Age estimation vs ID checks: how to choose

Published 1 July 2026 · General information, not legal advice

Australia’s age rules are technology-neutral: regulators name several acceptable ways to check age and let you pick something proportionate to your risk. In practice the choice comes down to two families of method, face-based age estimation and document-based verification. They solve the same problem in very different ways, and the trade-offs decide which belongs where in your flow. This guide compares them honestly, including what the Australian government’s own technology trial found. It is general information, not legal advice.

The two methods, precisely

Age estimation analyses a live image of a face and returns an estimated age range. No document is involved and no identity is established: the system doesn’t learn who you are, only roughly how old you look. The result is an estimate with a stated confidence, not a date of birth.

ID verification establishes age from an authoritative document such as a driver’s licence or passport, usually by extracting the date of birth and matching the holder’s face to the document photo. It is precise, but the user must present identity data, and you must handle it.

What the government’s own trial found

Before the current rules took effect, the government commissioned an independent Age Assurance Technology Trial that tested more than 60 solutions from 48 providers, including estimation, document verification and layered systems, in field trials and mystery-shopper exercises. Its headline conclusion: age assurance can be done, and a wide variety of technologies already meet meaningful thresholds for accuracy, security and privacy when carefully selected and implemented.

The report was also candid about limits. Estimation accuracy varies near age boundaries (16 to 20 year olds are the hard cases) and performance can differ across demographics and lighting conditions. Its recommended answer to those limits was not to abandon estimation but to layer methods: the report calls this “successive validation”, where no single method has to carry every case.

The trade-offs that actually decide it

• Friction. An estimation check is a selfie that takes seconds; most users pass without leaving the flow. A document check means finding an ID, photographing it, and waiting. Every extra step loses real customers at checkout.
• Privacy and breach exposure. Estimation collects no identity data: there is nothing to retain and nothing to breach. Document flows collect exactly the data Australians have learned to be wary about handing over. Whatever you collect, you must protect under the Privacy Act, and the safest data is the data you never had.
• Precision. A document gives an exact date of birth; an estimate gives a range. For a threshold question (“clearly over 18?”) a range with a buffer answers most cases. Right at the boundary, documents win.
• Coverage. Not everyone has an ID to hand, and some legitimately adult customers will abandon rather than upload one. Estimation covers them; documents don’t.

The pattern that usually wins: layered

The practical answer, and the trial’s recommended one, is a waterfall. Start with the lowest-friction check and escalate only when needed:

• Estimate age from a live selfie, with liveness confirming a real person is present.
• Clearly above your threshold, with a sensible buffer: proceed. This is most adult customers, and they never touch a document.
• Borderline or below: step up to an ID check, or decline.

The buffer does the compliance work. If the legal line is 18, requiring an estimate of, say, 25 or older to pass without documents means the borderline cases (where estimation is weakest) all flow to the stronger check. You choose the buffer to match your risk appetite and document your reasoning.

Where eaglecheck fits

eaglecheck provides the face-analysis layer of that waterfall: age estimation from a live photo, liveness detection to confirm presence, and face matching to compare a selfie against an ID portrait when a flow steps up. Images are processed in AWS’s Sydney region and are not retained. The boundaries, stated plainly: we do not verify document authenticity, and we are not an accredited digital identity provider. Whether a given flow meets your specific obligations is a question for your regulator or lawyer; our part is making the face checks fast, private and easy to integrate.

Choosing for your case

• 18+ gate with high traffic (retail, content, gaming): estimation first, buffered, with document step-up. Friction kills conversion here.
• Boundary near 16 (social features): expect more step-ups; the 16 to 20 band is where estimation is least certain, so size the buffer accordingly.
• Regulated document pathways (e.g. some NSW delivery options): the document flow is prescribed, but estimation still helps at the handover and as a first screen.
• Whatever you choose, write it down. “Reasonable steps” and “proportionate” are judgements; a documented rationale is your best evidence you made them properly.

Sources

• Department of Infrastructure: Age Assurance Technology Trial final report
• Age Assurance Technology Trial
• eSafety Commissioner: Industry codes