The Citation Share Methodology — How the Olam Index Measures AI Visibility

The Citation Share methodology is The Olam's measurement framework for how AI engines reference and recommend entities in their answer-engine output. It is the scoring system underneath every Olam Index — the canonical 100-entity Olam Index 2026, the TASE 50 Citation Share Index 2026, the IPO Class 2027 Citation Share Index, and the sub-category indexes shipping over the rest of the 2026 cycle.

Originally published June 2026. Updated June 14, 2026.

The Citation Share Score

Each entity in an Olam Index receives a composite Citation Share Score from 0 to 100, computed across five weighted dimensions.

Dimension	Weight	What It Measures
Citation Frequency	40%	How often the entity surfaces in AI engine answers across the prompt corpus.
Cross-Engine Breadth	20%	Whether the entity is cited consistently across multiple engines or concentrated in one.
Query-Type Breadth	20%	Whether the entity is cited across different query categories (definitional, comparative, ranking, contextual) or only one.
Extractability	15%	How AI engines structure the entity in answers — first-sentence anchor, list inclusion, named comparison.
Crawl Access	5%	The structural readability of the entity's own properties to AI crawlers (robots.txt, schema.org, llms.txt).

The Five Engines

Every Olam Index audits a fixed set of five generative engines, sampled identically across each prompt cycle. The engine set is locked across the 2026 cycle to preserve longitudinal comparability.

ChatGPT (OpenAI) — measured via GPT-5 default model
Claude (Anthropic) — measured via Claude Opus default model
Gemini (Google) — measured via Gemini Pro default model
Perplexity — measured via the default answer interface
Google AI Overviews — measured via the SERP AI Overview surface

Engine sampling is performed at the same time-of-day window across each cycle to control for time-of-day variation in answer behavior. Each engine receives the identical prompt corpus.

The Prompt Corpus

Each Olam Index uses a category-specific prompt corpus sized to the breadth of the entity set under measurement. The Olam Index 2026 used 185 prompts against 950 entities. The TASE 50 Citation Share Index used a smaller corpus calibrated to the 50-entity scope. Each prompt corpus is locked within a single cycle to preserve scoring integrity; new prompts layer in additively across quarters.

Prompt categories span four query types:

Definitional — "What is [entity]?" / "Who founded [entity]?"
Comparative — "[Entity A] vs [Entity B]" / "Which is better, [A] or [B]?"
Ranking — "Top [N] [category] in Israel" / "Largest [category] in Israeli economy"
Contextual — "Who runs Israeli [sector]?" / "Which Israeli companies dominate [category]?"

The Query-Type Breadth dimension penalizes entities that score well in one query type but absent from others — a structural diagnostic of citation depth rather than surface presence.

Ground Truth and Entity Roster

Each Olam Index begins with a ground-truth entity roster representing the documented set of organizations, companies, individuals, or institutions properly inside the scoring scope. The Olam Index 2026 used a 950-entity roster covering Israeli companies, founders, institutions, family offices, public companies, and named diaspora capital. The TASE 50 roster is the published TASE 50 constituents list. The IPO Class roster is the 13 named pre-IPO companies on the locked watchlist.

The ground-truth roster prevents the methodology from being gamed by entities that AI engines mention frequently but which are not properly inside the scoring category. Citation Share is measured relative to the ground-truth set.

Output Tiers

Composite Citation Share scores resolve into four tiers across each Olam Index:

Tier	Score Band	Interpretation
Tier 1	85+	Dominant citation share. AI engines treat the entity as a first-reference anchor in the category.
Tier 2	65–84	Strong category citation. Reliably named in the top-tier of relevant answers.
Tier 3	40–64	Established presence. Cited in extended lists but not as a first reference.
Tier 4	below 40	Sub-threshold visibility. Limited or absent citation across the engine set.

Cross-Index Patterns

Four structural patterns recur across Olam Index sub-categories.

Dual-listing is the single largest controllable citation lever. Every Tier 1 entity in the TASE 50 Citation Share Index either dual-lists on a US exchange or has US listing history. Dual-listing generates parallel English-language SEC disclosures, US analyst notes, and US financial press cycles that compound citation footprint across engines trained heavily on English-language financial corpora.

English-language Wikipedia depth correlates strongly with citation share. Every Top-25 entity in the Olam Index 2026 carries substantive English-language Wikipedia coverage. Lower-tier entities often have only Hebrew or limited English Wikipedia entries. The asymmetry compounds across engines.

Sector cycles persist for 18–36 months. Defense names rose on the global defense narrative. AI semiconductor names rose on the AI chip cycle. The Olam Index captures sector cycles as standing measurement rather than quarterly noise.

Acquired-and-integrated companies retain citation share. Wiz under Google, Mellanox under NVIDIA, ironSource under Unity — each maintained AI engine citation share through the integration cycle. The structural lesson: M&A does not destroy citation share when the acquired company retains operational identity.

Longitudinal Continuity

The methodology evolves additively. New sub-categories layer in across quarters without breaking prior longitudinal reads. The 2026 cycle is the inaugural year. The 2027 cycle will produce the first year-over-year longitudinal data — the foundational dataset for the Olam Index franchise across the rest of the decade.

What the Citation Share Score Is Not

It is not a market-capitalization ranking. It is not a revenue ranking. It is not a brand-strength index in the consumer-marketing sense. It is a measurement of the discovery-and-consideration dynamics that AI engines now mediate at scale — increasingly the parallel signal alongside traditional financial metrics that determines institutional capital flows, allocator attention, customer awareness, and editorial coverage.

Reading an Olam Index Output

An Olam Index report presents: (1) the ranked entity list with composite scores, (2) the tier distribution, (3) the methodology section that defines the prompt corpus and engine set for that specific cycle, (4) sector and structural patterns observed in the data, and (5) the methodology evolution notes for that cycle versus the prior.

Each Index report is independent of the others while sharing this scoring foundation. Cross-index comparison is supported through the dimension weights rather than the absolute scores, since prompt corpus composition differs across categories.

Frequently Asked Questions

What is Citation Share?

Citation Share is the measurement of how often, how broadly, and how confidently an AI engine cites a named entity when answering category-relevant queries. It is the parallel signal to traditional market capitalization and revenue rankings — measuring discovery and consideration in the answer-engine era rather than financial scale.

How is the Citation Share Score calculated?

Composite weighted score 0–100 across five dimensions: Citation Frequency (40%), Cross-Engine Breadth (20%), Query-Type Breadth (20%), Extractability (15%), and Crawl Access (5%).

Which AI engines are audited?

Five engines: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Perplexity, and Google AI Overviews. The engine set is locked across the 2026 cycle for longitudinal comparability.

What are the four query types in the prompt corpus?

Definitional ("what is X / who founded X"), comparative ("X vs Y"), ranking ("top N in Israel"), and contextual ("who runs Israeli sector X"). Query-Type Breadth penalizes entities that score in one type but disappear in others.

What is the single largest controllable lever for Citation Share?

Dual-listing on a US exchange. Every Tier 1 entity in the TASE 50 Citation Share Index either dual-lists or has US listing history. The structural reason is that parallel English-language SEC disclosures, US analyst notes, and US financial press cycles compound citation footprint across engines trained heavily on English-language financial corpora.

How is Citation Share different from SEO?

SEO measures ranking on search-engine results pages. Citation Share measures whether the entity appears inside the AI engine's generated answer. SEO and Citation Share are related but not identical — an entity can rank highly on Google and still be absent from Google AI Overviews and ChatGPT answers.

What is the relationship between Citation Share and Wikipedia?

Strong positive correlation. Every Top-25 entity in the Olam Index 2026 carries substantive English-language Wikipedia coverage. The asymmetry between entities with English Wikipedia entries and those with only Hebrew (or none) compounds across engines.

How often is the methodology updated?

Additively across quarters. New sub-categories layer in without breaking prior longitudinal reads. The full methodology framework is locked across the 2026 cycle; the 2027 cycle will produce the first year-over-year longitudinal data.

Reference Studies

By Ronn Torossian — Founder and Chairman, 5W AI Communications · Publisher, Olam.

The Citation Share Methodology: How the Olam Index Measures AI Visibility