Mobile driver's license verification: Why digital IDs break plastic-first onboarding

A driver in Phoenix signs up for a delivery platform. Her license lives in Apple Wallet. That stopped being unusual in Arizona some time ago: 1,227,833 Arizonans hold a mobile driver's license, a 21% conversion of the state's licensed drivers.

The signup flow asks her to photograph her driver's license. She now has two options. She can photograph her phone screen using a second device. Or she can find the plastic card she stopped carrying.

If she picks the first option, most fraud engines flag the image instantly. A photo of a screen is exactly what presentation attack detection is trained to reject. If she picks the second, a digital-first onboarding flow just sent a legitimate user to dig through a drawer.

Either way, the platform loses. And the platform's fraud dashboard records it as a win.

At Signzy, we verify driver's licenses for banks, fintechs, and delivery platforms across the US, which is exactly where this collision between wallet-based credentials and plastic-first capture shows up first.

Related Solution

Driver's license verification APIDocument OCRIdentity Verification

Mobile driver's license adoption in the US

Mobile driver's licenses (mDLs) are state-issued credentials that live in Apple Wallet, Google Wallet, Samsung Wallet, or a state app. They follow the ISO/IEC 18013-5 standard, the same one TSA uses to accept them at airport checkpoints. As of mid-2026, 21 states plus Puerto Rico have live programs. TSA recognizes digital IDs from 20 states plus Puerto Rico.

Published enrollment figures show where this is concentrated:

Sources: ITU presentation of state DMV figures (2026); GovTech national estimate (2025); state wallet announcements.

The full list of live programs: Alaska, Arizona, Arkansas, California, Colorado, Delaware, Georgia, Hawaii, Illinois, Iowa, Kentucky, Louisiana, Maryland, Montana, New Mexico, New York, North Dakota, Ohio, Utah, Virginia, West Virginia, and Puerto Rico.

Two percent nationally sounds ignorable. It is not, for two reasons. First, adoption is not evenly spread: in Arizona, one in five drivers already converted, and California added Samsung Wallet support in April 2026 to accelerate its 3.3 million. Second, the growth curve is steep. ABI Research forecasts 143 million US mDL holders by 2030. That is more than half of the country's licensed drivers, four years out.

If your onboarding volume includes Arizona, California, Colorado, or Maryland, mDL holders are already entering your funnel today.

Why mobile driver's licenses fail traditional document verification

Here is the mechanical problem, step by step.

Document-capture verification was built around one assumption: the user photographs a physical card. Every downstream check depends on that assumption.

Read that table from a fraud engine's point of view. A genuine, state-issued, cryptographically signed credential presents through a channel that scores like a fraud attempt on all five layers. The system is not malfunctioning. It is correctly rejecting the wrong object, because the right object was never designed to be photographed.

This is the part most teams have not internalized: an mDL is not a picture of a license on a phone. It is a signed data payload. The ISO 18013-5 standard defines how it is presented in person, over Bluetooth or NFC, to a reader that validates the state's cryptographic signature. Verified correctly, an mDL is stronger evidence than plastic, because forging it means defeating the issuing state's cryptography rather than a hologram.

Photographed off a screen, it is weaker than plastic. All the cryptography is stripped away, and what remains is an image with every spoof signal present.

How remote mobile driver's license verification works

In-person mDL verification is settled. ISO 18013-5 works, TSA uses it daily, and retail age checks are being piloted on it.

Remote verification, the kind onboarding flows need, is not settled. The relevant standard, ISO/IEC 18013-7, was published as a technical specification in October 2024 and defines how an mDL is presented over the internet to a verifier that is not in the room. The specification exists, interoperability events have run, and a next edition is in progress. But as of August 2026 it remains a moving target: NIST notes that its annexes depend on protocols still co-evolving, and wallet support for unattended presentation is inconsistent across states and platforms.

The practical consequence for anyone running remote onboarding:

So a fintech, marketplace, or insurer verifying users remotely today sits in a gap: a growing share of applicants hold a credential that is cryptographically superior to plastic, and no photograph-based flow can accept it.

Where mobile driver's license verification creates onboarding friction

Gig and delivery onboarding. Driver supply is won and lost on activation speed. An applicant whose license is wallet-only either abandons signup or completes it days later with plastic. In Arizona, that friction now applies to one in five potential drivers. The platforms that onboard in minutes hold the supply advantage.

Age-restricted commerce. A 22-year-old buying online is the demographic most likely to have converted to a wallet credential and least likely to carry a physical card daily. A hard "photograph your ID" gate at checkout is a paid-traffic write-off for that segment.

Banking and fintech KYC. A rejected capture does not read as "unsupported credential type" to a compliance analyst. It reads as a failed document check, which routes to manual review. Each mDL holder pushed into the review queue costs analyst minutes and adds a day of onboarding delay for a customer who did nothing wrong.

Car rental and mobility. The person standing at the counter can present an mDL in person under 18013-5. The same person pre-registering through the app cannot. The channel with the lower fraud exposure gets the better credential; the channel with the higher exposure gets a photo.

How to detect mobile driver's license failures in your onboarding funnel

You do not need a new vendor to answer this. You need one query against your own capture logs. mDL holders hitting a plastic-only flow leave a recognizable signature:

Run the state-level cut first. National averages will hide this: at 2% national adoption the signal drowns, but Arizona at 21% will show up clearly if it is happening.

One more reason this matters beyond conversion: the fraud asymmetry. Fraudsters do not photograph wallet screens; they submit AI-generated images of plastic licenses, which we have covered in depth in our fake ID detection and AI-generated fake ID analyses. So a plastic-only capture flow now faces the worst version of both worlds. Legitimate users increasingly hold a credential the flow cannot accept, while attackers submit exactly the artifact the flow was built to trust. The gap between what your verification accepts and what honest users carry is a gap fraud fills.

The near-term answer for most teams is layered rather than pure: extract and verify the plastic when it exists, validate the underlying record against issuing sources such as DMV-aligned databases, and match the face to the credential holder, so the verification does not stand or fall on the photograph alone. That is how driver's license verification works within a complete onboarding check today, alongside the rest of the US verification API stack, and the API documentation shows what is returned field by field. As 18013-7 implementations stabilize, native mDL presentation becomes another input into the same decision, not a rebuild.

How to prepare your KYC flow for mobile driver's licenses

Before the end of this week, pull ninety days of capture rejections and segment them by issuing state. If Arizona, California, Colorado, or Maryland over-index on presentation attack flags or capture abandonment, you are not looking at fraud. You are looking at your earliest mDL adopters failing to hand you a credential that is better than the one you asked for. Count them, cost them, and put a number on what "wait for the standard to settle" is worth against 143 million holders by 2030.