Here is my observation: AI did not break the marketing funnel, it took over its first step. Awareness is not disappearing, it is moving from the consumer's memory to the model's memory. The marketer's question shifts with it. Next to "Does the consumer know me?" there is now a second question: "Does the model know me, and what does it believe about me?"
The funnel was never a law, it was a cost calculation
Awareness sits at the top of the funnel for an economic reason, not a psychological one. Search is costly. Consumers have limited time and attention, so they cannot examine every option and they evaluate only the brands that come to mind. Coming to mind was the gatekeeper, and brands spent decades of budget to get through that gate.
AI pushes the cost of search toward zero. Consumers no longer retrieve the options from their own memory. They hand that job to an intermediary, and the intermediary builds the consideration set. The step is not skipped, it changes hands.
We can already see the handover in the field. In the August 2026 wave of YouGov Türkiye's "Türkiye's Online Search Inventory" study, 51% of respondents say they would consider a brand they did not know before if AI recommended it. And 38% say they are less likely to consider a brand that does not appear in the AI answer.
AI opens the door, the brand closes the sale
The same study has a finding that looks contradictory at first: 52% of respondents find an AI recommendation more convincing when it names a brand they already know. Unknown brands get into consideration, yet known brands win more easily. This is not a contradiction. It is a picture of two separate stages.
In the first stage AI builds the list, and familiarity is not required. In the second stage the consumer chooses from the list, and familiarity still decides. The job of brand equity is changing: it used to get you onto the list, now it gets you chosen from it.
The competitive consequence cuts both ways. For the market leader, awareness is no longer the safe moat it used to be, because a rival can join the same list with a single recommendation. For the newcomer, the price of entry has dropped, but it is possible to make the list and still not be chosen. Both sides arrive at the same place: being recommended is necessary, and on its own it is not enough.
What changes for brand equity?
eBrandValue's method rests on the idea that consumers learn about a brand from the signals they receive (Akçura, Gönül and Petrova, 2004, Marketing Science). With each signal the consumer updates a belief about the brand, and each signal is weighted by how credible it is. Advertising is a signal with a known and interested source, so its weight is low. An AI recommendation is perceived more like the opinion of a neutral expert. That is why it is not surprising that 48% of respondents say such a recommendation can be more effective than the brand's own advertising.
The real novelty is this: two parties are now learning the brand. Alongside the consumer, the model also learns the brand from signals: news coverage, user reviews, independent write-ups, everything written about the category. What a model believes about a brand is the sum of the traces that brand has left in the world over the years. That is very close to the definition of brand equity. Brand equity is now kept in two ledgers: the consumer's mind and the model's memory.
The second ledger runs on its own clock. The memory a model acquires in training updates slowly, while the part fed by live search changes quickly. This has a hard consequence for budget decisions: the cost of brand investment cut today is not paid in this quarter's sales, it is paid as a missing brand in tomorrow's models. Brand equity is a stock that builds slowly and erodes slowly. AI does not remove that lag, it adds a new layer to it.
The third change is in where trust comes from. AI's credibility is borrowed from the independent sources it relies on. That is why a recommendation cannot be bought, it can only be earned. Consumers do not grant this trust unconditionally either: according to the study, after an important answer 27% of respondents check the cited source and 26% search again in a search engine. At the moment of verification, what they meet is the brand's real reputation.
The shelf got shorter
A supermarket shelf holds dozens of products in a category. The first page of a search engine holds ten results. An AI answer usually names three to five brands. As the shelf gets shorter, share concentrates: the gap between being on the list and being off it is sharper than the gap between the top result and page two.
We see this in our own measurements. For a beverage brand, we asked eight AI assistants the same set of questions twice, about six weeks apart. The share of answers that mentioned the brand barely moved: it was named in roughly three out of four answers both times. Yet the brand's overall score fell by about one fifth over the same period. The assistants still knew the brand, but they had started giving first place to competitors. Most of the loss came from a single assistant.
The same measurement showed something else: the brand led the list in one category and was almost never recommended in a neighboring one. The type of question mattered too. A brand that was strong on a general recommendation question fell behind on a price-focused one.
The most striking observation was this: one of the brand's products, the market leader in its own category, did not stand out in most of the assistants. Only three of the eight recommended it first, and in the first measurement one assistant presented the product as if it belonged to a competitor. Leading on the shelf and in the consumer's mind does not guarantee leading in the model's memory.
A classic awareness study would read this picture as "no change." On a short shelf, what counts is not whether you are mentioned but where in the order you are mentioned. Being known and being recommended are different things, and the gap between them shows up only when you look at the model through the questions consumers actually ask.
The consideration set is observable for the first time
For decades marketing science worked without being able to see the consideration set directly. It lived in the consumer's mind, and we estimated it with surveys. AI makes this hidden variable visible for the first time: we can ask a model the same question many times and count how often each brand makes the list.
At eBrandValue we run this measurement under the name AI Mindshare. It rests on four principles:
- Repetition. A model can answer the same question differently each time. A brand that appears in nine answers out of ten is not in the same position as one that appears in three, and a single answer cannot show that difference.
- Relative share. What matters is not whether the brand was mentioned, but how much of the category's recommendations go to it and how much go to its competitors.
- Question variety. Share is measured separately across question types, such as general recommendation, single best pick and price-focused questions. The same brand can be strong on one and weak on another.
- Link to sales. Saying "I would consider it" in a survey is a stated intention, not behavior. AI share turns into a budget decision only when it enters the sales forecast and the marketing mix model as an input.
For years awareness budgets invested in a single memory. Now there are two, and the second one can be measured. Consumers may well remember your brand. The real question is who will remember it once they stop searching and start asking.
To see your brand's share in AI answers, get in touch with the eBrandValue team.
Sources
- YouGov Türkiye, "Türkiye'nin Çevrimiçi Arama Envanteri" (Türkiye's Online Search Inventory), third wave, August 2026.
- Akçura, M. T., Gönül, F. F. and Petrova, E. (2004), "Consumer Learning and Brand Valuation", Marketing Science 23(1), 156-169.
To learn more about eBrandValue Platform, and how such metrics influence brand value, contact us using the form below.
