A standings table that records every cited source, not just brands — which is why its findings are uncomfortable.
The Citation Index answers a question benchmarks cannot: not how a brand performs against a distribution, but who actually holds the ground in a category. It is a standings table — who is named, how often, for what, and how concentrated the field is. The uncomfortable finding it consistently produces is that the leaders are frequently not brands at all, which changes what competing there means.
A benchmark tells a brand where it sits in a distribution. An index tells everyone who holds the category. Those are different questions with different uses: the first is diagnostic and private, the second is competitive and public, and the second is what determines whether entering a contest is realistic.
The unit differs accordingly. A benchmark measures citation rate — the proportion of your own priority prompts where you appear. An index measures share of citation — the proportion of all citations across a category’s question set that go to each source. One is about your hit rate; the other is about the division of a fixed pool.
The consistent and initially unwelcome result is that commercial brands frequently do not lead their own categories. Reference entries, community discussion, independent publishers, and specialist resources take a substantial share of citations on exactly the questions brands assume are being contested between companies.
This is not a measurement artefact. It reflects what answer engines need for the questions people actually ask: comparative candour, definitional clarity, and experiential detail, which non-commercial sources supply more readily than brands do. A brand entering an index at a low position is frequently not losing to competitors — it is losing to a category of source it had not counted.
How standings are constructed. Published so a reader can assess whether a position means what the table implies.
Standard question set
A fixed set of category questions built to a published template, not supplied by any brand in the index.
Prevents self-selected prompts producing flattering positions.
All cited sources recorded
Every source named or cited is captured, including non-brand sources.
Brands-only tables misrepresent the actual competitive field.
Share of citation
Position determined by proportion of total citations across the question set, not by hit rate.
Makes positions comparable and additive.
Per-surface reporting
Standings computed per answer surface, since surfaces cite different sources.
No blended cross-surface ranking is published.
Concentration reported
The distribution of share, alongside the ranking itself.
A rank without concentration hides whether the field is open.
It would be commercially easier to publish a table of brands. It would also misdescribe the competitive reality, because it would exclude the sources actually taking the citations and present a brand ranked fourth among companies as fourth in its category when it might be twelfth among all cited sources.
A brand acting on the first number would conclude it is close to leading and needs incremental improvement. Acting on the second, it would understand that most of the category’s answer share sits with source types it must either match or route around. Those are entirely different strategies, and only one of them is available from an honest table.
The measure that determines whether a position is worth contesting is not rank but concentration. A category where the top three sources hold most of the citations is closed — entering requires displacing an incumbent, which is rare. One where share spreads across twenty sources is open, and joining the set is achievable.
We report both because rank without concentration is misleading in a specific direction: it makes closed categories look enterable. A brand ranked eleventh in a category where the top three hold three quarters of the share is not eleven places from leading; it is competing for a fraction of a quarter of the pool. That reframing changes investment decisions.
Reference entries, communities and publishers routinely hold the largest share of a category’s citations. A brand entering low is often losing to a class of source it never counted, which is a different problem entirely.
It does not tell you commercial performance. A source holding the largest citation share in a category may be a community forum monetising nothing, and its position says nothing about whether a brand at position eight is doing well commercially. The index measures answer-share, which is one input to commercial outcome rather than a proxy for it.
It also does not measure quality of citation. Being named as an option and being named as the recommended choice count identically in a share calculation, and they are commercially very different. We report description sentiment separately for this reason rather than folding it into position, since combining them would produce a number that means neither.
Standings differ by surface because surfaces cite different sources — demonstrably so, including between two operated by the same company on the same index. A blended ranking would average genuinely different competitive fields into one that describes no actual surface.
So the index publishes standings per surface, and the divergence between them is itself informative. A brand leading on one surface and absent from another has a source-mix problem rather than an evidence problem, which is a diagnosis available only from disaggregated standings.
Most brands have never seen the full cited set for their category. DUNkē records every source cited for your prompts across eight engines — which is where an honest competitive picture starts.
Three uses. First, reality-testing the competitive set: the sources in the table are who you are actually competing with, which frequently differs from the slide. Second, assessing enterability: concentration tells you whether the category is contestable at all. Third, identifying what the leaders supply — because studying the source currently taking your category’s citations teaches more than any framework.
What it should not be used for is a target. Aiming to reach a given position without understanding what the sources above you provide is how brands end up producing content that resembles a community thread badly, which serves nobody. The table is diagnostic input, not an objective.
The index describes the question set it was computed on. A different set of category questions would produce a different table, which is why the question construction is published and why we describe positions as being on a defined question set rather than in a category absolutely.
It is also a snapshot of a field that moves, and answer variance means positions near boundaries are unstable between editions. We report share alongside rank so that a movement of two positions on a fractional share difference reads as what it is rather than as a competitive shift.
A definitional decision that materially affects the table: we count domains rather than pages, and treat a brand’s own properties as one source. A brand cited from three of its own pages in one answer counts once, because the alternative would let sites with many pages accumulate share without holding more of the category.
Reference sources are counted as themselves rather than being excluded or grouped, community platforms are counted per platform rather than per thread, and publishers are counted individually. Every one of these is a judgement that changes the table, which is why they are published rather than assumed.
Alongside the full table we report what proportion of a category’s citations go to commercial brands at all, versus reference, community, editorial, and institutional sources. That single number frequently tells a brand more than its own position does.
A category where commercial sources hold a small minority of citations is one where the contest between brands is a contest for a fraction of the pool, and the larger strategic question is whether a brand can become the kind of source that holds the rest. That reframing — from beating competitors to changing what kind of source you are — is the most valuable output the index produces.
The obvious objection is that comparing a brand against a reference entry is unfair, since they are not competing for the same thing. Commercially that is true. In an answer they are competing for exactly the same thing: a citation slot in a response to a buyer’s question.
Excluding them would make the table more comfortable and less true. A brand needs to know that most of its category’s answer share sits with sources it never considered, because that determines whether its strategy should be to compete with those sources, to differentiate from them, or to accept a smaller pool and contest it properly.
Share concentration against commercial share
Two-by-two: x-axis concentration (dispersed to concentrated), y-axis proportion of citations going to commercial brands (low to high). Plot categories as points. Quadrants indicate strategic conditions: open commercial, closed commercial, open non-commercial, closed non-commercial — each requiring a different approach.
A position in this table means: on this question set, on this surface, in this period, you received this share of citations. It does not mean you are the fourth best option in your category, that buyers rank you fourth, or that your commercial position is fourth.
That precision matters because standings tables invite over-reading. The number is a measurement of answer-share on a defined question set, and every one of those qualifiers changes what it supports. We state them repeatedly because a table without them gets quoted as a market ranking, which it is not.
Since share is a proportion of a fixed pool, one source gaining necessarily means others losing, which makes index movement zero-sum in a way citation rate is not. A brand can improve its own citation rate while losing index share if others improved faster.
That distinction confuses people and is worth stating plainly: the benchmark measures your hit rate, the index measures your share of a pool. Both can move in opposite directions simultaneously and neither is wrong. Reading them together is what produces an accurate picture of whether you are improving and whether you are keeping up.
Some categories will not support one. Where coverage is minimal, where citation is unstable across runs, or where the question set returns uncited answers, computing standings would produce a table with no underlying signal.
Those categories are reported as unsuitable rather than published thinly, and that is itself an informative finding — a category where an index cannot be computed is one where the answer layer has not yet arrived, and brands in it should read that as a timing signal rather than as missing data.
Three concrete uses. If commercial share is low, the strategic question is whether to become a different kind of source rather than to outcompete existing brands. If concentration is high, the question is whether the category is worth contesting at all versus finding narrower ground.
And if a non-brand source leads, study what it supplies — usually candour, definition, or specificity that brands avoid for understandable reasons. Each of those is a strategic decision rather than a tactical one, which is why an index is a planning input rather than a performance report.
The index depends entirely on the question set, which we construct and which could be constructed differently to produce different standings. It is a snapshot of a moving field. And it measures answer-share, which correlates imperfectly with commercial outcome.
We publish the construction so a reader can assess whether our questions represent their category, and we would rather be argued with on the question set than have the table accepted uncritically. A standings table is the format most likely to be over-trusted, which is why its limits need stating more insistently than the others in this series.
The productive response depends on what the table shows above you. If non-brand sources hold most of the share, the question is whether you can supply what they supply — candour, definition, experiential specificity — or whether you should contest narrower ground where those source types are weaker.
If commercial competitors hold it, the question is corroboration: they have evidence you do not, and closing it is a quarters-long programme rather than a content one. And if concentration is high and you are outside the top few, the honest answer is frequently that this category is not currently contestable and adjacent, more specific questions are the better investment.
A standings table is commercially useful to whoever holds it privately, which is an argument for not publishing. We publish for the same reason we publish methodology: a category cannot be assessed by the people operating in it if the only people who can see the field are vendors selling into it.
It also invites correction. A brand that believes its position is wrong can run the published question construction and demonstrate it, which is a better outcome than a table nobody can check. The risk of publishing something contestable is smaller than the cost of asking readers to trust a number they cannot verify.
The Citation Index reports standing and share rather than distributions: who holds a category’s answer share, measured as proportion of total citations on a published question set, computed per surface, with concentration reported alongside rank. It records every cited source rather than only brands, which is what makes it honest and uncomfortable.
The recurring finding is that commercial brands frequently do not lead their own categories. A brand entering low is usually not losing to competitors but to a class of source it never counted — and recognising that is the difference between a strategy aimed at the right opponent and one aimed at the wrong one.
Source types and what they supply
Rows for reference sources, community platforms, independent publishers, specialist resources, and commercial brands. Columns: typical share held, what the engine appears to draw from them, and what a brand would need to supply to compete on that ground. Turns the standings into a strategic map.
The most valuable output of an honest index is not a position but a question: given that a reference entry, two communities and a publisher hold most of my category’s answer share, what kind of source do I need to be? That is a different question from how do I outrank my competitors, and it has different answers.
Usually the answer involves supplying something those sources cannot: proprietary data, direct experience at scale, or specificity only an operator has. That is a content and evidence strategy rather than a competitive one, and it is available precisely because brands have advantages non-brand sources lack — they simply rarely use them.
Two measures we do not currently produce and would like to. Citation persistence: how long a source holds its share, which distinguishes durable positions from recent arrivals. And answer-position weighting, since being named first in an answer plausibly differs in value from being named fifth.
Both require longitudinal data the series does not yet have. Stating them as gaps rather than implying the current measures are complete is the same discipline applied throughout this publication, and readers should treat an index without persistence data as describing a moment rather than an equilibrium.
Third of four. The trends report covers the environment, the benchmark covers distributions, this covers standings, and the movement report covers deltas. This is the one most likely to be over-read, because standings tables invite treatment as market rankings and this measures answer-share on a defined question set.
Read with the benchmark it locates you twice — once against a distribution of peers, once against the full field of cited sources. Those two positions frequently differ substantially, and the gap between them is usually explained by how much of the category’s share sits with non-brand sources.
Find out what is actually being cited for your category’s questions, including everything that is not a brand. Most organisations have never seen that list, and it routinely contains sources they did not know were competing with them for the attention of their own buyers.
That single observation reframes more strategy than any framework in this publication, because it replaces an assumed competitive set with an observed one — and the two are frequently not the same set at all.
That commercial brands hold a minority of citations in many categories will be contested, usually by brands whose own tracking shows otherwise. The likely explanation for the discrepancy is prompt construction: a brand tracking its own prompt set will have built questions its content addresses, which selects for questions where brands are cited.
That is not a criticism — a brand should track the questions it cares about. It does mean self-tracked citation data systematically overstates commercial share relative to a neutral question set, which is exactly why an index needs a construction nobody in it supplied. Where our table and a brand’s own data disagree, the question set is usually the reason.
League tables are the most persuasive and least honest format in industry publishing. They compress complex measurement into an ordinal ranking, invite comparison across things that are not comparable, and get quoted long after the underlying data has moved.
We publish one anyway because the alternative — brands operating with no view of who holds their category — is worse. But the format deserves more caution than we can build into it, and readers should hold a position here as a measurement on a defined question set at a moment in time, which is considerably less than a table implies.
The most common misuse we anticipate is a brand seeing a competitor above it and concluding that competitor is better. Position here reflects accumulated corroboration and question-set fit, not product quality, service, or commercial performance.
A brand can be materially better and cited less, because citation measures documented evidence rather than merit. That gap between being good and being documented as good is the subject of most of this publication, and an index is the clearest available demonstration that the two are separate things.
A standings table for who holds a category’s answer share, recording every cited source rather than only brands — which is why its findings are uncomfortable and why they are worth having. Commercial brands frequently do not lead their own categories.
Read concentration before rank, because it determines whether the category is contestable at all. Study what the leading source supplies, because it is usually candour, definition, or specificity that brands avoid. And treat a position as a measurement on a defined question set at a moment in time, which is considerably less than a league table implies.
Standings invite a conclusion about who is winning, and the honest answer is that this measures one thing about winning. A source holding the largest citation share may be monetising nothing; a brand at position ten may be converting the citations it does receive at a rate that makes it the commercial leader.
The index is a measure of answer-share and nothing else. Read alongside your own commercial data it is informative; read as a proxy for market position it is misleading. We report it because nobody currently sees the field at all, not because share is the outcome that matters.
How are category question sets constructed?
To a published template covering intent types a real buyer moves through, built without reference to any brand in the index. The template is published so a reader can assess whether the questions represent their category fairly.
Can a brand be excluded or added?
Positions are computed from what is actually cited, so inclusion is an outcome rather than a decision. A brand absent from the table was absent from the answers, which is the finding rather than an omission.
How often does the index update?
On the same quarterly cadence as the benchmark, because the underlying positions move slowly and more frequent publication would report variance as movement.
What if we disagree with our position?
Run the published question construction yourself. If your result differs materially, either the question set does not represent your category as we constructed it, or sampling variance is at play — and both are worth knowing.
The Citation Index reports standings rather than distributions: who holds a category’s answer share, measured as proportion of total citations on a standard question set, computed per surface and published with concentration alongside rank. It records every cited source rather than only brands, which is the decision that makes it honest and the reason its findings are uncomfortable.
The recurring result is that commercial brands frequently do not lead their own categories — reference entries, communities, and publishers hold ground that companies assume is being contested between companies. A brand entering low is usually not losing to competitors but to a class of source it never counted, and recognising that is the difference between a strategy that can work and one aimed at the wrong opponent.
Methodology note: this edition publishes the construction method rather than completed standings. Producing an index requires a question template validated across categories and a sample large enough that positions are stable rather than variance-driven, and publishing a thin first table would establish rankings that later corrections would have to undo. Categories will be released as their question sets and sampling reach the stated thresholds.
“Brands assume their category is contested between companies. The table usually shows a reference entry, two forums and a publisher holding most of it — which is a different problem than the one they were solving.” The Age’X Research Team
Get a free GEO audit — the same analysis behind every article here.