Recommendations: the project's action plan

The Recommendations page (Home menu) measures nothing new. It re-reads what the project has already gathered — the SEO × GEO matrix, mentions, scan history, Search Console, rankings and Google Analytics 4 — and turns it into a list of things to do. The analysis is rebuilt after each scan and then served from cache; only your « completed » and « not relevant » statuses are re-read every time you open the page.

The seven blocks

Every block folds and unfolds. On load, only the summary, the GEO score and the action plan are open.

• Monthly summary: a dated write-up, clickable key-figure tiles (each one opens the page that details it) and three shortcuts — View full report, Activity timeline, Active alerts.
• GEO optimization score: a mark out of 100 and a letter from A to F, broken down by category (AI visibility, share of voice, Google presence, organic traffic, keyword coverage). Every weak category shows the button that takes you to the matching page.
• Priority action plan: actions sorted by priority, each with its category, its impact and the effort involved, then the concrete steps to take.
• Detected opportunities: gains spotted in your own data, each with its Action: line.
• Content suggestions: format and priority for each topic, with the Create this content (PivotSEO), NeuronWriter (when enabled on the project) and Keyword research buttons.
• Competitive analysis: for each competitor, their mentions against yours, the engines they show up on, their sentiment, the Google positions you are fighting over, and counter-actions.
• GEO-optimized FAQs: questions written to be picked up by AI engines, with a Copy the full <script> block button that outputs ready-to-paste FAQPage markup.

Marking a recommendation as done

Every recommendation carries two buttons. That is what separates this page from a list of tips: it keeps track of the work.

• Completed: the recommendation leaves the list.
• Not relevant: it leaves the list too, but without counting as work done.
• The Show completed / dismissed items checkbox at the top right brings them back; a Reopen button undoes the decision.
• Marking something Completed creates a dated event on the Impact proof page; reopening it removes that event.

The buttons take you to the right page

Each recommended step is wired to the screen that lets you carry it out, with the project — and, where it matters, the keyword or the competitor — already loaded: schema checking, page audit, content brief, sources and citations, rank tracking, simulator, alerts or gap analysis. You never have to go hunting for the page yourself.

Cross-cutting recommendations

On top of the base engine, the server adds actions drawn from the caches of other pages — with no external call. They therefore only appear once the matching data has been collected somewhere in the project.

• Clicks to defend when Google answers with an AI Overview on queries you rank for.
• Cannibalizations to resolve as a pillar-and-spokes cluster, and queries waiting in the content queue.
• Reddit threads in your market worth joining — Reddit is heavily cited by ChatGPT and Perplexity.
• AI crawlers blocked in robots.txt, or content invisible without JavaScript.
• No Wikidata entity, no YouTube presence, and visits already coming from AI engines that deserve follow-up.

If the page stays empty

Everything comes from the project's own data: without at least one completed scan there is nothing to recommend. A brand-new project, or one whose scan failed, shows a blank page rather than an error message.