AI Always Names Anderson First.
River Rat Motorsports scored 27 out of 30 — the highest result we've published, and not the good news it sounds like. Across ten prompts on ChatGPT, Claude and Gemini, this two-store Arizona powersports dealer was named in nine of ten questions, with both addresses and phone numbers correct every time. Then you read what the engines actually said. In all nine answers where a competitor appeared, Anderson Powersports was named first and described with model names, while River Rat got "and others." One engine merged the two dealers into a single entry. Another sent a Laughlin buyer 45 miles past the closest store. ChatGPT queried the Mahindra dealer locator for Mohave County and returned a tractor dealer 200 miles away, missing both River Rat locations. And the same dealership came back at 2.3, 3.5, 4.1, 4.3, 4.4 and 4.8 stars depending on which aggregator the engine happened to read. Being findable and being preferred are different problems. Here's the fifth failure mode we've documented, with every transcript published unedited.
AI Can Find Victory SK.
Thirty prompts across ChatGPT, Claude and Gemini. Victory SK scored 11. Every point came from a question that hands the engines the company name or the DOT number. Every question asking for a category in a territory returned zero — while the engines named roughly sixty other carriers and described Victory SK's exact business model doing it. The engines can look this company up. Nothing on the open web gives them a reason to.
The Default Stack: AI's Recommendations Are Quietly Building Billion-Dollar Companies. Here's the Mechanism — and What It Means for Yours.
Ask an AI to build you an app and watch which tools it reaches for — the same stack, nearly every time. Being that default answer has become a measurable growth channel, and the receipts are now public: 10% of Vercel's signups arriving from ChatGPT, a Fortune 500 CEO calling AI answers her fastest-growing channel on an earnings call, conversion rates that embarrass organic search. We break down the five mechanisms that form a default — and why, outside developer tools, the cement in your market is still wet.
AI Knows Exactly Who Mount Holly Nissan Is
We ran 10 prompts a real South Jersey car buyer might type through ChatGPT, Gemini, and Claude — 30 data points, every transcript published unedited. Mount Holly Nissan scored 18/30, our best result yet, hiding the strangest pattern we've captured: every identity question went a perfect 12/12, every category question went 0/9, and retrieval fired on all of them. The engines know exactly who this dealership is. That's the problem.
Troncalli CDJR Is Invisible to AI — Except When It Isn't
Ask an AI assistant for the best Jeep dealership in Cumming, Georgia, and Troncalli CDJR comes back first — called "the standout," with the right rating and the right review count.
Ask the same assistant where to buy a Jeep Wrangler in North Georgia, and it names twelve dealerships. One in Dalton. One in Rome. One in Chattanooga, Tennessee. One entry isn't a dealership at all — just a suggestion to "check Jackson County."
Troncalli isn't on the list.
We ran 10 prompts to find out why. The answer splits along one line: whether the model looked them up or remembered them. When it searched, Troncalli scored 43 out of 48. When it worked from memory, they scored zero across four prompts — and had an address invented for them in a fifth.
All 10 full responses are included below.