Resource · citations

Where AI answers get their citations.

By Reviewed & updated Measurement-first: figures are median share-of-model with 95% confidence intervals. How we measure.

If you know which sources an engine trusts, you know where to be. Here's what the research shows about how AI answers are sourced - and what it means for your visibility.

The short version. Engines cite only a handful of domains per answer, split roughly evenly between your own pages and third-party listings - so being absent from the listing and review sites in your category quietly costs you the citation, even with the best product.

The reconciliation

Almost all AI citations are off your site - but most are within your control

Two well-cited studies measure different cuts of the same reality - they don't conflict. ~95% of AI citations are off your own website (Otterly.AI, 2025). And ~86% of AI citations are within your control or influence - 44% first-party site, 42% listings, 8% reviews/social, 6% news/forums (Yext (6.8M citations), Oct 2025).

~95%

off your own website (Otterly.AI)

~86%

within your control or influence (Yext (6.8M citations))

The reconciled takeaway. Almost all AI citations live off your own site (~95%, Otterly) - but most are within your control (~86%, Yext). Off-site isn't out of reach.

Our framework

The Clear Cited Citation Control Map

We brand the reconciliation as a 4-tier model and map each tier to a service - the differentiator is that we work all four tiers, not just the ~44% you own.

~44%

Full control

your site - owned pages, docs, benchmarks

we control · content + schema

~42%

Controllable

listings - G2, Capterra, directories

we control · listings

~8%

Influenceable

reviews & social

we influence · reviews + social

~6%

Earned

news & forums

we work (earned) · digital PR + community

Clear Cited works all four tiers

Coverage 100% · Control ~86% · Influence ~94% · Earned (best-effort)

Most AEO optimizes the ~44% you own. Clear Cited works all four tiers.

Control ~86% · influence ~94% · the last ~6% earned — worked, not guaranteed.

Source: Yext (6.8M citations), 2025-10 — a third-party citation-tier mix, not measured by our AI Visibility Index (which tracks share-of-model); illustrative of the tier split until we publish our own measured citation-source mix

Our honest coverage. We CONTROL ~86% (your site + listings), can INFLUENCE ~94% (adding reviews/social), and WORK the last ~6% (news/forums) through PR + community - influenced, never guaranteed. The differentiator: we work all four tiers, not just the ~44% you own. (Control vs influence are distinct - never ~100%, never guaranteed; the earned ~6% is worked, not promised. Our own niche breakdown is our own niche citation-source mix — illustrative until we publish our own measured citation-source data.)

Engines cite only 2-7 domains per answer

An AI answer doesn't link ten blue results - it cites a handful. LLMs reference only 2-7 domains, so the bar isn't "rank on page one," it's "be one of the few sources the model trusts enough to cite." (Source: LLM citation-pattern analysis, 2025.)

Structured data and freshness win citations

Structured data outweighs SERP snippets for extraction, and recency matters: roughly ~85% of AI-answer citations are under two years old and ~44% are from the current year, so content refreshed on a recurring cadence earns materially more AI citations than stale pages. Retrieval weights relevance, recency and trust - E-E-A-T is decisive. (Source: AI-citation recency study, 2025.)

What we do with this. It's why our work pairs owned answer-first pages (schema-marked, kept fresh) with earned placement on the listing and review sites engines already cite - and why retainers re-measure monthly rather than set-and-forget. See the full method on our methodology page.

See which sources cite you today.

A teardown shows where engines source your category - and where you're absent.

Get a free teardown

Keep reading

Methodologyhow we measure - 5 engines, adaptive per-engine runs, metric-matched confidence intervals + the 7-KPI modelHow we analyzethe intelligence loop - measure, attribute, learn, refresh, proveGlossaryplain definitions for every term we use

The methodology, as a film

How the measurement behind this actually works.

The measurement deep dive — sampling, intervals, reproducibility. (2:37) · captioned film — no audio by design · illustrative data, labelled in-film