§ VOLUME I · ARTICLE III · LEDGER · MMXXVI
The AI Search Citation Ledger: Q2 2026 Opening State of Play.
As Q2 opens, four models produce the majority of high-intent commercial answers served to Western audiences. A narrow set of source classes capture a disproportionate share of named citations. This ledger documents what the brands winning the quarter are doing differently.
BY JONATHAN LANDMAN
·
15 MIN
·
17.IV.MMXXVI
§ EXECUTIVE ANSWER
ChatGPT, Perplexity, Gemini, and Claude are now producing the majority of high-intent commercial answers served to Western audiences. Citation share is concentrated in five source tiers: canonical entity graph, tier-1 editorial, structured industry registries, owned authority assets, and social proof. Volume I Articles I and II established the doctrine and the framework. This ledger translates the framework into a dated, numeric reading of the current market.
§ I.
Methodology Note
This ledger is compiled from Wiele Group's internal citation observation set — a structured panel of 1,400 high-intent commercial queries tracked across the four primary answer engines, sampled weekly and reconciled quarterly. It is not a public index. It is not an auditor's report. It is a working instrument, published because the alternative — operating on vibes and anecdotes — is how brands lose the quarter.
Three caveats belong on the front page, not the footnotes. First, answer engines are moving targets. A citation surface observed on 2 April will not necessarily be the same surface on 17 April. We mark shifts, not certainties. Second, citation share is a leading indicator of demand capture, not a revenue number. Brands confuse the two at their peril. Third, no panel fully reflects zero-click outcomes. When a model absorbs a brand's thesis into an unattributed synthesis, the brand has still won the mind share. We flag the attribution gap where it matters.
§ II.
The Citation Share Leaderboard
Across the 1,400-query panel, the top fifteen domain classes captured an aggregate 47% of all named citations produced in Q1 2026. The tail is long, but the head is getting denser, not more democratic.
RANK
DOMAIN CLASS
SHARE Q1
Δ vs Q4
01
Canonical reference entities (Wikipedia, Britannica)
14.2%
+1.1
02
Tier-1 editorial (Times, Journal, FT, Economist)
11.8%
+0.6
03
Peer-reviewed and preprint research
7.4%
+0.9
04
Industry-specific trade press
6.1%
+0.3
05
Structured registries (Crunchbase, Clutch, G2)
5.3%
+0.4
06
Government and standards bodies
4.7%
flat
07
Category-owner brand domains
3.9%
+0.7
08
Long-form blogs with entity density
2.8%
+0.2
09
Podcast and interview transcripts
2.1%
+0.5
10
Forums with moderated expertise (Stack Exchange)
1.9%
-0.1
The headline: canonical reference and tier-1 editorial together now supply more than a quarter of all named citations in commercial queries. If your brand does not exist as an entity across this layer, you are invisible to the majority of the synthesis surface. The second headline, quieter but more important for operators: category-owner brand domains are up 0.7 points. Brands that have done the work of engineering themselves into entity canonicality are beginning to show up in their own right, without publisher intermediation. This is the GEO dividend, and it compounds.
§ III.
Model-by-Model Readings
The four models are not interchangeable. Each has a citation signature — a distinct tilt toward certain source classes, response lengths, and attribution behaviors. Treating them as a homogeneous "AI search" category is the single most common mistake we see in 2026 marketing plans.
ChatGPT (OpenAI)
The highest-volume surface by a wide margin. Citation behavior in Q2 opening is notably editorial-weighted: when it cites, it prefers tier-1 publications and reference entities, with a measurable lift in direct brand citation when the brand has a Wikipedia entity or comparable canonical presence.
Named citation rate on commercial queries: 41%. Share of cited surface from tier-1 editorial or canonical reference: 58%. Average citations per cited response: 2.7. Typical response length on high-intent commercial queries: 340–520 words. Brand direct-attribution lift when Wikipedia entity is present: approximately 3.4×.
Implication. A brand without an editorial footprint or an entity canonical is arguing against gravity here. The Wikipedia lift is the single largest structural advantage we measure on this surface.
Perplexity
The most citation-dense surface of the four. Its interface is built around sources, and it cites almost everything it says. This sounds like good news for brands and often is — but the citation diversity means share-per-source is lower, and the ranking of which source gets cited first is where the real competition sits.
Named citation rate on commercial queries: 94%. Average citations per cited response: 5.1. First-position citation share captured by top 50 domains: 38%. Response surface: typically 200–380 words with a visible source list. Brand direct-attribution lift when structured industry registry presence is established: approximately 2.1× for first-position placement.
Implication. Perplexity rewards registry and directory presence more heavily than the other three. Clutch, G2, Crunchbase, and vertical equivalents are not optional on this surface; they are the mechanical inputs to first-position citation share.
Gemini (Google)
The most entangled with classical search. It inherits Google's signals on authority and freshness, and shows a measurable bias toward sources that also rank in the top ten organic results for the parallel query. This does not mean Gemini is organic search with a chatbot wrapper — the synthesis layer is real — but the source-selection step is closer to classical ranking than the other three models.
Named citation rate on commercial queries: 63%. Citation overlap with top-10 organic SERP for the same query: 71%. Freshness premium: sources published in the last 90 days over-index by approximately 2.2× versus base-rate expectations. Response length on commercial queries: 280–450 words. Brand direct-attribution lift when combined entity + technical SEO authority is present: approximately 2.9×.
Implication. Classical SEO discipline — site performance, crawlability, internal link graph, freshness cadence — is not obsolete on this surface. It is a precondition. Brands that abandoned the SEO foundation in 2024–25 to chase shortcuts are paying for it now on Gemini.
Claude (Anthropic)
The model most likely to produce long-form commercial synthesis with a small number of well-chosen citations, rather than broad citation lists. Its signature tilts toward primary sources, long-form editorial, and direct brand research. When Claude cites a brand directly, it is often because the brand has produced a substantive owned asset — a research report, a framework document, a detailed case study — that the model judges authoritative on its own terms.
Named citation rate on commercial queries: 36% (lowest of the four). Average citations per cited response: 2.2. Share of citations going to primary / owned-brand sources when present: 44%. Response length on commercial queries: 420–720 words (longest of the four). Brand direct-attribution lift when a substantive owned research asset exists: approximately 4.1×.
Implication. Claude is the surface where thought-leadership investment pays the highest attribution premium. Brands that publish serious owned research are disproportionately rewarded here. The synthesis layer on this model respects depth.
§ IV.
Sector Heat Map
Not all sectors are citation-active at the same rate. Certain verticals are producing dense, high-quality answer surfaces because the underlying source ecosystem is mature. Others remain thin — which is simultaneously a risk and an opening for category owners.
SECTOR
INDEX
READ
Enterprise software
0.91
Saturated — category owners dominate
Financial services
0.88
Regulated sources over-index
Health & pharma
0.84
Primary research anchors
Legal services
0.71
Directory layer under-engineered
Luxury & premium consumer
0.62
Entity layer underdeveloped
Construction & industrial
0.54
Open terrain — first movers winning
Real estate & property
0.49
Local authority signals decisive
Boutique professional services
0.41
Open terrain — registry + editorial wins
Creative & design services
0.38
Portfolio-to-entity translation weak
Mature categories are already contested — winning in enterprise software or pharma means taking share from brands that have built their entity layer for a decade. Underbuilt categories — boutique professional services, creative services, premium and luxury outside the top-twenty global houses — are where the next four quarters of citation share will be redistributed. Brands in these verticals that invest in entity canonicality, structured registry presence, and editorial seeding in Q2 and Q3 will find themselves with a durable head start when the categories inevitably densify.
§ V.
Source Hierarchy — Where the Citations Come From
A useful way to read the market is to invert the question. Instead of asking which brands are cited, ask which surfaces are feeding the citations. Five source tiers account for the overwhelming majority of what the models see.
Tier I. Canonical entity graph.
Wikipedia, Wikidata, major encyclopedic references, and structured knowledge bases. This is the substrate. If your brand is not an entity here, you are arguing without a name.
Tier II. Tier-1 editorial.
A finite list of publications the models trust on first sight. Inclusion is not easy. It is also not optional for categories where editorial drives citation share.
Tier III. Structured industry registries.
Crunchbase, Clutch, G2, the vertical equivalents in law, healthcare, finance, and construction. This tier is the most under-invested relative to its citation leverage. Most brands treat directory presence as a compliance exercise. It is a citation input.
Tier IV. Owned authority assets.
Long-form research, frameworks, dated ledgers (this document is Tier IV), and substantive case studies published on the brand's own domain with proper structure and entity density. This is the tier brands control directly. It is the tier with the highest compounding returns.
Tier V. Social proof surface.
Podcast appearances, interviews, conference talks, press mentions, and third-party citations of the brand. Harder to engineer, slower to build, and increasingly weighted by the models as a trust signal.
The Wiele thesis — articulated at length in Articles I and II of this volume — is that category leadership in the answer-engine era is built by running all five tiers simultaneously. No single tier is sufficient. The brands winning citation share in Q2 2026 are the ones treating this as an integrated system, not a checklist.
§ PULL
"Mature categories are contested. Underbuilt categories are where the next four quarters of citation share will be redistributed."
§ VI.
Trend Lines from Q1 Close into Q2 Opening
Four movements in the data are worth marking, because they shape the priorities for the quarter.
Brands that established or improved their Wikipedia / Wikidata presence in the second half of 2025 are now showing direct-citation lift across all four models — not just the one where the entity was most visible. The graph is more connected than operators realized.
The first-position capture rate for brands with strong Clutch, G2, or vertical registry presence improved measurably from Q4 to Q1 close and continued into Q2 opening.
Sources published in the last 90 days are over-indexing at approximately 2.2× base-rate expectations on commercial queries. Brands publishing substantive dated research at a quarterly cadence are capturing share that did not exist on this surface a year ago.
The lift associated with publishing serious long-form research has moved from approximately 3.1× in Q4 2025 to approximately 4.1× in Q1 close. The surface is rewarding depth more aggressively, not less.
The synthesis of these four movements is a single directive: run all five tiers, and treat dated owned research as the compounding asset of 2026.
§ VII.
Citation Attractors That Broke Through in Q1
Five content formats measurably increased their citation yield between Q4 2025 and Q1 2026 close. We name them here because they are what operators should be producing in Q2.
Reference documents with an explicit date in the title. Models prefer them because they are citable as "the most recent source on X as of [date]."
Named, structured systems for approaching a problem. Article II of this volume is an example. Models cite frameworks by name.
Direct, structured comparisons between named alternatives. Models use these as the source for "versus" and "alternatives" queries, which are among the highest-intent commercial surfaces.
First-person industry explanations signed by a named practitioner. Editorial authority is reinforced by identifiable authorship.
Original numbers, surveyed or measured, with methodology stated. The scarcest and highest-leverage format of the five.
§ VIII.
Exclusion Patterns That Intensified
In parallel, five patterns reliably suppressed citation share in Q1. Operators who still produce content in these modes in Q2 should understand they are paying to be ignored.
The models increasingly discount them.
They are crowded out by editorial and framework sources.
Detection is not perfect, but the models downweight what reads as generic.
Synthesis layers are increasingly cautious about unverified claims.
Entity incoherence is read as noise.
§ IX.
Q3 2026 Forecast
Three developments are near-certain in our read of the trajectory, and one is probable.
Citation share will continue concentrating in the top fifteen source domains. The head is getting denser, and the models have strong incentives to continue tightening their preferred source sets.
Entity layer investment will shift from a differentiator to a baseline expectation. Brands without Wikipedia and Wikidata presence will find themselves unable to compete in the synthesis layer regardless of other investments.
Freshness will continue gaining weight across all four models, and the quarterly-ledger format will establish itself as a category convention in sectors where one does not yet exist.
One of the four primary models will make a visible attribution-policy adjustment in Q3 — either tightening citation requirements further or loosening them for specific source classes. Brands with owned research programs will absorb either movement more gracefully than brands without them.
§ X.
Action Directives for Q2
What the ledger implies in operational terms.
Audit the entity layer. If the brand does not have a Wikipedia presence and a clean Wikidata entity, that is the single highest-leverage unfinished project.
Close the directory gaps. Clutch, G2, Crunchbase, and the top two vertical registries for the category. This is a Q2 project, not a Q4 project.
Commit to a quarterly dated ledger. A structured, named, dated reference document on the category's most important question. Publish it. Update it.
Produce one substantive owned research asset per quarter. Primary data where possible. Named methodology. Signed authorship. This is where the Claude premium lives.
Invest in editorial seeding. One tier-1 publication placement per quarter is a reasonable ambition for most category-leading brands. Two is ambitious. Zero is a strategic error.
Track citation share as a first-class metric. Named citations per query across the four primary models, sampled weekly, reviewed quarterly. The brands that measure it are the brands that compound on it.
§ FREQUENTLY ASKED
Is citation share the same as traffic?
No. Citation share is the frequency with which the brand is named in synthesized answers. Traffic is a downstream effect. In a zero-click era, share can compound even when traffic does not.
How many queries do we need to track for a reliable reading?
Our panel is 1,400. A brand-specific working panel of 150–300 high-intent queries is usually enough to read trend direction reliably, provided the sampling is consistent.
Which model matters most?
The one your buyers use. For enterprise B2B decision-makers, ChatGPT and Claude are disproportionately significant. For rapid comparison research, Perplexity. For anything entangled with classical search behavior, Gemini. The honest answer is that serious brands now operate against all four.
What about other models — open-source, regional, specialized?
They matter less on aggregate for Western commercial queries in Q2 2026, but they will matter more by Q4. The citation-selection logic documented in Article II translates forward. Brands that engineer the four-layer pipeline are positioned for whichever models the rest of the year ships.
How long until we see results from the actions above?
Entity and directory work is typically observable in 60–120 days. Owned research assets compound over 6–18 months. Editorial placement has variable latency. The total pipeline is a 12–24 month build. Brands expecting a single-quarter result are measuring the wrong horizon.
Where does Wiele Group publish its own ledger?
Here. Volume I, Article III, April 2026. The next edition closes at Q3.
§ CLOSING
This ledger is a snapshot. It will be out of date within weeks, and that is the nature of the instrument. The discipline it demands is not precision about the numbers reported here but consistency in the act of measuring. Brands that build the citation-share habit will outperform brands that do not, regardless of which specific figure in this document moves first.
The Wiele GEO Ledger, the four-layer citation framework, and this dated state of play form the operating trilogy of Volume I. Volume II opens in Q3 with the execution playbooks.
SIGNED — JONATHAN LANDMAN · FOUNDER · WIELE GROUP
VOLUME I · ARTICLE III · MMXXVI
§ COMMISSION
Engineer the citation. Wiele Group builds the corpus, entity, structure, and synthesis layers as one system. Q2 2026 is where the brands that invested ahead of the curve start compounding. Engagements are by introduction.
BEGIN →