Borrowing the State's Front Door: GOV.UK One Login and the Commercial App
Photo: Flame, not lame, CC0, via Wikimedia Commons
Somewhere in a discovery workshop, someone always asks it: 'Can we just use the government login?' It's a fair question. GOV.UK One Login is familiar to millions of UK users, it's been built to a standard most commercial apps would envy, and it removes a password from someone's life.
The honest answer is 'maybe, and not for the reason you think'. Here's how to work out where you stand before you put a logo on a slide.
Where One Login sits, and where it doesn't
GOV.UK One Login is the single sign-in being rolled out across central government services — HMRC, DVLA, DWP and a growing list of others. It does two distinct jobs, and mixing them up is the most common mistake in early scoping.
Authentication answers 'is this the same person who came back?'. Identity verification answers 'is this person who they say they are?', usually by checking a photo ID and matching a face.
Most commercial teams actually want the first and assume they need the second. Almost every product that asks for verified identity when it only needs a returning-user token is over-collecting, and paying for it in drop-off.
The eligibility question, asked properly
One Login is built for public services and for services that deliver a public function, onboarded as a relying party through GDS. Being a supplier to the public sector does not automatically make your app eligible. Nor does having a public-sector client. The test is about the service, not the company logo on the contract.
So the realistic routes are narrower than the enthusiasm suggests:
- You operate a service on behalf of a public body, under a contract that names it.
- You run something with a clear statutory or public-benefit function where government has agreed the connection.
- You're building a white-label app for a council, NHS trust or arm's-length body and One Login is the identity rail they've chosen.
If none of those describe you, stop here. Use a commercial identity provider, do it well, and revisit in a year. Chasing an ineligible integration will cost you a quarter and a lot of goodwill.
And don't confuse neighbours: NHS login is a separate service for health, and it has its own onboarding, its own attribute list and its own clinical context. Turning up with the wrong one in an architecture diagram is a quick way to lose the room.
What you actually get back
Attributes matter more than protocols. Authentication typically returns an email address and a stable pairwise identifier — a subject value that is unique to your service, which is exactly what you want for account linking. Identity verification adds verified personal details, with the exact attribute list agreed during onboarding.
Two habits to build now:
Key on the identifier, not the email. People change email addresses. If your account model is an email primary key, a One Login user who switches address becomes two accounts and a support ticket.
Read the current scope list before you promise anything. That blog post from 2024 is not a specification. Attribute availability and scope names change, and your integration spec should cite the version you were onboarded against.
Then apply the boring discipline of data minimisation: request the smallest set that delivers the service, write the purpose in language a user can understand, and make sure the consent screen isn't asking for five things when one would do.
Consent that survives contact with users
One Login shows the user what's being shared before they're sent back to you. That screen is your shop window, and it's read more carefully than your privacy notice.
Keep it short. Be specific about why. If you're asking for verification to meet a regulatory requirement, say so in one sentence — users are far more tolerant of a stated reason than of a vague one. And make sure the consent journey is tested with a screen reader, a keyboard and a Welsh-language user if your service touches Wales. Accessibility and language parity aren't garnish on a public-facing identity flow; they're the baseline.
The fallback is the design
This is the part that gets skipped, and it's the part users will remember.
One Login will occasionally be unavailable, or a user's account will be locked, or their verified identity will fail a check they can't fix at 8am on a Monday. If your only route into the app is a rail you don't control, an outage becomes a business interruption.
Design for it deliberately:
A second door. Magic link, phone plus one-time code, or a partner identity provider. Not a hidden one — a visible, equally tested option from day one.
Save and resume. If someone bails mid-journey, let them come back to where they were rather than starting again.
One person, one account. Decide what happens when someone with a live account signs in through a different route. The wrong answer is a silent duplicate.
A deletion story. If a user deletes their account, what happens to the stored identifier? And when they re-register and the same identifier returns — do you re-attach old data to a new account? Think that through before it happens in production.
An outage runbook. Who watches the status page, what you show in-app, and who's allowed to flip the feature flag.
What to put in the integration brief
Before anyone writes code: eligibility confirmed in writing; onboarding contact named; attribute list versioned; the authentication-versus-verification decision recorded with a reason; the fallback path specified and tested; the account-linking rules written down; session length and step-up re-authentication agreed; and legal sign-off on the privacy notice, data-retention period and deletion behaviour.
Borrowing the state's front door is a genuinely good idea when it fits. It just needs to be a decision you can explain to a product board — and to a user who's forgotten their password at the worst possible moment.