What a brand still owns

In brief. AI assistants now write the last sentence a buyer reads before deciding. AYAN's position: visibility scores measure how often a brand appears, which is the wrong question. What matters is whether the brand being described is the brand the company built, and whether the four teams who own the cited sources, content, media, retail and frontline, can act on the gap.

Where this started

We did not start AYAN from a market map. We started it from a repeated, uncomfortable observation in rooms we knew well: brands that had spent two decades governing every word about themselves, every pixel, every pack shot, every adviser script, had no idea what the most consulted source about them was saying. They had governance everywhere except where the decision now happens.

Our backgrounds are in marketing, data and brand operations. That matters for what we built. We came to this problem from the side of the people who have to act on a finding: the brand director who owns the positioning, the country manager who owns the numbers, the field teams who meet the customer after the answer has already been read.

Conviction 1: what did brands actually lose?

For a century, whatever else changed, a brand had the last word before a decision. The label, the shelf, the site, the adviser. That sentence belonged to you.

Fig. 1. The last mile, before and after.

The channel did not change. The narrator did.

The model does this in seconds, with no brief, no media plan and no negotiation. It is not hostile. It is indifferent, which is harder.

The behaviour is no longer marginal. Forrester's 2025 buyers' survey finds 94% of B2B buyers using generative AI somewhere in their buying process, and twice as many naming generative or conversational AI as more meaningful than any other information source. 6sense puts buyers at roughly 70% through the journey, with 85% of their requirements set, before they contact a vendor at all.

We think this is the most significant change in brand communication since the arrival of search, and that it is being under-reacted to precisely because it does not look like a channel. There is no login, no dashboard by default, no invoice. Nothing arrives to tell a CMO that a third of their category conversation now happens inside a paragraph they have never read.

Conviction 2: why is visibility not the goal?

The category calls itself AI visibility, and we use the term because clients do. But visibility, on its own, is a vanity metric with better manners.

Fig. 2. Two different questions.

The visibility question: how often do we appear? Mention rate, share of voice, rank against competitors, movement week to week. The answer is a number that goes up or down, and the failure mode is a high score on attributes you never chose.

The sovereignty question: is the brand described the brand we built? What the model credits you for, which category it places you in, which sources it believes, whether it can justify choosing you in one sentence. The answer is a gap between the brand you decided and the brand being narrated, and no number alone would reveal it.

A brand cited constantly for discounts, in a category framing it never chose, is highly visible and quietly losing. A brand mentioned less often but described precisely as what it set out to be is compounding. The number that matters is not how often you appear. It is the distance between the brand you decided to build and the brand being narrated on your behalf.

We are not in the business of telling brands how often they appear. We are in the business of telling them whether the brand being described is the one they built.

Conviction 3: why does the model have to be able to say why?

An AI answer is a judgment, not a listing. It names a brand and then justifies it, usually in one sentence. That sentence is the whole game.

Which is why explainability is, in our view, the most underrated metric in the category. If a model can state clearly why you are the right choice for this buyer, in this context, you survive comparison, you survive a hedge between two brands, and you arrive in the conversation pre-qualified. If it cannot, you are a name in a list waiting to be filtered out. In our work with a beauty retailer, ranking first on explainability in beauty AI queries came with an 8.6% conversion uplift from frontliners. Not more mentions. A better reason.

Conviction 4: why is an insight that stops at a report not a product?

This is the conviction that shaped the platform most, and the one we argue about most with the market.

Every finding about an AI answer belongs to someone else. The content team owns the page the model should be citing. PR owns the article it cites instead. E-commerce owns the product data behind a commerce answer. The field team owns the conversation with the customer who arrives holding a wrong answer. A report that names a gap without reaching those four owners has done the easy 20% of the work and left the rest to a meeting that will not happen.

The fix itself is well documented. A study presented at KDD '24, across roughly 10,000 queries, found that adding quotations, statistics and citations to source content lifted visibility in AI answers by about 40%. Knowing that is not the hard part. Getting it written, published and verified by the team that owns the page is.

So we built for the handoff: findings that become briefs and tasks, in the tools those teams already use, with a verification date attached. If an activation did not change the answer, we would rather the platform said so than have it produce a better-looking chart.

Conviction 5: why is a brand never one market, one language or one model?

From the start, we worked with brands that operate across Arabic, English and French, in several markets at once, in front of several models at once. That was not a constraint. It was a clarifying condition.

A brand that is the default answer in one market and invisible in the market next door does not have a visibility problem, it has an evidence problem in one language. A brand described accurately by one model and generically by another does not have a brand problem, it has a source problem. Measuring brand by brand, market by market, model by model is not a feature list. It is the only resolution at which any of this is actionable.

What this makes us

Not a monitoring tool with a strategy layer on top. Not a consultancy with a dashboard attached. A platform built on a simple sequence: understand how AI shapes and recommends your brand, judge it against what you stand for, then turn it into omnichannel activation: content, media, retail and the frontline.

We are aware of how much of this the market currently takes on faith. The honest position is that the proof is operational, not rhetorical: whether the answer changed after the work, and whether the commercial number moved with it. That is the standard we want to be held to, and the one we would apply to anyone else in this category, including the people writing the comparison articles.

A brand still owns its evidence. That is now the whole of the territory, and most of it is unoccupied.

Three terms, defined

AI visibility. How often, and in what position, a brand appears in the answers generated by AI assistants such as ChatGPT, Gemini and AI Overviews, for the questions that precede a purchase in its category.

Explainability. Whether an AI model can justify recommending a brand in one clear sentence, rather than merely listing it. A brand the model can justify survives comparison; a brand it only mentions is filtered out.

Omnichannel activation. Turning a finding about an AI answer into executed work on the sources that answer is built from: content, earned and paid media, retail and marketplace data, and frontline briefing, each with a named owner and a verification date.

Questions we are asked

Is AI visibility just SEO under a new name? No. Search returns a list the buyer judges; an AI assistant returns a judgment with reasons already attached. Ranking logic is replaced by source selection and justification, so the work shifts from winning a position to being the evidence a model relies on, and to correcting the sources it cites when they are wrong.

Can a brand pay to appear in an AI answer? Not in the answer itself. Sponsored units sit below the generated response, and OpenAI states that advertisers cannot shape, rank or alter what the model says. Paid AI media can buy the slot underneath a verdict, never the verdict, which is why organic representation has to be measured before any budget is committed.

How do you know an activation worked? By re-running the same questions after it ships. Every activation carries a verification date: either the answer changed, or the work did not do what it was meant to. Commercial effect is tracked separately, through AI-referred traffic and by asking buyers at the point of contact what they checked before coming.

References

Forrester, Buyers' Journey Survey, 2025 (94% of B2B buyers use generative AI in the buying process; twice as many name generative or conversational AI as more meaningful than any other source).

6sense, B2B Buyer Experience Report, 2024 (buyers around 70% through the journey, 85% of requirements set before contacting sales).

Aggarwal et al., "GEO: Generative Engine Optimization," KDD '24 (around 40% visibility lift from quotations, statistics and citations across roughly 10,000 queries).

AYAN pilot, beauty category (first on explainability in beauty AI queries, 8.6% conversion uplift from frontliners).