Category Design Is the Moat: Why Feature Lists Lose in an AI-Flattened Market

The claim: A longer feature list is not a moat. Features are copied in weeks and flattened by any AI that can read a comparison table. The durable advantage is positioning, the category you define and the frame buyers, and now the models, use to evaluate. Compete on features and you are playing a game the best-positioned company already won. What would make it wrong: feature-parity products with weak positioning consistently beating better-positioned, weaker products. Markets usually reward the better-framed product, not the longer spec sheet.

Most product marketing is a list of features wearing a nicer font. The team ships, the team announces, the team adds a row to the comparison table. It feels like progress because the list gets longer. It is not a moat, and in an AI-flattened market it is barely a speed bump.

Two forces have made the feature list nearly worthless as a defense. The first is speed: any feature worth having is reverse-engineered and shipped by competitors in a sprint or two. The second is newer and underrated. When a buyer wants to know which tool is best, they increasingly ask a model, and a model’s job is precisely to read every comparison table and flatten it into a shrug, “they all do roughly the same thing, here are three options.” A feature war is a war fought on the one terrain where AI is strongest and your differentiation is weakest.

Positioning is not a tagline

The way out is the part of marketing teams treat as decoration: positioning. Not a slogan, not a brand color. April Dunford defines positioning as a small set of deliberate choices (Obviously Awesome):

• the competitive alternatives, what a customer would do if you did not exist,
• your unique attributes, what you have that those alternatives lack,
• the value those attributes enable, with proof,
• the target customers who care most about that value,
• and the market category you place yourself in, the frame that makes the value obvious.

(She adds one optional ingredient on top, a relevant trend that makes you matter now.)

Read that list again and notice what it is doing. Four of the five components are about context, who you are compared to, who you are for, and what game you have decided to play. Only one is about features. Positioning is the act of choosing the context in which your product is the obvious answer, instead of letting the market, or a model, drop you into a context where you are one of nine.

The strongest version: name the category

The most durable form of this is category design: do not just position inside an existing category, define a new one whose problem statement makes you the answer. The canonical plays are canonical for a reason.

Salesforce did not enter the CRM software market, it declared war on it. Its 1999 mission was literally “The End of Software,” and its logo was the word software in a red circle with a line through it (background). Buyers were no longer asked “which CRM software,” a question with incumbents, but “software or no-software,” a question only Salesforce answered. HubSpot did the same with “inbound marketing,” a phrase Brian Halligan put a name to around 2006 and a book behind in 2009; notably, the founders describe it as naming something they were already doing rather than a calculated land grab, which is its own lesson, sometimes the category finds you. Drift was more deliberate, coining “conversational marketing” and cementing it with a 2019 book. Gong was the most deliberate of all, formally launching the “revenue intelligence” category in October 2019, moving up from the narrower “conversation intelligence” it could have been boxed into, and getting the analysts to ratify the new frame.

What these companies own is not a feature. It is a question. They got to define the question buyers ask, and they made themselves the obvious answer to it. That is a moat a competitor cannot ship in a sprint, because it does not live in your product, it lives in the buyer’s head.

The economics back the instinct. The team behind Play Bigger studied category dynamics and found the prize is winner-take-most: the category king commands roughly 76% of the market capitalization of its category (Play Bigger). Three quarters of the value accrues to the company that defined the game. Everyone else splits the remainder for building, often, a better product.

Now the model is a buyer too

Here is the part that makes this urgent rather than evergreen. The frame used to live only in the buyer’s head. Now it also lives in the model’s. When someone asks ChatGPT, Gemini, or Perplexity “what’s the best tool for X,” the answer is assembled from how the category is described across everything the model read, and from who is consistently named as the reference point.

We measured how engines pull those answers together. In our cross-engine citation study, the three major AI engines agreed on almost nothing, only 28% of cited domains appeared in more than one engine, and each leaned on whoever the surrounding corpus framed as the authority on a topic. In our extractable-sentence study, we found engines pull the short, clear, framing sentence, the definition, not the dense feature spec. Put those together and the implication for positioning is direct: the company that owns the category’s language, the clean sentence that defines the problem and names the answer, is the company the model reaches for. Own the frame across the corpus and you own the AI’s default answer. Show up as a feature list and you get flattened into “one of several options.”

Positioning was always about controlling the frame in which you are evaluated. The buyer doing the evaluating is now sometimes a model, and the model is even more frame-dependent than the human, because it has no loyalty, no relationship, and nothing to go on except the words the market uses to describe the problem. The category you design is the prompt the model answers.

The honest caveat

Category design is not a license to invent a name and declare victory. Most attempts fail, because most are a slogan with no real wedge behind them, a new word for an old thing nobody was confused about. The frame has to be earned: a genuine difference in how you solve the problem, stated as a category the buyer recognizes as real. Salesforce had a genuinely different delivery model behind “no software.” Gong had genuinely different data behind “revenue intelligence.” Without the wedge, category design is just expensive branding, and the model will see through it faster than the buyer.

So the work is not “make a category.” It is: find the real wedge, choose the frame that makes it obvious, and then make every surface, your site, your docs, the way you describe yourself everywhere the corpus is read, reinforce the same clean sentence. Do that and you have built the one moat that survives both a fast-following competitor and a summarizing machine.

The line that matters

A feature is a sprint away from being copied and a prompt away from being flattened. The moat is the frame: the category you define and the words buyers and models reach for to describe the problem. Compete on features and you are playing a game the best-positioned company already won, in the buyer’s head and now in the AI’s answer.

Want your category and frame engineered for how buyers and AI engines actually evaluate you? → Book a growth audit

Related: AI engines barely agree on sources · The extractable sentence · More from Meriin Labs