AI Citation Tracking - See Which Sources AI Models Cite | Trakkr
When AI answers expose sources, track them.
Trakkr runs one controlled prompt set across all 8 models. Its citation table records the exact source links exposed by ChatGPT Search, Google AI Overviews, and Perplexity, alongside mentions and competitor answers from every model.
Citation share
18%+3 pts
Cited pages
23+4
Sources in category
212+11
Slots lost to rivals
6-2
| Source | Type | Citations | Share | Cited by |
|---|---|---|---|---|
| yourbrand.com | Your domain | 142 | 18% | ChatGPT Search, AI Overviews |
| g2.com | Review site | 118 | 15% | ChatGPT Search, Perplexity |
| reddit.com | Community | 96 | 12% | Perplexity, ChatGPT Search |
| competitor.com | Competitor | 84 | 11% | AI Overviews, ChatGPT Search |
| techradar.com | Publisher | 61 | 8% | AI Overviews, Perplexity |
| wikipedia.org | Reference | 47 | 6% | ChatGPT Search, Perplexity |
Why citations
Mentions and source links answer different questions.
Mention tracking records whether your brand appears in an answer. Citation tracking records the exact links a surface exposes beside that answer, including the domain, page, prompt, provider, and run date. A model may mention you without a link, or link to your page without recommending your brand.
Mention tracking vs citation tracking
| What it measures | Your name in answers | Exposed source links |
|---|---|---|
| What it tells you | Observed brand presence | Observed page-level evidence |
| The unit | A mention + position | A URL + prompt + provider |
| The action | Review the answer change | Inspect the page and source type |
| The limit | No mention is not universal absence | A link does not prove causation |
Prompt-to-citation workflow
- Freeze the prompt controls
Save the exact prompt, surface, market, language, and run date so the next result is comparable.
- Capture the answer
Record the full response, named brands, their order, and whether the surface exposes any source links.
- Normalize each source
Store the exact URL, domain, prompt, provider, first-seen date, and last-seen date for every exposed link.
- Classify and recheck
Separate owned, competitor, publisher, review, and community sources, assign one action, then rerun the same controls.
What Trakkr tracks
Six citation signals
The answer is what buyers see.
Cited domains
Every source domain exposed by the supported citation surfaces, ranked by observed link count across the controlled prompt set.
Your citation share
What percentage of observed source links point at pages you control, tracked over time without treating correlation as causation.
Competitor citations
The exact pages linked beside answers that name competitors, ready for a page-level source-gap review.
Page-level performance
Which of your URLs are linked, on which prompts, and by which citation-capable surfaces, so you can choose what to investigate.
Source-type mix
Whether observed source links point to reviews, communities, publishers, or brand domains. Use the mix to frame the next investigation.
Citation alerts
Notifications the day you lose a citation slot, win a new one, or a new source starts speaking for your category.
How each model cites
No two surfaces expose sources the same way
Source behavior decides what can be measured on each platform. Each guide below covers one model’s mention and citation mechanics in depth.
How it works
- Add your brand
Enter your domain. Trakkr generates a buyer-intent prompt set for your category and identifies your competitors automatically.
- Collect answers and source links
Trakkr runs the fixed prompt set across all 8 models. The citation table normalizes exposed source URLs from ChatGPT Search, Google AI Overviews, and Perplexity.
- Win the slots that matter
Citation gaps become a ranked to-do list: pages to fix, sources to earn placements on, and alerts when a slot changes hands.
FAQ
What is AI citation tracking?
AI citation tracking means monitoring the source links an AI surface exposes when it answers questions in your category: which domains, which pages, and how often yours appear in the controlled prompt set. Those links are observable evidence, but they do not reveal every retrieval candidate or prove why a model produced an answer.