Because AI Doesn't Bowl

One of the more frustrating aspects of Brand AI strategy is an inherent tension.

Imagine someone asks AI for the best brands of bowling shoes. AI doesn’t bowl, so it reads whatever it can about bowling shoes: from structured and official sources, to the self-described, to the bowling banter of social media, and everything in between.

If you’re a bowling shoe brand that deserves to be in that answer, you have to be present across the places AI looks. AI is looking for evidence that you belong in the category; that you are classified as a bowling shoe brand and repeatedly compared, described, sold, and discussed alongside other brands already known to be bowling shoes –- and recently.

Getting grouped with the competition, in all the same AI sources, sounds like a recipe for sameness, not brand differentiation. If you stopped there, did the bare minimum of managing your brand in AI sources, you would probably be right.

But the information a brand provides across those sources can and should have real dimension -- depth, detail and differences that most humans are unlikely to ever read, but that can give AI specific reasons to distinguish one brand from another.

It is these dimensions that brands can lean into. AI won’t care much about a new ad campaign itself, but it will notice the numbers and other data that show what the brand does right. For example, patents, annual reports, testimony at government hearings, new materials used, engineering tests, range of products offered, support for local bowling leagues, and maybe the measured performance of pro bowlers who wear the shoes: all sorts of specific, citable evidence will begin to separate the bowling shoe brands.

Not all that detail belongs in all sources. The choice of what goes where is certainly strategic, and potentially creative -- in the service of positive differentiation in AI.

Once AI has enough evidence to place a brand in the category, those with the most citable, corroborated evidence in trusted sources are more likely to appear in its response to ‘the best.’ Category association gets a brand considered. Differentiating evidence gives AI a reason to choose it.

If only it were that easy.

Unfortunately, many of the sources that can confer the most authority are also the ones brands can’t inform themselves. That is part of why they carry such authority.

Qualifying and differentiating? High authority and low? To keep it all clear, I think in terms of “Tiers”, like rock strata or Jenga -- pick your metaphor.

Don’t read too much into the numbers. These are tiers for managing, not scores or a fixed order that AI follows. A government filing may be relevant to one prompt and not at all to another. A product page might provide just the right arcane data about material performance. To answer “Where’s the best place to buy bowling shoes in Berkeley?”, AI would probably check Reddit and Google Maps over Wikipedia.

The names of the sources change with the category, of course, and their influence on the answer changes with every prompt. But the source-types themselves that define each Tier don’t change much. That stability is how brands can manage their representation across AI sources.

These definitions are not meant to be comprehensive, just illustrative. Yours may certainly vary, as mine often do, but generally:

Tier 1 — Structured and official knowledge.

Wikidata, Wikipedia, government filings, official registries, and the structured systems used by search platforms (like the Google Knowledge Graph).

These sources help AI answer basic questions: What is its legal name? Where is it based? What category does it belong to? Is it real, correctly identified, and accurately described?

These sources are not all the same. Some can be edited by independent contributors (though usually not by brands). Some are official records. Some are facts that a company has supplied in a form.

Their common value is that they make information easier to identify, compare, and confirm. Something else they have in common: they are slow to inform and hard to edit.

In Wikipedia, for example, brands are strongly discouraged from directly editing articles about themselves. If they do, the accepted practice is to first disclose their “Conflict of Interest”, then propose well-sourced changes for independent editors to consider. And those editors can refuse or delete anything they deem to be promotional.

Tier 2 — Major editorial, government, and regulatory sources.

National press, business press, analyst reports, regulatory documents, and academic publications.

Like Tier 1, this process is also slow, laborious, and, even if successful, the result is written by people who don’t work for you and aren’t on your schedule. However, there is an expectation that the information in Tier 2 has been researched, reviewed, or produced by experts or under professional rules. That’s probably why they carry so much weight with AI.

Tier 3 — Registries, trade, and local press.

Industry directories and ratings bodies, trade publications, and regional press.

They may carry less broad authority than Tier 2, but in Tier 3 sources AI can see how others have associated a brand with its peer group, often in a specific geography.

Earning your way into Tiers 2 and 3 takes some doing.

Years ago, building my own agency, when we won a major account, I would email ADWEEK, hoping for some coverage -- which never happened. Eventually, I called the reporter of an article I thought we belonged in and asked what the famous agencies she covered had that we didn’t. “Fame. No one wants to read about an agency they’ve never heard of.”

But when a respected publication does write about what makes you interesting, the reporter’s language -- the angle, your newsworthiness -- can become AI’s language about you in response to a prompt, for better or worse.

These first three tiers can make you easier to verify, categorize, and include with confidence. They are slow by design -- official, difficult to earn, and painful to change. They can establish that you are real, relevant, and part of the category. But presence alone will not differentiate you. Dimension will.

Tier 4 — Owned media.

Your website, your blog, your press releases, your podcast. The grist of modern marketing. SEO or not, AI reads a lot of it.

When presence in the more authoritative Tiers 1-3 is thin, owned media can supply much of a brand’s detail, and AI will very likely use it. But if you are perennially #5 in AI’s list of the 4 “best brands”, i.e., invisible in every response, it is unlikely that your owned media alone will tip the scales in your favor.

Even so, Tier 4 is where differentiation lives.

The magnitude of your successes, products, leadership, and growth can shine because this is the one layer you fully control. It is where you plant a flag on the part of your category you own — your authority.

For some brands, that authority is more obvious. It is a showcase for their purpose, people, science, or accolades. For others -- service companies especially -- it takes real work to articulate a differentiating authority.

But work it out they must.

Because for AI to answer, “What are the best brands in your category?”, the question beneath it, for brands to answer first, can be brutal:

“With all the available evidence, why would a machine recommend us, instead of, or alongside, our competitors?”

Owning a narrow and defensible patch of your category -- publishing and speaking about it often, and connecting your successes, products, and reviews to it -- gives you a plausible path into the handful of brands AI can confidently recommend. But it has to be authentic, consistently demonstrated, and eventually supported by sources beyond your own.

Tier 5 — Commercial placement.

Sponsored answers, e-commerce integrations, and the platform deals that inform “Where can I buy...?” prompts. Present in the answer, but paid, rather than earned. The placement itself builds little authority for tomorrow, although the attention and customer experience it generates could.

Tier 6 — Social.

Reddit, LinkedIn, review sites, and forums.

When the prompt asks, “What are people saying about your category?”, Tier 6 is likely to matter quite a bit. AI uses these sources to gauge recency, experience, engagement, and sentiment, not necessarily to confirm facts.

Less authoritative for some questions, but when the question is about the buzz, AI knows where it lives.

Tier 7 — The junk layer.

Content farms, scraper sites, link-only directories, and AI-generated spam.

AI systems may discount Tier 7, but those sources can still be cited in AI responses. Even so, if your brand visibility rests here, you have bigger fish to fry than AI.

Tier 0

There is one layer beneath all of these -- “Tier 0” -- each model’s own training data. The bad news: it is untouchable by brands. The good news: during the infrequent rounds of AI training on new data, repeated and consistent representation across sources may help a brand’s identity and associations become part of a model’s learned picture of the world.

So what’s more important? The dry registry no human will ever read, or another quippy post from Cannes? Both. They are not competing. And these tiers run on different clocks. The slow tiers are your Brand AI infrastructure that you check quarterly. The fast ones are the habit that feeds your brand’s recency and relevance.

Coverage makes a brand legible. Corroboration makes it credible. Differentiating evidence makes it recommendable.

You can control, to varying degrees, how your brand is represented in AI sources. But you can’t know, let alone control, AI’s response to someone using AI to search for your category of products or services. You can see the tiers, how each one represents you today, and put the work where your gaps are.

Most brands have never looked.

This post first appeared in the Brand AI Report, my newsletter on how AI describes and recommends brands.