Here’s an experiment worth ten minutes of any founder’s week. Take a genuine hero SKU — the one that pays the wages — and ask three AI assistants the question a real buyer would type. Nothing clever. For a supplements brand: “what’s a good magnesium supplement for sleep and recovery?” Put it to ChatGPT, Perplexity and Gemini, one after the other.
When we ran exactly this for one of our customers’ bestsellers, the result was: three assistants, three different answers, and the product showing up in exactly one of them.
That alone is worth sitting with. If this were Google, a result appearing on one engine and missing on two others would be a five-alarm fire — someone would be in a meeting about it by lunch. Because it happened inside a chat window, nobody at the brand knew it had happened at all.
The follow-up is where it gets worse
Then do the thing most people skip: ask the follow-up a picky buyer actually asks. What dosage? What form — citrate, glycinate, oxide? Is it third-party tested?
The answers degrade. Not dramatically wrong, which would at least be easy to catch. Just vaguer — and in places subtly off: a dosage lifted from an old blog post, a “tested” claim stitched together from a marketplace listing that no longer matches the current product. The assistant isn’t lying. It’s doing its best with a set of sources that quietly disagree with each other.
None of this shows up anywhere the brand can see. No ranking dropped. No traffic dipped. No line on a dashboard so much as twitched. The whole thing — the miss on two engines, the degraded follow-ups, the stale dosage — happened in a room the brand isn’t allowed into and doesn’t measure.
Search told you when you were losing. This doesn’t.
The thing everyone in ecommerce quietly took for granted is that the discovery channel tells you when you’re losing. Search does this beautifully: rankings fall, impressions drop, click-through sags, and the numbers arrive before the revenue does. You get a warning. Most of SEO is built on that feedback loop — you can see the wound.
The AI channel doesn’t give you the wound. It gives you silence. A buyer asks, gets an answer that isn’t yours or is a muddled version of yours, and moves on. There’s no impression to count, because you were never in the set. You don’t lose the sale in a way that registers as a lost sale. You simply never enter a consideration you didn’t know was happening.
That asymmetry is the heart of the problem. Not that AI is a magical new channel — but that it’s the first major discovery surface in twenty years that doesn’t come with its own alarm system. The old world handed you a dashboard for losing. This one makes you go looking.
Search isn’t dead; that’s not the claim. There are now two discovery systems running side by side, and most brands are fully instrumented for one and completely blind in the other. The blind one is the one growing.
Run it yourself
Take your own bestseller — the real one, not the one you wish were the bestseller. Ask an assistant what it would recommend for the job that product does. Then ask the follow-up a fussy customer would ask. Read what comes back, across ChatGPT, Perplexity and Gemini.
Most founders go quiet halfway through. Not because the answer is a disaster — it rarely is. Because it’s the first time they’ve watched a channel they had no idea they were already competing in.
Instrumenting it continuously — engines queried on a schedule, citations tracked, misdescriptions flagged — is what Geoffy’s Monitor capability exists for; it’s in early access today. But the first look costs nothing but ten minutes and a little comfort.
Get the first look done properly: your free GEO Score.