Skip to main content
AI engines cite pages when they answer your prompts. The domains they cite most are the ones shaping the answer. This recipe pages through every cited source for a brand, rolls them up by domain, and prints a leaderboard with your own domain highlighted.

What you need

How it works

  1. GET /projects/{project_id} returns the brand’s domain, so the script can spot your own pages.
  2. GET /projects/{project_id}/citations?limit=100 returns cited URLs, most cited first. The script follows next_cursor until it’s null.
  3. Each URL has times_cited (how often answers cited it) and times_retrieved (how often an engine’s web search retrieved it, cited or not). The script sums both per domain.

Code

Expected output

Read the results

  • Share is the domain’s portion of all citations for this brand’s prompts. A few domains with a large share are where you should aim to be cited, through PR, reviews or partnerships.
  • Retrieved much higher than cited means engines find that domain’s pages but rarely use them in answers. Compare their content with the pages that do get cited.
  • Your own rank is the quickest health check for content work. Track it over time by saving each run’s output.

Variations

Pass domain= to filter server-side, for example GET /projects/{project_id}/citations?domain=acme.com. The filter is a case-insensitive substring match, so it also matches subdomains such as blog.acme.com.
Filter with domain= set to your own domain and print each url with its times_cited. These are the pages AI engines already trust.
Each source has first_seen and last_seen. Keep only sources with a last_seen in the last 30 days to see who is cited now, not historically.
Last modified on November 18, 2026