Your Inventory Feed Was Built for Humans. ChatGPT Can't Read It.

The pitch this week: AI is everywhere in your dealership now. Meta glasses doing inspections. A free chatbot that prices your used cars. A Swedish SaaS platform promising 30% faster lot turns. A Chinese dealer-tech company reporting 4,100 stores burning 150 million AI tokens a day. It reads like the inflection point vendors have been promising for three years.

The less flattering story, also in this week's news, is that 61% of shoppers say they'll use AI during their next purchase — and most of your inventory is invisible to it.

The feed problem nobody's fixing

Lectrium's research lands on something operators should sit with: dealership inventory largely exists in feeds and databases designed for human browsing and third-party syndication — not for an AI agent to query in real time. When a shopper asks ChatGPT or Gemini for "CPO EVs under $35,000 within 50 miles," the assistant needs a live, structured connection to your actual stock. Most stores don't have one. They have a VDP built for a person clicking through photos, not a machine calling an API.

This isn't hypothetical plumbing talk. Google's Universal Cart and Universal Commerce Protocol, launched at Google Marketing Live in May, and Anthropic's Model Context Protocol, now adopted across major AI assistants, are the pipes that let an agent pull your live pricing and availability instead of a stale cache — or, more likely, a marketplace listing that scraped your inventory better than your own site exposes it. The dealers who don't structure their data for machine retrieval aren't losing a marketing channel. They're becoming invisible at the exact moment a shopper is ready to transact, while a listing aggregator answers the question for them.

Lectrium's suggested audit is worth stealing regardless of vendor allegiance: ask ChatGPT, Gemini, and Perplexity the five questions your best shoppers actually ask, and see whether your store shows up at all. Do it monthly. The uncomfortable stat here is that a third of dealers can't even tell you whether AI traffic is working for or against them. That's not an AI strategy gap — that's a measurement gap, and it's fixable this quarter without buying anything.

Meanwhile, the vendor pile-up

Set against that gap, this week's product news is a useful reality check on what's actually proven versus what's a press release. CardinaleWay Hyundai & Genesis of Mountain View is running Meta AI glasses for hands-free digital multi-point inspections through UpdatePromise's Symphony platform — a specific, named store, doing a specific workflow. That's the kind of detail that separates a real deployment from a concept video. Affinitiv's CapSure is chasing the same hands-free inspection idea for service video. Worth watching for actual before/after cycle-time numbers, not just the announcement.

Lotlinx's LotGPT and sMedia's Smart Offer 2.0 are squarely in "free or cheap AI advisor" territory — low-risk to try, but neither comes with independent proof yet of moving units or margin. DSC's earnings call is the most data-rich item of the bunch: 4,100 dealerships on AI assistants, average daily token use rising from zero to 150 million since March. That's genuine adoption at scale — though it's happening in China's used-car market under DSC's own ecosystem, and most of those tools are still free or subsidized, which tells you adoption is easy when the price is zero. The harder question, which the earnings call doesn't answer, is retention once DSC starts charging for the "digital employee" tier.

What this means for your store

CarDekho's Amit Jain is right that the direction of travel is AI agents handling discovery, pricing, and service retention as specialized functions rather than one big system. But that's a 2027 roadmap conversation. The 2026 conversation is narrower and cheaper to act on: can an AI assistant actually see your inventory today? Run the five-question audit before you sign anything else. A store that's invisible to ChatGPT doesn't need a better chatbot on its own website — it needs its data exposed to the ones shoppers are already using.