Sources overview
Which domains AI cites when answering prompts about you and your space.
Last updated: 2026-05-05
Sources overview
When AI models cite the web in their answers, those citations are gold. They tell you exactly which domains the AI trusts in your category — and therefore exactly where you need to be published, mentioned, or quoted to influence future answers. The Sources page surfaces every cited domain across your scans with frequency, influence, and competitor context.
How sources are extracted
After each AI Scan, Pranas:
- Parses every response for URLs and citation markers.
- Strips hallucinated domains using a multi-layer validator (fail-strict — if a domain doesn't resolve or doesn't actually appear in the surrounding text, it's rejected).
- Normalises each URL to its root domain (e.g.,
g2.com/articles/best-crm→g2.com). - Persists the cleaned source list to a dedicated table so the same numbers appear in the UI and in PDF exports.
Hallucinated URLs are common in older models — Pranas's validation prevents them from polluting your source intelligence.
Extraction pipeline reference
| Step | What happens | Outcome |
|---|---|---|
| Parse | Every response is scanned for URLs and citation markers | Raw candidate list |
| Validate | Fail-strict check rejects domains that don't resolve or don't appear in surrounding text | Hallucinated URLs removed |
| Normalise | URLs collapse to root domain (g2.com/articles/best-crm → g2.com) |
One row per domain |
| Persist | Cleaned list is written to a dedicated table | UI and PDF exports match exactly |
Column reference
| Column | What it tells you |
|---|---|
| AI Citations | How many times the domain was cited in scope |
| Influence score | Outreach priority, 0–100 |
| Model chips | Which models cited it; all-model coverage is most reliable |
| Brands mentioned | Which brands that domain tends to name |
What you'll see on the page
- Total citations across the active scope
- Unique domains count
- Top sources table ranked by citation count, with influence score, model coverage, and the brands each domain tends to mention
- Domain favicon and chip for fast visual scanning
- Model chips showing which models cited each domain (a domain cited by all models is more reliable than one cited only by Perplexity)
How to use it
- Pitch list builder. Sort by citation count, filter to "doesn't mention us yet", and you have a target list for PR and link-building.
- Competitor reverse-engineering. Click a competitor in the Top Brands widget to see which sources cite them — that's their distribution channel.
- Content opportunities. A domain that cites the category often but cites no specific brand is an opportunity to be the brand they cite next.
- Citation defence. A domain that used to cite you but no longer does is an early warning of relevance decay.
Filters
The page respects the global filter bar (date range, country, models, prompt type). Source data scales with scan volume — small prompt libraries produce sparse source data.
Related articles
- Source influence scoring
- Citation optimization tactics
- Top Brands widget