Changelog | Seenlark

NOVEMBER 9, 2026

Alias suggestions — Spending limits now apply before web search, including shared search keys used alongside a workspace's own AI key. If optional shared search exceeds its allowance, generation continues with the workspace's own AI key without search. Completed searches remain accounted for if AI generation fails.
AI narratives — Manual workspace and topic generation now follows the scheduled generation entitlement: Free workspaces cannot generate narratives, including with their own API keys. Refused refreshes create no snapshot, and plan restrictions open the upgrade sheet. Saved narratives remain readable.
Demo workspaces — Source recommendations now enforce the existing prohibition on AI spending.
Closed signup — Google and GitHub cannot create accounts while signup is closed. Existing users can still sign in.

NOVEMBER 9, 2026

No more orphaned workspaces — you can no longer delete your account while you are the only owner of a workspace, including one with nobody else in it. Delete the workspace (which cancels its subscription) or make someone else its owner first. Account settings now says this instead of a generic error. The same rule applies when an admin removes a user.
The rule now applies in account settings too — the existing check that stopped the only owner of a team workspace from deleting their account only ran when an admin removed the user. It now also runs when you delete your own account.
Seats follow deleted teammates — when a member's account is deleted, their membership is removed and the workspace's seat count is updated, on both paths.

NOVEMBER 9, 2026

One dispatch per distinct platform — repeated selections are deduplicated in request order across individual scans, bulk runs, retries, previews, and schedules. Unknown platforms and empty selections are rejected; schedules can still use null to select all platforms.
Explicit input limits — direct prompts accept up to 2,000 characters, matching saved queries. Custom system prompts accept up to 2,000 characters, and requested output limits accept up to 4,000 tokens, including BYOK scans. Managed-key scans retain their existing lower output limit.

NOVEMBER 9, 2026

Resume workspace setup — Refreshing before brand setup resumes at the brand step using the existing workspace. Unsaved brand, topic, and competitor drafts must be entered again.
Retry the first scan — A failed scan can be retried in the same review session without repeating successful workspace setup.
Keep the first query — The first scan uses the first saved seed query, or a saved starter prompt when no seeds were supplied, linked to the onboarding topic. Scans still start only when you click the button.
Consistent routing ? Opening the app root also resumes unfinished setup. Demo workspaces do not count as configured workspaces, and pending invitations retain precedence for users without a real workspace.
Safe review retries ? Successful setup locks earlier steps; completion restores the resumed workspace as active.
Count saved queries ? Setup initializes quota counters, counts its topic and queries, and accepts at most ten seeds within the tier?s remaining keyword quota.

NOVEMBER 9, 2026

Every component is driven by a real check — scan execution, scheduled scans, insights, alerts, email delivery and webhook delivery now reflect the same health checks our own monitoring alerts on: whether queued scans and notifications are being picked up, and whether the background jobs behind each surface have run on time. Before, those six rows were always shown as operational.
"Unknown" and "Not monitored" instead of a default green — if the status page can't reach a check, the affected rows say "Unknown" and the summary banner says some systems could not be checked. Documentation is checked by contacting the docs site, and is shown as "Not monitored" where no docs site is configured.
Daily digests no longer look like a stuck email queue — notifications waiting for the daily digest were counted as an email backlog in our internal monitoring, which could raise a false alert every day. They are now left out until the digest is composed.

NOVEMBER 9, 2026

One-click unsubscribe — alert and digest emails carry List-Unsubscribe and List-Unsubscribe-Post headers (RFC 8058), so the Unsubscribe button in Gmail, Yahoo and Apple Mail works. Each email also links to your notification settings and to an unsubscribe page. Unsubscribing turns off email for the notification types in that message only; they still appear in the app.
Legal footer — every email now ends with the operator's name and postal address (as in the imprint) and links to the privacy policy and imprint.
Language tag — emails declare the language they are written in instead of always en.
Postmark — plain-text bodies are now sent alongside HTML.

NOVEMBER 8, 2026

NOVEMBER 8, 2026

NOVEMBER 7, 2026

OCTOBER 24, 2026

OCTOBER 23, 2026

OCTOBER 5, 2026