What we’ve shipped lately

The admin console gains a customer-success surface aimed at the support team. It pulls from the audit log and account activity to surface accounts that need attention, and it does the boring work — picking who to reach out to and when — so the team can focus on the conversations.

Risk scoring blends sign-in cadence, audit frequency, content approval throughput, and integration health. Anything trending the wrong way gets flagged in a single sortable list.

Impersonation is read-only by default and every action takes an extra confirmation. Every impersonated session is logged with the admin user, the target workspace, and the start/stop timestamps.

The most common workflow on the platform was: a reviewer leaves notes, the original author reads them, opens the editor, makes the changes by hand. Auto-revision collapses that into a button.

Click Auto-revise on a draft with revisions requested and the content engine reads the original brief, the current draft, and every reviewer note. It regenerates while preserving anything not flagged, then re-runs compliance and files the new version for re-review.

We've watched internal pilots cut review-to-approval time by roughly 4x on long-form content. Short LinkedIn drafts are sometimes faster to hand-edit — for everything else, this is now the default.

WordPress publishing has been live since launch. This release adds two more channels: Webflow and Shopify.

Webflow uses an API token scoped to CMS write, and we map standard fields automatically (name, post-body, main-image). For non-standard collection schemas, advanced override lets you point each draft field at the right Webflow field.

Shopify uses an Admin API custom app with write_content and read_content scopes. You provide the store domain, token, and blog ID. We verify the token has the right scopes during the connection test, so you find out immediately if something's missing — not on the first failed publish.

Credentials for both adapters are AES-256-GCM encrypted at rest. See the security release notes from earlier this month for context.

Connecting WordPress, Webflow, Shopify, or LinkedIn means handing us long-lived credentials. Until this release, we stored them in the database at rest as plaintext alongside the workspace record — protected by infrastructure controls but not by the application layer.

Every credential is now encrypted with AES-256-GCM before being written to the database. Keys are derived from a workspace-scoped master key plus a per-record nonce, so a leaked row is useless without the master key, and a leaked master key doesn't break the per-row authenticated-encryption guarantee.

The one-shot migration re-encrypted every existing record in place on deploy. The plaintext now only exists in memory during a live publish call, and we wipe the buffer when the call finishes.

No customer action required. If you're on Enterprise and want the threat-model writeup, ping support.

The visibility tester used to run all prompts against Anthropic and surface a single rank. This release fans the tests out across four engines in parallel, with two more coming online in rotation.

Why this matters: a brand can rank 1 on Perplexity and unranked on Gemini for the same prompt. The aggregate-only view hid that. The per-engine breakdown surfaces which engines you're losing on, which is exactly where the next round of content investment should land.

Non-determinism is the other thing this release addresses. We now send each prompt 3 times per engine and use the modal rank. A 1–2 rank wobble between runs is normal and the data is much cleaner.

Operationally, if one engine returns errors, the audit still completes — we log the failure and continue. You'll see a small "engine unavailable" badge on that engine's column.

The old audit experience: kick off, watch a spinner, wait 5+ minutes. If you closed the tab, you got no signal that it finished. That was the most common support ticket.

Audits now run as background jobs and stream progress to the page via SSE. You see each step tick over in real time — Crawl → Analyze → Test → Score — and the page is safe to close at any point. The job continues. You get an email when it lands.

This also unblocked a roadmap item we couldn't ship before: scheduled audits don't need an open browser tab anymore, because nothing on the audit path is browser-blocking.

The platform shipped as single-user. This release adds full team collaboration: invites, roles, and a workspace switcher for agencies and consultants who operate in multiple client workspaces.

Roles are role-based, not permission-based — so onboarding a new reviewer takes one click, not a permission matrix. Owners can change roles any time; admins can do everything except change ownership or billing.

Seat limits are part of your plan tier. The invite flow detects when you're at the cap and pops the upgrade dialog inline — no dead ends.

First-run experience used to drop new users into an empty dashboard. The new tour walks through the five rooms — audits, content, publishing, integrations, settings — and seeds a demo company so the dashboard is never empty on first load.

For users on a paid plan, the tour pulls their first real audit forward and finishes once they've seen the report. For free-trial users, the demo company gives them something to click before kicking off the real audit.

Numbers from internal pilot: median time from signup to first audit dropped from ~12 minutes to ~6 minutes.