Recommendations

Work the recommendation queues to turn what the scans and audits found into things you actually do.

Recommendations overview

Recommendations are split into three queues in the sidebar — Context, Technical and Content — with a count of open items next to each. There’s no combined list: you work one category at a time, because the three need different people and different amounts of effort.

Cards are generated from your data, not written by hand. They refresh after each scan and each audit, and the screen tells you when that’s happening rather than showing you stale cards.

Each card carries a priority band, an estimated impact and an effort label, and the queue arrives already ordered — so the top of the list is where to start.

This page covers what’s common to all three. For what lands in each queue, see Context, Technical and Content.

How to work a card

1. Click View details to open the card.
2. Read The problem — what was found, and where.
3. Read Why this matters — what it costs you to leave it.
4. Follow How to do it. Context cards link straight to the screen where the fix lives; Technical cards carry a Ready-to-use snippet with a Copy button; Content cards lead to a brief.
5. Set the status when you’re finished.

Back to recommendations returns you to the queue.

How to change a card’s status

The buttons on a card set where it stands:

• In progress — you’ve started. Use it so a teammate doesn’t pick up the same card.
• Done — finished.
• Skip — you’re not going to do this one. It leaves the queue without pretending to be done.
• Reopen — puts a closed card back.

You’ll also see Resolved on cards you never touched. That’s the app closing a card by itself because the next scan or audit no longer detects the problem — the usual outcome for a Technical fix you deployed.

Good to know A card you’ve started is protected. Once its status leaves Open, the regeneration that runs after each scan stops overwriting its contents, so the instructions you’re working from don’t change under you mid-task.

How to record what you published

Content cards close differently: they don’t resolve on their own, because nothing in a scan can tell that the article you wrote is the one that was asked for.

When you publish, mark the card Done. The dialog offers to record the link:

1. Paste the URL.
2. Choose its role — Original for the main version, Rewrite for a version adapted natively to another platform, Cross-link for a link to your original from an existing page.
3. Add more URLs if you published in several places, then save.

You can leave it empty and add links later. Adding one puts that page into your tracked sources, so future scans can tell you whether it’s being cited.

For the publications checklist and the briefs themselves, see Content briefs.

What the queue tells you when it’s waiting

The screen is explicit about why it has nothing for you, rather than showing an empty list:

• Before your first scan finishes — “Preparing your recommendations…”
• While a new scan is running — the queue refreshes automatically when it completes.
• While cards are being regenerated — “Updating recommendations for the latest data…”, usually under a minute.
• When there’s genuinely nothing — each category says so in its own words.

You’ll also see a note when there are more items waiting than the screen shows. They appear as you complete the ones in front of them, so the queue stays a to-do list rather than a backlog.

Good to know If your plan runs few samples per prompt, the screen warns that these recommendations are lower-confidence. The measurements behind them are noisier, and content briefs are the most affected.