The AI shortlist is the new page one.

01 — The answer is a shortlist, and the shortlist is the market.

When a buyer asks an AI assistant "what should I use for X", the answer isn't ten blue links — it's a verdict. In our scans, a typical answer names about four brands; 78% of answers named at least one of the brands we tracked. Buyers take that list and go search the names on it (B2B practitioners describe exactly this behavior). If search ranking was about being on page one, AI visibility is about being one of the four names in the paragraph. There is no page two.

02 — Position inside the shortlist is winner-take-most.

Even inside the list, the default slot concentrates: on ChatGPT, Stripe is the first recommendation in 50% of payment answers; Paddle, named almost as often overall, gets the top slot just 18% of the time. Being mentioned and being the recommendation are different businesses.

03 — Size helps, but it doesn't guarantee a slot. That's the opportunity.

Bigger brands do get written about more, so size and visibility broadly correlate — but our data shows the mapping is loose enough to matter. Paddle out-recommends Stripe on this buyer-question set (73% vs 66%) despite being a fraction of its size. Lemon Squeezy — a well-known brand — is invisible to ChatGPT (5%). The leaders of the AI-visibility industry itself score ~30% in their own category. Incumbency isn't a lock, and absence isn't destiny. In search, outranking an incumbent took years and a content army; in AI answers, the leaderboard is demonstrably still soft.

04 — The race has already started — your competitors are feeding the sources.

AI answers are assembled from sources, and the sources are being worked. In payments, two rival platforms' blogs were cited on more questions than stripe.com or paddle.com themselves. In the GEO category, the answer to "is this industry trustworthy?" is largely sourced from that industry's own content. Reddit threads, YouTube videos and comparison listicles dominate coverage. Whoever writes the pages AI reads writes the answer — and in every category we've measured, someone is already doing it deliberately.

05 — "Your AI visibility" isn't one number. Engines are parallel universes.

The same brand, same questions: 93% on Google AI Overviews and 53% on ChatGPT (Paddle). 69% vs 5% (Lemon Squeezy). 40% vs 6% (Otterly). As buying questions spread across ChatGPT, Perplexity, Gemini, Google and whatever ships next, visibility becomes a per-engine property — a brand can be dominant and invisible simultaneously. Single-engine spot checks tell you almost nothing.

06 — And the ground moves: answers drift week to week.

Ask the identical question again and 15% of our question×engine checks flipped — recommended in one pass, absent the next. AI answers are a distribution, not a fact. That has two consequences: any single screenshot of "what ChatGPT says about us" is close to meaningless, and visibility is something you hold, not something you win once.

A real drift pattern from our scans — identical question, identical engine, minutes apart. 15% of all checks behaved like this.

The part that is extrapolation — labeled as such

Everything above is measured. Here is the argument, which is not: the share of buying journeys that start and end inside an AI answer is growing. Independent analyses found ~58% of Google searches already end without a click (SparkToro/Datos, 2024); Gartner projected traditional search volume to fall 25% by 2026; OpenAI reported ChatGPT passing 800 million weekly users. If even a fraction of those sessions are purchase-shaped, the shortlist mechanics we measured above are compounding under them. We can't measure the future. We can measure that the mechanism is already here, already concentrated, already being gamed, and already unstable enough to need watching.