Skip to content

Concept · GEF

Generative Engine Friendly

Generative Engine Friendly (GEF) is the measurable degree to which a page or brand can be accessed, understood, trusted, and cited by generative engines — quantified 0–100 by the GEF Score.

04
Dimensions
Accessibility · Extractability · Authority · Citability.
16
Signals
Pre-registered on-page signals.
0–100
Scale
A predictor of citation rate.

Generative Engine Friendly (GEF) is the measurable degree to which a page or brand can be accessed, understood, trusted, and cited by generative engines — ChatGPT, Claude, Gemini, Perplexity — when composing their answers.

Where SEO produced search-engine-friendly pages to rank, GEO produces generative-engine-friendly pages to be cited. The GEF Score quantifies that property on a 0–100 scale, computed from the page itself — a predictor of AI Visibility (citation rate), not a record of it.

I.

The four GEF dimensions

Composition

What the GEF Score measures

Four weighted dimensions, built from 16 pre-registered on-page signals. Accessibility acts as a gate — a page a bot cannot read caps the whole.

  1. Evidence block 20%

    Accessibility

    The engine can reach and parse the page: AI-bot access (GPTBot, ClaudeBot, Google-Extended, PerplexityBot), HTTPS, server-rendered or static HTML, valid JSON-LD, correct canonical, sane robots.

    • S1 — JSON-LD @graph completeness
    • S2 — llms.txt presence & validity
    • S6 — Canonical URL correctness
    • S7 — AI user-agent crawlability (gate)
    • S9 — BreadcrumbList presence
    • S15 — Image alt-text coverage
    • S16 — HTTPS & Core Web Vitals

  2. Evidence block 30%

    Extractability

    Content is structured for atomic extraction: a direct-answer paragraph early, a regular heading cadence, appropriate length and reading level — passages a model can lift whole.

    • S3 — Direct answer within first 120 words
    • S4 — Heading every ≤ 300 words
    • S13 — Length 300–2,500 words
    • S14 — Reading-level accessibility

  3. Evidence block 25%

    Authority

    E-E-A-T plus verifiable author and entity identity: byline markup, sameAs entity links, freshness, internal link importance — reasons for the engine to trust the source it is about to quote.

    • S8 — sameAs entity links
    • S10 — Author / byline structured data
    • S11 — Freshness (dateModified, lastmod)
    • S12 — Internal linking density

  4. Evidence block 25%

    Citability

    The page carries citable units — FAQ / HowTo structured data, definitions, statistics, named entities — ready to be lifted into an answer.

    • S5 — FAQPage / HowTo structured data

II.

The GEF Score in practice

illustrative

82/ 100

GEF Score

Illustrative breakdown — not a measurement of this page.

  • Accessibility20%95

  • Extractability30%88

  • Authority25%70

  • Citability25%75

How the GEF Score is computed

The score is computed from the page itself, never by querying an engine. Each of the 16 pre-registered on-page signals (OSF: 10.17605/OSF.IO/XCG7J) scores on the registered 0 / 0.5 / 1 rubric; each dimension is the registered-weighted average of its signals, and the four dimensions combine by weight — with Accessibility acting as a gate:

based=100×∑(ωi⋅gi)∑ωi(gi∈{0,0.5,1},  ωi=registered weight)Subd=clamp(based,0,100)GEF=20⋅SubAcc+30⋅SubExt+25⋅SubAuth+25⋅SubCit100×gate(SubAcc)\begin{aligned} \text{base}_d &= 100 \times \frac{\sum (\omega_i \cdot g_i)}{\sum \omega_i} \quad (g_i \in \{0, 0.5, 1\}, \; \omega_i = \text{registered weight}) \\ \text{Sub}_d &= \text{clamp}(\text{base}_d, 0, 100) \\ \text{GEF} &= \frac{20 \cdot \text{Sub}_{\text{Acc}} + 30 \cdot \text{Sub}_{\text{Ext}} + 25 \cdot \text{Sub}_{\text{Auth}} + 25 \cdot \text{Sub}_{\text{Cit}}}{100} \times \text{gate}(\text{Sub}_{\text{Acc}}) \end{aligned} gate(SubAcc)={1if SubAcc≥60SubAcc60if SubAcc<60\text{gate}(\text{Sub}_{\text{Acc}}) = \begin{cases} 1 & \text{if } \text{Sub}_{\text{Acc}} \ge 60 \\ \frac{\text{Sub}_{\text{Acc}}}{60} & \text{if } \text{Sub}_{\text{Acc}} < 60 \end{cases}

A hard floor is applied: if all four AI bots are blocked AND there is no SSR/static HTML, the score is capped at ≤25\le 25 ("GEF-blocked").

A page AI bots cannot read caps the whole score — which is why Accessibility gates the rest.

Common questions about GEF

What is Generative Engine Friendly (GEF)?

Generative Engine Friendly (GEF) is the measurable degree to which a page or brand can be accessed, understood, trusted, and cited by generative engines — ChatGPT, Claude, Gemini, Perplexity — when composing their answers.

What is the GEF Score?

The GEF Score quantifies generative-engine-friendliness on a 0–100 scale across four dimensions — Accessibility, Extractability, Authority, and Citability — built from the 16 pre-registered on-page signals (OSF: 10.17605/OSF.IO/XCG7J). It is computed from the page itself, never by querying an engine, so it is a predictor of citation rate, not a record of it.

How is the GEF Score different from an AI Visibility score?

Most AI scores measure realized presence — how often you are actually cited. The GEF Score measures the property instead: whether a page can be, and deserves to be, cited, computed without querying any engine. AI Visibility (citation rate) — the proportion of LLM answers that cite your domain — is the separate outcome the Score is built to predict.

How do I improve my GEF Score?

Raise the weakest dimension first — Accessibility acts as a gate, so a page a bot cannot read caps the whole. Ensure AI-bot access and server-rendered HTML, structure content into self-contained answer-first passages, strengthen author and entity authority, and add citable units like FAQ blocks, definitions, and statistics. Prime Sentia audits each of the 16 pre-registered signals and prioritizes the fixes that move the number most.

Related concepts

Get your GEF Score

Run a free audit to score your pages across the four GEF dimensions and the 16 pre-registered signals — with the fixes that move the number most.

Run a free audit