Social login

The offer is plain: you already have an account somewhere, so use it. The site sends the reader to a provider, the provider asks whether this site may know who they are and what it may see, and the reader comes back with an identity nobody had to invent. No new password, no verification mail, no forgotten-password flow six months later. For most consumer products it is the highest converting sign-in route on the screen, which is exactly why it needs designing rather than dropping in.

The failure that costs the most is account linking. Someone signs up with Google in March, comes back in September, types the address into the email field, and meets either a stranger’s empty account or an error that tells them nothing. Match on the verified email address, say which route was used the first time (“you signed up with Google”), and offer to link the two rather than making the reader guess. Provider order is a real decision too: the row should read as a set of equals rather than a recommendation with alternatives beneath it, unless the audience genuinely has one dominant account. Show the route the reader used last if you can remember it, since that answers the question the buttons are really asking.

The buttons themselves are the most rule-bound artwork in any sign-in screen. Each provider publishes brand guidance covering the mark, the wording, the minimum size, and which backgrounds are allowed, and those rules override house button styling, so the craft is in giving visually different marks equal weight rather than in restyling them. Keep the verb consistent across the row, put the email route below a labelled divider instead of hiding it, and never let the row grow past the three or four providers the audience actually uses.

Two honest costs. Delegating identity means the provider can see where their users sign in and can end an account you depend on, which is why the email address should be stored as your own record from the first session rather than fetched on demand. And a row of provider buttons is a privacy statement whether or not it was meant as one, so the scopes requested should be the smallest set that works: a name and an email address, not a contact list. For a sign-in that leaves nothing with a third party at all, the alternatives are a passkey, a magic link, or a one-time code.