Citations and the advisor

Citations

In grounded mode the assistants search the web before answering and their answers carry the sources they used. Every citation is stored with the run: URL, domain and the position it appeared in. The “Where the AIs look” card aggregates them by domain across your prompts — the review sites, directories, comparison pages and publications the assistants trust for your market.

This is the part of the metric you can act on: get covered on the domains that already shape the answers and the grounded score follows. Citations also feed the competitor view — the domains that cite the brand ranked first are the map of where you are not.

The advisor

Starter and Pro projects get the GEO advisor: for the prompts where your brand is absent or ranked low, an agent reads the winning answers, their citations and your own site, and produces a short list of concrete recommendations — the comparison page to publish, the directory to get listed on, the review site to pursue, the llms.txt to add, the claim on your homepage the assistants keep missing. Evidence-based: every recommendation points at the answers and sources that motivated it.

The advisor runs on demand from the dashboard (one run at a time per project) and its output is kept with the date, so you can measure what a recommendation did.