Shufti Named Leading Vendor in Liminal’s 2026 Age Estimation Index

Key Takeaways

• Liminal evaluated 80 age estimation solutions in 2026. Only 17 cleared the leader thresholds.
• Shufti is named a Leader, sitting in the upper cluster on Product Execution and Strategy.
• 62% of practitioners are reviewing alternatives after running into problems with facial age estimation, citing accuracy and transparency issues.
• 87% want one platform for age estimation compliance in 2026; most still run four.
• Shufti ships risk-based estimation with on-device inference already in production.

In its 2026 Age Estimation Index, Liminal scored 80 vendors against a 33% Product Execution cutoff and a 46% Strategy cutoff. Only 17 cleared both to qualify as Leaders. Shufti is one of them, positioned in the upper cluster of the leadership matrix on Product Execution and Strategy.

The placement matters because of Liminal’s underlying data documents. Practitioner confidence in facial age estimation accuracy is under pressure. 62% of practitioners using it report that problems with the technology have triggered internal reviews of alternative age assurance methods, citing limited model transparency, difficulty integrating with existing workflows, and inaccurate results across specific demographic groups.

The leader cohort isn’t a list of who’s biggest. It’s a list of vendors equipped for where the market is going, which is somewhere different from where it sat 18 months ago.

What Liminal’s 2026 Age Estimation Index Actually Measures

The index scores 80 vendors across three axes: Product Execution, Strategy, and Market Presence. Two carry hard thresholds. A vendor that falls below either threshold doesn’t qualify as a Leader, regardless of brand or scale.

The Two Scores Buyers Should Care About

Product Execution measures verification accuracy under load, image and sensor input quality, liveness and presentation-attack detection, demographic performance consistency, and end-user experience, including completion rate. It captures what the platform actually does when a real user lands on it.

Strategy measures forward-looking investment: scalability across regions and languages, automation and orchestration across assurance steps, integration with identity, trust-and-safety, and fraud systems, resilience against synthetic media and other evasion techniques, continuous model governance with drift monitoring and bias testing, policy and jurisdictional adaptability, and controls for data minimisation. It captures where the platform is going.

Why is the Leader Cut Narrower than it Looks

63 of 80 vendors didn’t clear the cut. Roughly 79% of the age estimation market failed thresholds that practitioners themselves told Liminal were the minimum bar.

The Product Execution cutoff (33%) filters out vendors that can’t reliably estimate age under production conditions. The Strategy cutoff (46%) filters out vendors with a working model today but no credible plan for continuous governance, drift detection, or jurisdictional adaptability. Most failures sit on the Strategy line.

Shufti cleared both thresholds with margin. Liminal placed Shufti in the upper cluster among 17 best age verification vendors filtered out of 80.

The Four Fears Reshaping Age Estimation Buying in 2026

Liminal’s age estimation buyer demand survey of 58 practitioners, combined with regulatory pressure from the UK Children’s Code and the California Age-Appropriate Design Code Act, points to four fears now dominating vendor selection. Each shows up in the index methodology and each is testable in a vendor demo.

The regulatory pressure shaping these fears is concrete. The UK’s Children’s Code, in force since September 2021 and enforceable under UK GDPR with fines up to £17.5 million or 4% of global turnover, sets 15 standards for online services likely to be accessed by under-18s. The ICO’s age assurance opinion names facial age estimation as currently the most widely used age assurance method and introduces the “waterfall technique” of combining estimation with verification fall-backs. California’s Age-Appropriate Design Code Act follows the same pattern. The EU AI Act treats age inference systems as a regulated category of biometric processing.

The trust crisis and the regulatory tightening aren’t separate stories. They’re the same story.

Where Shufti landed in Liminal’s 2026 Age Estimation Index

Shufti is named a Leader, positioned in the upper cluster of the matrix on both Product Execution and Strategy. The placement reads cleanly because the analyst note lays out specific capabilities mapped to where buyer demand is going.

What the Analyst Note Calls Out

Liminal credits Shufti with a risk-based, multi-method age assurance architecture built on its in-house identity verification platform. The configuration combines AI-based facial age estimation, government ID document verification, and configurable liveness checks, with step-up paths to higher-assurance methods when estimates fall near regulatory boundaries rather than enforcing a single gate. Privacy-preserving age checks delete end-user data immediately after a decision. Fraud and integrity controls (spoofing, presentation attack, replay, deepfake, and duplicate account detection) run by default.

On accuracy, Shufti reports its age estimation model outperforms a major industry benchmark vendor across multiple accuracy metrics based on internal testing on a consistent 100,000-image benchmark spanning diverse demographics. Performance is tracked by race and gender rather than aggregate averages. Training uses large consented adult datasets and augmented synthetic data for minors, with active bias monitoring to support tighter confidence thresholds and reduced reliance on document fall-backs.

Capabilities mapped to buyer demand

Shufti’s privacy-preserving on-device age estimation is in production today. Inference runs locally without internet connectivity, transmits no facial data, and issues reusable age tokens. That’s a direct answer to the privacy and on-device demand of the 2026 buyer-demand survey found, and a capability most leaders in the cohort don’t yet have shipping. Shufti also holds KJM approval for the German market and has published NIST FATE evaluation results, including top-15 placements in several Challenge 25 and Child Online Safety Act metrics.

How Shufti handles age estimation challenges faced in real world

If you’ve shipped age estimation to a production audience, you’ve felt at least three of the four fears Liminal documented in its buyer demand survey. The age estimation model returns low-confidence decisions on hard demographic edges. Bias monitoring is either absent or ceremonial. The vendor count used by each company keeps climbing. Privacy reviews keep flagging facial data transmission.

Shufti’s age estimation is structurally designed against each of those failure modes.

Confidence crisis: Shufti runs age estimation inside a risk-based multi-method architecture rather than as a single gate. When the estimated age sits close to a regulated threshold (16 or 18 in most jurisdictions), the platform steps up to document verification, authoritative database lookup, NFC chip read, or liveness, invoked only when the confidence profile warrants it. The fall-back is engineered, not bolted on, which is the gap Liminal’s data points to in the 62% who’ve lost trust.

Static model risk: Training uses large consented adult datasets and augmented synthetic data for minors, with bias monitoring tracked by race and gender. Internal benchmarking on a 100,000-image dataset shows Shufti’s age estimation outperforms a major industry benchmark vendor across multiple accuracy metrics. Shufti’s age estimation algorithm is continuously updated, validated and monitored for demographic bias. It’s the structural answer to a market where 62% are running models they don’t refresh.

Vendor sprawl: Age estimation is one capability inside one Shufti platform that also covers KYC, KYB, AML screening, document verification, face verification, deepfake detection, and fraud prevention. The 87% of practitioners who want one platform but currently run four close that gap with one integration.

Privacy + on-device demand: Shufti’s on-device age estimation is in production today. Inference runs locally (device/browser) without internet connectivity, no facial data is transmitted, and reusable age tokens replace repeated checks across sessions. That maps directly to the Children’s Code’s data minimisation requirements, CAADCA’s privacy-by-default expectations, and the EU AI Act’s biometric processing controls.

The independent validation behind the platform: 17-Leader recognition from Liminal in the 2026 Age Estimation Index, KJM approval for the German market, iBeta Level 3 conformance under ISO/IEC 30107-3 for liveness attack detection, and published NIST FATE evaluation results.

One platform. Fully owned technology. Global coverage with real local depth.