Citations and sources
Where citation data comes from, the difference between quoted and listed, and how sources roll up to domains.
Citations are the evidence trail behind an answer: the sources an engine displayed when it wrote one. They are the most actionable data in the platform, because a citation is a page you can influence.
Where the data comes from
Citations are read from the captured answer itself — the sources the engine actually showed, in the order it showed them.
Geonimo does not scrape cited pages to infer citations, and does not ask a model to guess which sources it used. Both are slow and unreliable; the rendered answer is neither.
Coverage varies by engine because each surface displays sources differently: AI Overviews carries a source panel plus inline references, Perplexity cites natively, ChatGPT differs again. Compare citation data within an engine over time rather than across engines in a snapshot.
Quoted versus listed
The distinction the citation data is built around.
- Quoted (inline) — the source was cited against a specific claim in the answer body. The engine attributed a sentence to that page.
- Listed — the source appeared in the panel beside the answer. The engine is saying it read the page.
These are not the same achievement, and they barely overlap in practice — measured across captured AI Overviews, on the order of 9 panel URLs against 134 inline references in the same corpus. Collapsing them into one list throws away most of the signal.
Being quoted is what moves an answer. Being listed is worth having and is not the same thing.
Each citation also carries its rank — the order it appeared in — and a source can be both quoted and listed.
Sources and domains
Citations normalize into two levels:
- Source — a specific URL, with its title and when it was first and last seen in your answers.
- Domain — the site it belongs to, with its type (Editorial, Directory, Community, Docs, Blog, Marketplace).
Domains tell you where influence concentrates and drive outreach strategy. Sources tell you which individual page is doing the work, which is what you pitch, correct or compete with. Both views are in Sources.
Citation index
The Citation Index on the Overview measures whether the pages engines read actually name each tracked entity: per day, the share of read-page citations landing on pages whose fetched body text names them.
Read from body text, not titles. A headline says what a page is filed under; the body says whether it talks about you, and the pages that decide answers ("The 10 Best Running Shoes of 2026") name nobody in their headline. Matching is conservative — only occurrences that positively read as the brand count, so a page that contains the word without meaning the company does not.
It is fed by pages Geonimo actually fetches. After each day's collection, cited pages are read most-cited first and top pages are re-checked every 14 days. Reading is paid for a page at a time, so the read set always trails the cited set — which is why the panel states its own coverage, N of M cited pages read, rather than letting a share imply a census.
Two population rules keep the number honest:
- Visibility prompts only, matching every other Overview metric. A perception probe names the brand, so the material read for it names the brand by construction; counting it would move the index with the prompt mix rather than with anything published.
- A fixed 30-day window, unlike the rest of the Overview, because a shorter slice would sample the reading schedule rather than the answers.
Read it against visibility:
| Pattern | Reading |
|---|---|
| High visibility, high citation index | Strong. Engines name you, and the material they read names you too. |
| High visibility, low citation index | Fragile. You are being named from the engine's memory while the pages behind the answers ignore you — a position nothing published is holding up. |
| Low visibility, high citation index | The material names you but the answers do not. The gap is the association engines make, not the content. |
| Low, low | Start with Pages. |
Making your pages citable
Citations follow structure. Properties that recur on pages engines quote:
- A direct, self-contained answer near the top — engines extract passages, not documents
- Clean heading hierarchy so passages have boundaries
- Structured data where it applies, especially FAQ markup on question-shaped content
- Explicit comparisons that name alternatives, for comparison prompts
- Current dates and current facts
- Content in the HTML, not assembled by client-side script
Third-party citations
Most categories are decided partly by pages you do not own. That is not a failure of your content — it is how buyers behave, and engines follow.
Two responses, both legitimate:
- Earn presence on the domains that already decide your category. PR Opportunities ranks them from your own citation data.
- Correct what is wrong on the pages already cited. A stale price or a wrong feature claim on a widely-read directory costs more than a missing blog post, and takes an email to fix.