Your dashboard tells you that Nike isn't being recommended for marathon shoes. Ranking advice tells you what to do about it.
It's an agent, not a report. It reads the citation data for one prompt, searches for content you've already published, fetches real pages, and returns a prioritised list of specific changes.
What you get back
A short summary, then a handful of recommendations in priority order — ordered by likely impact rather than by how easy they are.
Each recommendation carries:
What to do — a specific action, like improving an existing page, adding a comparison, or getting listed in a roundup
The topic it applies to
Why, in a couple of sentences grounded in what it actually found
Supporting evidence — the pages it looked at
Difficulty — low, medium or high
How it works, and why that matters
Three behaviours make the advice more useful than a generic checklist.
It looks for your existing content first. Before suggesting you create anything, it searches your site for pages already covering the topic. Improving a page you have usually beats writing a new one, and the agent is built to prefer that.
It checks before recommending outreach. If a roundup article looks worth being in, it fetches the page to see whether you're already listed. That avoids the classic wasted recommendation of pitching a publication that already covers you.
It won't send you to a competitor. Recommendations to earn a mention never point at a domain belonging to a brand you're tracking as a competitor.
It also weighs its suggestions by what actually drives citations — earned mentions, expert quotes, statistics, answer-first structure, freshness — and treats technical markup as hygiene rather than a growth lever. If you were expecting schema advice at the top, that's deliberate.
Before you can run it
The agent needs a certain amount of citation data on the prompt before it can say anything specific. With too little, its tools come back empty and the advice degrades into "create more content", which helps nobody.
The panel tells you how many citations the prompt has and how many are needed. If you're short, keep the prompt tracking and come back after a few more checks.
Ranking advice is available on paid plans, and it costs credits — the price is printed on the Get advice button, and there's no separate confirmation step. See How credits work.
How long it takes
Often under a minute. On prompts with a lot of sources to check it can take several minutes, because it's genuinely fetching and reading pages. You can leave the page and come back — the run continues, and we email you when it's done.
Running it again
If nothing has changed in the underlying data, running again returns your previous advice rather than charging you for the same analysis twice. You'll see a note saying so.
New citations on the prompt are what make a fresh run worthwhile — so the natural rhythm is to run it, do the work, and come back after the next few rounds of checks.
Reading it well
Start at the top and take one recommendation seriously rather than skimming all of them. They're ordered by impact, and the first one is where the agent thinks your leverage is.
