Categories form when attention moves faster than capability. This one has a property the others did not: the assets cannot be bought.
Every marketing category that became large did so the same way: a new interface accumulated attention faster than anyone had built a discipline for reaching it. Search, social, and mobile each ran that pattern. AI visibility is running it now, at an earlier stage than most people realise — and the reason it will become a category rather than a feature is that the assets it requires cannot be bought quickly, which is exactly what makes markets form around them.
The pattern is consistent enough to be predictive. A new interface starts accumulating attention. Existing disciplines cannot reach it with existing methods. A gap opens between where audiences are and where practitioners know how to operate. Early specialists develop methods, terminology stabilises, tooling appears, budgets get named, and within a few years there is a category with agencies, job titles, conferences, and software.
Search followed it. Social followed it. Mobile followed it. Each looked like a feature of the existing discipline before it became a discipline of its own, and in each case the practitioners who moved before it was obvious captured a disproportionate share of what followed. The pattern is not a forecast; it is a description of something that has happened repeatedly.
Early. Attention is demonstrably moving — that part is measured. Methods are forming but contested. Terminology is unstable enough that four acronyms compete for the same practice. Tooling exists and is immature. Budgets are mostly borrowed from search rather than allocated. Job titles are beginning to appear and are not yet standard.
That combination places it roughly where search was before anyone had settled what to call it, which is the stage at which positioning is cheapest and least contested. It is also the stage at which the category is most easily dismissed as a rebranding of what already exists, which is what people said about search marketing too.
Five stages through which a marketing discipline becomes a category. AI visibility is at stage two, which is where positioning is cheapest.
Attention moves
Audiences adopt a new interface faster than practitioners develop methods for it.
Marker: measurable behaviour change, no established practice.
Methods contest
Competing frameworks and terminology. Early specialists, no consensus.
Marker: four acronyms for one practice. We are here.
Terminology settles
One vocabulary wins. Roles get named. Budget lines appear.
Marker: job titles standardise; procurement categories exist.
Tooling consolidates
Measurement standardises. A few platforms define the workflow.
Marker: comparable metrics across vendors.
Discipline matures
Established practice, training, benchmarks, commoditisation at the low end.
Marker: the work becomes a line item rather than an argument.
Not every emerging practice becomes a category. Many are absorbed as a feature of an existing discipline, and the argument that AI visibility is simply SEO doing its job is reasonable and may partly prevail. What suggests otherwise is the shape of the work.
Three of the four inputs — access, entity resolution, corroboration — span functions that do not currently share an owner. Search owns none of corroboration; communications owns none of technical access. Work that spans functions and belongs to none is exactly the condition under which a new discipline forms, because somebody eventually gets hired to hold it.
Corroboration accrues over years and cannot be purchased. That is what turns a capability into a market — scarcity of something valuable that money alone cannot compress.
This is the economically interesting part. Most marketing advantages can be bought: media, tooling, talent, even rankings to a degree. Entity confidence and independent corroboration cannot, because their value derives from being independently produced over time. Money accelerates the work that earns them and cannot substitute for it.
That produces genuine scarcity, and scarcity of something valuable is what markets form around. It also determines the shape of the market: services and measurement rather than media buying, because the constraint is expertise and time rather than budget. That is the same shape SEO took, for the same reason.
Practically, this is the question that determines how fast the category forms. Currently most spend is redirected from search budgets, which constrains it, because the work that most needs funding — earned corroboration — sits outside what a search budget conventionally covers.
The organisations making real progress have generally moved a portion of communications or brand budget under the same objective. That reallocation is what turns a search sub-practice into something with its own claim on resources, and it is the transition to watch. When AI visibility gets a budget line rather than a share of one, stage three has arrived.
The assets that decide this accrue over years and cannot be bought later. DUNkē tracks where you stand across eight AI engines — per prompt, against competitors — which is where positioning starts.
Intellectual honesty requires the counter-case. If answer engines stabilise into a modest feature of search rather than a primary interface, the work is absorbed into SEO and no category forms. If platforms build tooling that makes visibility management trivial, the services market compresses. And if licensing arrangements restructure how content reaches answer engines, the entire mechanism could change shape.
None of these is unlikely. What makes the position defensible under all of them is that the underlying assets — documented expertise, original information, independent reputation, machine-legible structure — retain their value in every scenario. Building them is a bet on the mechanism rather than on the category forming, which is the safer of the two bets.
The headline number in claims like this is usually invented, so it is worth being explicit that we are not making one. We have no basis for sizing this market and we are not going to produce a figure that looks authoritative and rests on nothing.
What can be said structurally: search marketing became a multi-billion-dollar services and software market because attention moved and reaching it required specialised, non-purchasable capability. Both conditions are present here. Whether the resulting market is comparable in scale depends on how much of information-seeking moves, which nobody knows. The pattern is the claim; the number is not.
Looking at how value distributed in the last three category formations is more instructive than the pattern itself. In search, the durable winners were not the earliest tool builders but the practitioners who accumulated domain expertise and the platforms that eventually defined measurement. Early tools were mostly acquired or displaced.
In social, the value accrued to those who understood the behavioural shift rather than those who mastered a particular platform’s mechanics, which changed repeatedly. The transferable lesson is that mechanics-specific advantage decays and behaviour-level understanding compounds — which argues for investing in why engines cite rather than in how any current one behaves.
The consistent sequence is services first, tooling second, and the reason is that tooling requires a settled workflow to automate. While methods are contested there is nothing stable enough to build against, which is why the current crop of AI visibility tools mostly measure and few prescribe.
That is the appropriate state. Measurement is genuinely automatable now; diagnosis and remediation are not, because they require judgement about which of four pillars binds and why. Tools claiming to automate the second are usually automating the first and describing it ambitiously, and the distinction is worth checking when evaluating vendors.
Every category that matured did so partly because measurement standardised. Comparable metrics allow budgets to be justified, vendors to be compared, and performance to be argued about productively. Until that happens, spend is discretionary and vulnerable.
AI visibility is pre-standardisation: different vendors measure differently, prompt sets are not comparable, and no shared definition of citation share exists. That is the main thing holding the category at stage two, and it is why publishing methodology openly — as we do with the research series — is more strategically useful than protecting it.
Category formation, three precedents and one in progress
Four horizontal tracks: search, social, mobile, AI visibility. Mark the five stages along each with approximate durations. Show AI visibility at stage two with the preceding three completed, and annotate what triggered each stage transition in the precedents.
An honest assessment of the competitive field: measurement vendors are furthest along, because measurement is the automatable part and the demand is immediate. Traditional SEO agencies are repositioning with varying degrees of actual capability. PR and communications firms have the corroboration capability and largely have not noticed it is relevant.
That last group is the interesting one. The pillar that binds most often is corroboration, and the firms that have always sold it are not currently in this market. When they arrive — and the incentive to arrive is substantial — the competitive structure changes considerably, because they own the capability that the search-derived entrants have to build.
Category formation usually produces an in-house-versus-agency equilibrium, and this one will settle differently from search. Measurement and technical work are internalisable quickly. Entity correction is internal by nature. Corroboration is the piece that most organisations will not build internally, because it requires relationships and sustained external work.
That suggests the durable agency position here is closer to communications than to technical SEO, which is not where most current entrants are positioned. Firms building a technical AI visibility practice may find themselves selling the part clients can do themselves, which is a poor place to be when a category matures.
Three developments would compress the timeline substantially. Standardised measurement, which would let budgets be justified and vendors compared. A high-profile case where a major brand’s absence from AI answers had visible commercial consequence, which is how categories usually get their forcing event. And agentic commerce reaching ordinary use, which converts visibility into distribution.
None is predictable and all are plausible within a couple of years. A brand positioning now is buying optionality against all three at a cost that is currently low, which is the entire argument for early accrual and is considerably more defensible than a forecast about which one arrives first.
If we are wrong, the most likely shape is absorption: answer engines stabilise as a feature, search practitioners extend their remit to cover entity and corroboration work, and no separate category forms. The work still gets done and no market emerges around it.
In that scenario the assets this article argues for — expertise, original information, reputation, structure — still hold value, and the only thing lost is the positioning premium of being early to a category that did not form. That asymmetry is why we would make the same recommendation while holding significant doubt about the category thesis itself.
Ignore the category question entirely and ask a narrower one: in two years, when a buyer asks an assistant which options to consider in your space, will you be named? If the honest answer is no, the work required to change it accrues over that period and cannot be compressed at the end of it.
That framing removes the need to believe any forecast about market size, category formation, or platform trajectory. It is a bet on a mechanism that is already observable rather than on a prediction, and it produces the same actions as the more elaborate argument at considerably less epistemic cost.
Strip out the category argument and the instruction is narrow: start accruing the non-purchasable assets now, because they compound and cannot be compressed later. That means entity clarity established, an evidence programme running, original information published, and measurement instrumented.
None of that requires believing the category thesis. It requires believing that being named when a buyer asks an assistant about your space will matter more in two years than it does today, which is a considerably weaker claim and one the current evidence already supports.
The honest version is uncomfortable for our own industry: build the capability before selling the positioning. The work spans technical, editorial, and communications, and most firms repositioning around AI visibility have one of the three. Selling a capability you do not have recreates the ranking-report problem with newer vocabulary.
The firms best placed are not necessarily the search agencies. Communications firms own the pillar that binds most often and largely have not noticed. Whoever assembles the full capability first — from either direction — is positioned for whatever this becomes, and assembling it takes longer than repositioning does.
Reduced to what we would defend without qualification: attention is measurably moving into interfaces where visibility is determined by assets that accrue over years and cannot be bought. That produces scarcity, scarcity produces markets, and the period before this is obvious is the cheapest time to accumulate.
Everything else in this article — the stage model, the historical pattern, the market-formation argument — is interpretation built on that. The narrow claim is enough to act on and does not require any of the interpretation to be right, which is deliberately how we have constructed it.
Marketing categories form when attention moves faster than practitioner capability, and AI visibility is at the stage where methods are contested and vocabulary unsettled — where positioning is cheapest and the category is most easily dismissed as a rebranding of something existing.
What suggests it becomes a category is that its assets cannot be purchased: entity confidence and independent corroboration accrue over years, which produces genuine scarcity and a services market rather than a media one. We state no market size because no defensible basis for one exists. The pattern is the claim, the number is not, and the narrow version — start accruing now — holds whether or not the category forms as described.
Where value accrued in three previous cycles
Rows for search, social, and mobile. Columns: who captured value early, who captured it durably, what decayed, and what compounded. The consistent pattern to surface: platform-specific mechanics decayed, behaviour-level understanding and accumulated assets compounded.
Articles predicting a new marketing category are usually written by people selling into it, which is a reason to discount this one and we would rather say so than have a reader notice. The historical pattern is real, our position in it is our assessment, and the conclusion happens to favour our business.
What makes the argument checkable despite that is the narrow version: attention is measurably moving, the determining assets accrue slowly, and starting later means arriving later. Each of those can be verified independently of anything we claim about categories, which is why we have separated them explicitly rather than presenting the whole as one thesis.
The observable marker that this thesis is holding: budget moving out of search and into a line of its own, job titles standardising, and procurement categories appearing. All three are visible from outside and none requires trusting a forecast.
If those do not appear within a couple of years, the absorption scenario is winning and AI visibility is a feature of search rather than a discipline. That would be a legitimate outcome and it would not change what a brand should do in the meantime, which is the property that makes this a low-risk position to take.
This is the widest-angle piece in the publication and the most interpretive. The research articles and the state-of-the-field assessment carry the evidence; this builds an argument on top of them about what the evidence implies for how the industry is structured.
Readers who find it overstated should read the research hub first, which is deliberately more conservative and reaches compatible conclusions from measurement rather than from pattern-matching. The narrow claim underneath this article — start accruing now — is supported there without any of the category theory.
Begin accumulating the assets that cannot be purchased — entity clarity, independent corroboration, original information, machine-legible structure — on the assumption that being named in an assistant’s answer will matter more in two years than it does today.
That instruction survives every scenario in this article, including the one where no category forms at all. It is a bet on a mechanism rather than on a market, which makes it the version worth acting on regardless of what you conclude about the rest.
Confident: attention is moving into interfaces where visibility is determined differently; the determining assets accrue slowly and cannot be bought; starting later means arriving later. All three are supported by published research or by straightforward reasoning about accrual.
Not confident: that a distinct category forms, what it would be worth, how quickly the stages progress, or which type of firm captures the value. Those are pattern-matching from three precedents, which is suggestive and is not evidence. Separating the two lists is more useful to a reader than presenting the whole as a single confident thesis.
The most reliable thing about every previous category formation is that it was obvious afterwards and contested at the time. That is not an argument that this one is real — the same could be said of every trend that did not materialise, and survivorship bias makes historical pattern-matching unusually seductive.
What makes the position defensible is that the recommended action is identical whether or not the category forms. Build the assets that compound, instrument what you cannot currently see, and stop treating your website as the whole of your visibility. Those are correct under the ambitious reading and under the dismissive one, which is the only kind of bet worth making under this much uncertainty.
Articles like this create urgency, and urgency produces bad decisions as readily as good ones. A brand that reads this and hires against an unstable job description, buys tooling before knowing what it measures, or commissions a repositioning it cannot support has responded to the argument badly.
The correct response is small and specific: instrument what you cannot see, fix your entity, start earning corroboration. Three actions, none requiring a strategic bet, all of which improve your position under every scenario in this article including the one where none of it happens.
Attention is moving into interfaces where being named is decided by evidence that accrues over years and cannot be purchased. That produces scarcity, and scarcity of something valuable is what markets form around. We are at the stage where methods are contested and vocabulary is unsettled, which is when positioning is cheapest.
Whether a distinct category emerges or the work is absorbed into search is genuinely uncertain and does not change what a brand should do. Build the assets that compound, instrument the surface you cannot currently see, and start earlier than the present impact justifies — because the one thing that cannot be bought later is the time it takes to accumulate.
Is it too early to hire for this?
It is early enough that the role does not have a settled definition, which cuts both ways — you will pay less and get less certainty. Most organisations are better served by giving an existing person the objective and the cross-functional mandate than by hiring against an unstable job description.
How would we know if this is not happening?
Watch answer coverage and referral trends. If coverage stabilises or contracts and referral decline reverses, the urgency reduces substantially. Both are observable and neither requires trusting anyone’s forecast.
Should agencies reposition around this?
Only if they can actually do the work, which spans technical, editorial, and communications capability. Repositioning without the capability produces the ranking-report problem one layer up — selling a new proxy for the same reason the old one was sold.
What is the single earliest-mover advantage?
Corroboration, because it accrues and cannot be compressed. A brand starting its evidence programme two years before a competitor holds an advantage the competitor cannot close with budget.
Marketing categories form when attention moves faster than practitioners develop methods for reaching it, and AI visibility is at the stage where methods are contested and terminology has not settled — which is precisely where positioning is cheapest. The pattern is descriptive rather than predictive: search, social, and mobile each ran it.
What suggests this becomes a category rather than a feature is that its determining assets cannot be purchased. Entity confidence and independent corroboration accrue over years, which produces genuine scarcity and a services market rather than a media one. We deliberately state no market size, because we have no basis for one and a confident figure resting on nothing would undermine everything else in this publication. The pattern is the claim; the number is not.
Methodology note: we state no market size figure. Sizing an emerging category requires assumptions about adoption, spend reallocation, and category boundaries that nobody can currently support, and published estimates in this space are generally constructed backwards from a headline. The historical pattern described is observational. The claim that we are at stage two is our assessment, not a measurement.
“Every marketing category formed the same way: attention moved faster than anyone knew how to reach it. The only unusual thing here is that the assets that matter cannot be bought later.” The Age’X Research Team
Get a free GEO audit — the same analysis behind every article here.