Sources
When AI assistants answer with evidence, they point at URLs. Presence records every one of those sources so you can see which pages the models treat as trustworthy evidence - yours, your competitors', and third parties you may never have heard of.
The Citations screen is where you read them. Three words carry the whole subject:
- A is one event: one answer used one URL as a source. The same URL used by ten answers is ten citations.
- A site is the domain those pages live on -
g2.com,reddit.com, your own domain. - A page is one distinct URL -
g2.com/products/yourbrand/reviews.
A mention is being named in the answer. Being used as a source is being used as evidence. You can be mentioned without being used as a source, and used as a source without being the brand the assistant recommends first. Both matter.
Retrieved is not the same as cited
Assistants do two separate things with the pages they read. A citation is a page the answer visibly points to - the reference attached to a claim. But some assistants also disclose pages they retrieved while composing the answer and then never referenced.
The gap between the two is a real signal. A page that keeps being retrieved but never cited is being found and passed over, which is a very different problem from a page that is never found at all. The first is a content problem, the second is a discoverability problem, and they need opposite fixes.
Only Gemini and Perplexity publish retrieval data, so it appears only on the detail view of an individual response - the source lists, counts, and charts everywhere else count cited pages only. An assistant with no retrieval detail is not reading nothing; it just does not say what it read.
Why sources matter
- They show which of your pages earn attribution when assistants explain a topic.
- They reveal competitor and publisher pages that keep winning the evidence slot.
- They connect Presence (who gets credited) to website work (whether those pages are reachable, fresh, and structured for AI).
For a city government, sources might cluster on permit instruction pages and recycling schedules. For a university, on program pages and research summaries. For a corporate brand, on product specs and comparison guides.
Learn more
- Digital PR - earning mentions inside the third-party sites assistants already use as sources
- Press kit and newsroom - the on-site facts packet journalists and assistants use to describe you
- Wikipedia and Wikidata presence - encyclopedia sources when notability is real
- Affiliate and sponsored placements - buying disclosed inventory on those sites when it is worth it
- Guest-post link schemes - thin contributed posts bought mainly for links (discouraged)
- Parasite SEO - recognizing rented publisher subfolders that crowd the source set
- Review platform presence - legitimate review sites that dominate buying-intent sources
- Fake and incentivized reviews - when a competitor's rating spike is manufactured
- Community seeding - planted forum/Reddit "peer advice" (discouraged)
- Private blog networks - when sudden ranking authority comes from owned link networks
Related
- Citations screen - the report, gaps, and sites
- Traits - what was said, alongside where it came from
- Metrics at a glance
- Measuring visibility
- Presence and AEO