Register as a relying party on the Yoti developer portal and obtain an API key and SDK ID for the Age Verification product.
Choose the appropriate age assurance method: document-based age verification (scan government ID), estimation-based age assurance (facial age estimation, no PII stored), or Yoti Digital ID (user presents a pre-verified credential).
Integrate the Yoti age verification via the hosted UI redirect or the Yoti.js / mobile SDK; configure the age threshold (e.g., age_over_18 or age_over_21) and the allowed verification methods in your Yoti application settings.
Handle the verification result callback: Yoti returns a boolean age_verified attribute and, for document-based paths, a date_of_birth or age_over_N claim; store only the minimum required claim as mandated by GDPR data minimization.
Confirm your implementation satisfies Ofcom's age assurance requirements under the UK Online Safety Act (enforcement began January 2025 for pornographic content, July 2025 for user-generated content platforms) by using a certified age assurance provider.
Conduct user acceptance testing with edge cases including expired documents and facial estimation at boundary ages (17/18) to validate threshold accuracy before launching.
Known gotchas
Ofcom's enforcement timelines under the UK Online Safety Act differ by service category; confirm which deadline applies to your service type before selecting a target go-live date.
Facial age estimation does not produce a verified date of birth — it produces a probabilistic age range; for services that require verified age (not just estimated age), document-based or credential-based methods are required.
Yoti age verification in the UK applies to UK-resident users; for US-resident users different state-level age verification laws may apply (e.g., COPPA for under-13, and various state laws for adult content), potentially requiring a different technical path.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp