Recommendations

A recommendation is a specific, actionable change with a predicted visibility uplift attached. Lokrix derives each one from the feature-level reasons behind your score, then ranks the whole list so you always know the single next thing worth doing.

How recommendations are generated

The predictive scoring model exposes which features hold a page back for a given prompt and engine. The recommendation engine maps those features to concrete interventions — add a cited statistic, earn a corroborating mention, improve freshness — and runs each candidate through a what-if simulation to estimate its effect. Recommendations are always tied to real signals; Lokrix never emits a generic checklist item it can’t attribute to your data.

The impact / effort matrix

Every recommendation is placed on two axes: predicted impact (how much score and presence it is simulated to add, with its confidence) and estimated effort. The top-left quadrant — high impact, low effort — is where you start.

Anatomy of a recommendation

• The change. A concrete instruction — what to add or edit, and where.
• Predicted uplift. The simulated change in your Lokrix AI Score and per-engine presence, shown with a confidence interval — never a bare number.
• Affected prompts & engines. Which of your prompt universe entries and which engines the change is expected to move.
• Evidence. The features and competitor gaps the recommendation is grounded in.

From a list to a plan

A ranked list tells you what to do next; a sequenced plan tells you how to reach a goal. Feed prioritised recommendations into a path to #1 for a target prompt, or preview any single item first with what-if simulation.