A large release covering security, several features the documentation described before they existed, and some things that were quietly not doing what their name said.
Security
Two-factor authentication now covers your project data. If you have enrolled an authenticator app, a session that has not completed the second step can no longer read or write any of it — projects, prompts, scans, raw answers, metrics, recommendations. It previously protected your account settings and stopped at the product itself. Accounts without two-factor enrolled are unaffected.
Browsers now get security headers from citerva.com: HSTS, clickjacking protection, MIME-sniffing protection, a referrer policy and a permissions policy.
Approvals and dismissals can no longer be attributed to the wrong person. Who approved a prompt set, and who dismissed a recommendation, is now recorded by the database itself rather than accepted from the request.
Ownership transfers now demote the outgoing owner. A transfer used to leave two people with full owner rights, including the right to transfer it back.
A workspace owner can no longer leave their own workspace. Doing so left the workspace with nobody able to manage billing or membership. Transfer ownership first.
Profile images are capped at 2 MB and limited to PNG, JPEG, WebP and GIF.
Features
Any scan in your history opens. The scans list showed the answers for the latest scan and nothing else, so an earlier one was a dead end — you could see that a rate had moved and could not read the answers that moved it. Every row now opens to that scan's runs, answers, sources and health.
Recommendations can be worked through. Mark one as being investigated, mark it done, change its priority, or dismiss it. The list previously only grew.
Your allowance is visible before it refuses anything. Scans, provider calls and active projects for the month now appear on your dashboard and on the scans page, against what your plan allows — the same figures the server checks before starting a scan.
Tell us the other domains you own. A documentation site, a help centre, a second brand: sources pointing at them are now attributed to you instead of being counted as somebody else's.
A checklist for getting to your first scan, on new accounts and new projects. It disappears when you have finished it.
Two things that were not doing what they said
Pausing a project now stops it. It used to change a badge and nothing else: the project kept its plan slot, kept accepting manual scans, and kept running its schedule. If you paused a project to stop it spending your allowance, it went on spending it. A paused project now runs no scheduled scan and refuses a manual one, and the schedule resumes at its next slot rather than firing everything it missed.
Scan history is never trimmed. The plans page listed a number of months of trend history per plan. It was never applied, and we have decided not to apply it — every scan you have run stays readable, on every plan. The row has been removed from the plans page rather than turned into a limit. The product exists to show a trend over time; putting your own history behind an upgrade would be a strange thing to sell.
Also
Team size now matches what your plan lists, and pending invitations count towards it — an invitation holds a seat until it is accepted or expires. The refusal names the limit and how many seats are left.
Five new documentation pages cover recommendations, scheduling, project settings, competitors and topics, and workspaces and members. One correction worth naming: the provider table said Google Gemini returns sources. It does so only when reached with a direct Google key, and not through the shared gateway — which decides whether its answers count towards your citation rate at all. That was true of Gemini and false of the product, and the page now says so.