Skip to content

Module 6 min

Why a Source Gets Cited

Relevance, authority, freshness and extractability multiply into citation odds, and a weak score on any one collapses the rest.

57%
In contention

Weak on any one factor and the odds collapse. Engines gate on the lowest link, not the average.

+40% visibility lift from adding statistics/citations/quotes (Princeton GEO study)
4.2× more likely to be cited when semantic completeness scores ≥8.5/10
96% of AI Overview citations come from sources with strong E-E-A-T signals

Why it mattersThis is the closest thing GEO has to a ranking function, and it is gated, not additive.

01

Definition & Foundation

What it is, in plain words

Across studies of AI Overviews, ChatGPT, and Perplexity, citation behavior converges on four factors: relevance (the passage matches the question), authority (the source is machine-verifiably trustworthy), freshness (the content is current), and extractability (the claim can be lifted cleanly). This is the closest thing GEO has to a ranking function, with one crucial twist.

The twist: the factors are gated, not additive. A brilliant, authoritative page that is hard to extract still loses; a fresh, extractable page from a no-name source still loses. You cannot compensate for a zero; you must clear every gate. The good news is that the levers are proven: the Princeton GEO study found that adding statistics, quotations, and cited sources lifts a page's visibility in generative answers by up to 40%, and passages scoring high on semantic completeness are 4.2× more likely to be cited. Fix the failing gate first; then stack the multipliers.

02

The Four Gates

Clear all four or the citation goes elsewhere

gate 1

Relevance

Your passage must semantically match the question, not the topic, the question. Pages "about" a subject without a passage that answers the specific prompt fail here quietly. Fixed by answer-first sections targeting real prompts.

gate 2

Authority

Roughly 96% of AI Overview citations come from sources with strong E-E-A-T signals: named authors, recognized entities, corroborating mentions. Trust the machine can parse, not prose that claims it.

gate 3

Freshness

Engines bias toward recently published or genuinely updated content, hardest on topics where answers change. A two-year-old page competes at a structural discount, whatever its quality.

gate 4 · most often failed

Extractability

The claim must survive being lifted: self-contained, answer-first, one claim per passage, ideally carrying a quotable statistic. This is the gate great content fails most, and the cheapest one to fix.

03

Myths vs Reality

Common misreadings, corrected

Myth"Strong authority compensates for weak structure; the factors average out."
RealityThey gate. Citation studies keep finding authoritative pages that never get cited because no passage can be lifted cleanly, and perfectly-structured pages cut for having no parseable trust signals. Diagnose the failing gate; averaging is how effort gets wasted on the wrong fix.
Myth"Each engine cites the same kind of sources; optimize once, win everywhere."
RealityOnly about 11% of cited domains overlap between ChatGPT and Perplexity for the same query, and 71% of cited sources appear on only one engine. The gates are shared; the winners aren't. Track engines separately.
04

Putting It to Work

Audit the gates, fix the failing one, stack the multipliers

This node turns into a repeatable per-page audit. For every page that should earn citations:

The four-gate audit

1

Score each gate red / yellow / green

Relevance: is there a passage answering the exact target prompt? Authority: named author, entity signals, outside mentions? Freshness: substantively updated in 12 months? Extractability: does the key passage survive out of context? Ten minutes per page.

2

Fix the reddest gate first

Gating means the weakest factor sets your ceiling. A page with three greens and one red gains more from fixing the red than from polishing the greens; resist the urge to do the comfortable fix instead of the failing one.

3

Stack the proven multipliers

Once no gate is red, add what the Princeton study validated: a sourced statistic, an expert quotation, and cited references in the passage itself, the up-to-40% visibility lever. Concrete numbers beat adjectives everywhere in GEO.

4

Verify per engine, not in aggregate

With ~11% citation overlap between engines, "are we cited?" is engine-specific. Track your target prompts on AI Overviews, ChatGPT, and Perplexity separately; a win on one says little about the others.

05

Verification Checks

How to know it's really done

0/4 verified

The ranking function is working for you when audits replace guesswork. Click a check to mark it verified:

06

Works Together With

The nodes this one leans on

Tools for this

Go deeper

Always current

These links resolve live: what you get is generated or filtered the moment you click.

Sources