Citations & Sources

Most AI answers are retrieval-shaped: the assistant searches, reads a handful of pages, and writes from them. The pages it read are the lever. Change what those pages say about you and the answer changes; argue with the answer directly and nothing happens.

Every probe therefore stores the full citation list of its answer, and the panel aggregates them.

Collection

Citations are captured verbatim from the engine’s response — URL plus title where one is given. They are stored per probe, so you can always trace a domain back to the exact conversation that cited it.

During grading, each cited URL is also classified by the analysis model into one of three raw categories:

Model category	Meaning
third_party_blog	Third-party blogs, review sites, listicles (“Best X”, “Top 10 X”)
brand_website	Official brand or vendor websites
community_discussion	Reddit, Quora, forums and similar

The three buckets

For display, domains are folded into three buckets by a hybrid rule that puts deterministic checks ahead of the model’s judgement:

1. Your own domain → Owned Assets Matched on hostname, including subdomains. This one is never left to a model.
2. A known social / UGC platform → User Consensus Reddit, YouTube, LinkedIn, Quora, X, TikTok, Medium, Substack, GitHub, Stack Overflow, Trustpilot, G2, Capterra, Product Hunt and similar.
3. Otherwise, use the model's category brand_website → Owned Assets, community_discussion → User Consensus.
4. Everything else → Authority Media Including third_party_blog and anything unclassified.

Bucket Authority Media News, magazines and third-party review sites. Earned through PR, review placements and being covered — the slowest to move and the most durable.

Bucket Owned Assets Your own site and domains. The one bucket you control outright. A low share usually means your pages don't answer the questions being asked.

Bucket User Consensus Reddit, YouTube, LinkedIn and other social/UGC platforms. Heavily retrieved by current engines, and where absence is most visible.

A healthy mix has all three. All-owned means engines only find you when they’re already looking for you; all-consensus means the narrative is being written by other people.

Cross-engine sources

The Cross-Model Cited Sources view lists domains cited by more than one engine in the same run.

These are the highest-leverage targets in the whole product. A page that ChatGPT, Gemini and Perplexity all read is a page where one change propagates to three engines at once — whereas a domain cited by a single engine may just be that engine’s retrieval quirk.

Shared vs region-specific sources

When a run covers multiple markets, cited domains are also split by where they were cited, using geo-adjustable engines only:

Group	Meaning	Usual response
Cited in multiple regions	Global leverage — one page, several markets	Prioritise; the return is multiplied
Cited in one region only	Local content opportunity	Local press, local marketplaces, local communities

A market where your competitors appear and you don’t, backed by region-only sources you have no presence on, is the most concrete finding the region split produces.

Source depth by plan

How many citation rows are surfaced per probe is a plan limit — the Free and Lite tiers surface a capped number of rows, and higher tiers give full traceability. See Plans & usage.

Turning sources into action

What you see	What it usually means
A listicle cited by several engines, and you’re not in it	The single highest-value placement available to you
Reddit threads dominating consensus, no brand presence	Your category is decided in community discussion
Your own domain barely cited	Your pages don’t answer the prompts being asked — see Content
Competitor’s domain cited on your category prompts	They’ve published the definitive page for that question

The Sources tab exports to Excel, so the domain list can be handed to a PR or content team as a work queue.

Competitors → Who is winning the answers these sources feed. Distribution → Placing content on the sources engines already read.