AI in identity verification: Improve pass rates and reduce fraud

What's changing about AI in identity verification?

In modern digital identity verification, AI is often used to enhance and automate checks that confirm someone is who they claim to be. It can, for example, review an ID document for signs of tampering, such as a missing security element or fonts that don't match, much faster – and with greater accuracy – than a human reviewer. We cover this in greater detail in a section below.

It can now help you understand why customers fail and provide recommendations to optimise your verification journeys and improve your pass rate.

However, while AI can be useful, it's important to remember that it isn't a replacement for human judgement. Most implementations still pair automated decisioning with human oversight for edge cases.

Where AI is already being used

Of course, AI has been used in IDV for a long time now. These are some of the most common ways AI is automating verification checks:

• Document verification: AI scans government-issued IDs like a passport or driver's license and pulls out the relevant fields like name and date of birth. It looks for security features like holograms and optically variable ink and spots tampering a human might not, such as portrait lighting that doesn't match the rest of the image or a barcode that doesn't line up with the printed text.
• Biometrics and liveness detection: Is the person in the selfie the same as the one in the ID? And are they actually there, not a photo, screen, mask, or deepfake impersonating them? Using facial recognition and liveness detection, AI can handle these checks, too.
• Fraud detection: Individual checks can miss patterns that only show up at scale (several accounts with different names and addresses that all share the same device, for example). AI helps by sifting through large volumes to surface what a single review might miss.
• Repeat-user detection: Biometric models use computer vision to recognize when the same face has shown up before, which helps flag attempts at identity theft, opening multiple accounts, or reusing a stolen identity.
• Decisioning and automation: Every check gets weighed against your rules and thresholds, so low-risk users can clear automatically and others get flagged for review.

The above scenarios use AI to help determine whether an individual verification attempt can be trusted. But when you use AI to analyse the data generated by thousands or millions of attempts as well, it can help you understand if your verification process is working.

How GBG helps businesses optimise identity verification with AI-powered insights

There's a difference between knowing your pass rate and understanding what you can do to improve it, and that's what our Foresight solution was built to address. It's an intelligence layer that sits on top of your identity journey performance, adds anonymised peer benchmarking, and turns both into recommendations tailored to where and how you can improve.

Foresight runs on top of our clients' GBG solution sets. For example, this can be GBG Go, our identity verification platform with more than 110 configurable modules like document and biometric verification that use AI to authenticate documents, match faces, and detect liveness in onboarding flows.

If you're currently running verification through GBG Go, you don't need new tools or new data to start using Foresight: it works with performance data you're already generating. What's more, Foresight is easy to implement.

Here are three ways you can use Foresight to improve your identity verification process:

Use AI to better interpret your verification data (and get tailored recommendations)

We process about 800 million identity verification transactions a year across more than 21,000 customers. This scale gives our Foresight solution's recommendations real context.

Instead of handing you raw performance data you have to interpret yourself, our solution layers AI-powered recommendations directly onto your existing GBG integration. This includes potential improvements to your configurations, data sources, data capture, and capability usage. And it doesn't only suggest what to add: it also points out what to remove, like a check that's actively hurting your pass rate.

 

These recommendations are tailored to your business, not generic, one-size-fits-all suggestions, and they're ranked by expected impact, so you can prioritise them and manage your resources better. View them in your dashboard or have them sent straight to email or Slack.

Every recommendation comes with reasoning behind it (like which data points and comparisons led to it) so your team not only knows what to change but also why it's likely to work.

Understand the gaps in your pass rate with peer benchmarking

Our Foresight solution measures your numbers against relevant peer groups and sector trends, drawn from anonymised data across our customer base.

Here's an example of what it can reveal: In financial services alone, under-25 pass rates range from 61% to 85%. This means that a 65% score might look fine in isolation – but in this context, you might realise you're actually leaving 20 points on the table.

 

The same goes for your configuration itself. Across our customer base, 85% of organisations require a 2+2 match, 84% enable refer decisions, 95% allow partial matching, and 42% perform mortality checks. "Standard" practice varies more than you might think, and seeing how peers configure their journeys can help you build a smoother experience for each of your customer groups.

See one of our client testimonials here:

 

Fight AI-enabled fraud with AI-powered insights

Deepfakes, synthetic identities, document manipulation, and injection attacks evolve quickly, which means your controls should, too. Our AI-powered monitoring watches for signals that something has shifted, such as a sudden increase in failed biometric authentication or an unexpected change in pass rates in one customer segment.

 

It also helps you respond more precisely when risks change. By analysing performance across customer segments and markets, you can identify where controls need tightening rather than introducing additional friction for every customer. Tracking rule effectiveness over time tells you whether existing controls continue to perform as expected as fraud patterns shift.

With ongoing monitoring and automated alerts, you don't have to wait for a periodic review to uncover an issue: your team can focus dashboards on the fraud and performance metrics that matter most and investigate sooner.

And because our Foresight solution keeps a record of configuration changes and their impact, you have evidence to explain why controls were adjusted and defend your decisions later if you need to.

How Foresight helped a customer increase its pass rate by 10%

Our client came to us with a 70% identity verification pass rate. Our Foresight solution's peer benchmarking showed an 80% pass rate was achievable.

Foresight analysed the client's performance and recommended adding a UK credit reference agency (CRA) check to its verification journey. Rather than implementing a change based on the AI recommendation alone, we worked with the client to test the proposed change and measure its impact.

Real-time performance metrics demonstrated a 10% uplift in pass rate, taking the client from 70% to 80%, which is exactly what the benchmarking predicted.

 

 

Turn identity verification data into continuous improvement with Foresight

AI is changing not only how businesses verify identities, but what they can learn from every attempt. There's an opportunity to use those insights to build an identity strategy that evolves alongside your customers and the threats you face.

By treating verification as an ongoing source of intelligence rather than a one-time check, you can make smarter decisions about where to focus and what to improve next.

Ready to see what Foresight can uncover in your identity verification data? Book a demo today.