiency, and enhanced customer experience. These systems quickly pay for themselves by reducing inventory losses while maintaining the convenience that customers expect. By investing in AI-driven fraud prevention, retailers can ensure that SCOs remain a secure, profitable and customer-friendly checkout option.
Get to know the entire GK product portfolio, including GK Vision, at the Retail Technology Show (RTS) in London on 2-3 April, 2025. Schedule a demo here:
Retail Technology Show 2025
https://www.gk-software.com/us/public/events/retail-technology-show-2025-en
Can’t make it to RTS? Enjoy an online demo of GK Vision below:
