Most buyer questions AI answers are crowded, not owned

Showing up in an AI answer for your category is the easy part. Owning the question is rare. Across 612,952 brand recommendations on ChatGPT and Google AI Overviews, the median buyer question had 10 brands competing for it, and the leading brand took under a quarter of the recommendations. Half were wide-open free-for-alls with no brand above 25%; only one in eight was locked by a dominant brand. The questions buyers ask most are the most crowded.

How many brands compete for the average AI buyer question?

Parse breaks every AI answer down to the buyer need behind it ("high-yield savings account," "visual sales pipeline management," "truck accident legal representation") and records which brands the model recommends for that need. Across the 2,661 needs with enough volume to measure reliably, the median had 10 brands competing, with a typical range of 7 to 14. The leading brand for a question held a median of 24.8% of its recommendations. More than four in five of these questions (86.1%) had at least six brands circling, and just 12.2% were locked, meaning one brand took half or more of the recommendations. Half (50.1%) were wide open: no brand above 25%. The headline most teams get wrong is that being recommended at all feels like progress, when in most categories it only puts you in a crowd.

How we measured contestedness

We used Parse's reviewed recommendation-evidence layer: each record links a brand the AI recommended to the specific buyer need it was recommended for, on ChatGPT and Google AI Overviews, for prompts run between October 19, 2025 and April 25, 2026. We kept only recommended_for_need observations and collapsed each to one record per brand, per need, per answer, so a brand named twice in a single response counts once. That left 612,952 brand-need recommendations spanning 338,580 distinct needs, 48,739 brands, and 113,670 answers. Contestedness, how many brands compete and how concentrated the leader is, was measured only on the 2,661 needs with at least 20 recommendations, because a need seen three times tells you nothing about who owns it. The long tail of rarely-asked needs is undersampled by design; do not read it as a field of uncontested wins. This is the recommendation layer, not raw mentions, and it covers two surfaces, not Perplexity or Gemini.

Features and workflows are 77% of what buyers ask AI

Before you decide which questions to fight for, look at what buyers actually ask. Parse classifies each need by type, and the mix is lopsided. Feature requirements ("ad-free streaming," "fluoride-free toothpaste") were 43.3% of all brand recommendations. Workflow needs ("AI adoption change management," "visual sales pipeline management") were another 33.3%. Together, feature and workflow questions account for 76.6% of everything AI fields about brands. General category questions were 12.3%. The needs buyers supposedly obsess over, price and integrations, barely register: pricing and contract questions were 4.4% of recommendations, integration requirements 3.2%. Audience, business-size, and location needs were rounding errors. This matters because the volume is also where the competition is. The lane everyone assumes is the main event, "which is the best [category] tool," is precisely the lane with the most brands already standing in it.

If you want to see how AI engines describe your own brand, run a free brand check — it takes a minute.

The questions AI answers most are the hardest to win

Split contestedness by question type and the trap becomes obvious. The high-volume feature and workflow questions are crowded battlegrounds: a dozen or so brands each, a leader holding under a third, almost nothing locked. The low-volume pricing and integration questions are where a brand can actually own a slot: the leader holds close to 39% and roughly a quarter are locked, but there are far fewer of them.

Read down the table and the strategy inverts: the more a question type gets asked, the harder it is to lead. A pricing or integration question is roughly twice as likely to have a dominant brand as a feature question, and it has fewer competitors to displace. Chasing your biggest feature keyword puts you in the crowdedest part of AI search; building an answer the model trusts for "[your product] pricing for 50 seats" or "[your product] Salesforce integration" puts you somewhere a recommendation can stick. The crowded lanes are not worthless, but they reward patience, and the open lanes reward speed.

Which industries are wide open, and which are locked up

Contestedness also splits hard by industry, and the gap is large enough to change where a brand spends. Service categories are knife-fights: in Professional Services, the median buyer question drew 12 brands and only 1.6% of questions had a dominant leader. Biotechnology, Transportation, Consumer Goods, and Manufacturing all ran a median of 13 brands per question. The opposite end is genuinely ownable. Consumer Electronics had a median of 5 brands per question, the leader held nearly half of recommendations, and 52% of its questions were locked.

The practical read: in a locked-up category like consumer electronics, displacing the leader is a multi-quarter project, but defending a lead you already hold is realistic. In a wide-open category like legal services or professional services, no incumbent is safe, and a focused brand can climb fast because no one holds the question. The same AI-visibility budget buys a very different outcome depending on which side of this table your category sits on.

What an owned buyer question actually looks like

The rare locked questions are instructive. They are narrow and specific, and one brand has become the near-default answer for them: "visual sales pipeline management" drew nine brands but the leader held 81% of recommendations; "design system documentation" sat at 84%; "small business help desk" at 81%. Compare those to the most contested needs we measured, which were lead-generation categories where dozens of brands get named for the same broad job. "Slip and fall legal representation" surfaced 176 distinct brands with the leader at 26%, and "custom engagement ring design" surfaced 58 brands with the leader at 11%. The lesson is not that workflows are always winnable; it is that AI crowns a leader when a brand becomes synonymous with one tightly-scoped job. Owning "the visual pipeline CRM" is achievable. Owning "best CRM" is not. The questions worth targeting are the specific ones your competitors treat as too small to bother with, including the buyer questions no brand consistently owns yet.

How to turn a contestedness read into a plan

Parse tracks AI visibility across ChatGPT and Google AI Overviews, covering a public index of more than 4.7 million AI responses, 603,000 brands, and 57 million citations. A contestedness read turns that index into a target list in three moves. First, separate the questions you can lead from the ones you can only join: pull the buyer needs in your category, sort by how concentrated the leader is, and stop pouring effort into questions where 12 brands split the recommendations evenly. Second, weight toward the specific over the broad: the narrow workflow, pricing, and integration questions are where a recommendation actually sticks, even though they get asked less. Third, defend what you already lead, because a locked question is a churning one: only 30% of brands stay visible from one answer to the next. Parse's Rankings view shows which buyer questions your category competes on and where the field is concentrated versus open, and the Compare tool maps your position against named rivals on each. For the why-behind-the-pick, our breakdown of how ChatGPT decides which brands to recommend covers the mechanics, and does AI search favor big brands covers who tends to win the crowded questions.