Decades of market leadership. AI recommends the newcomer instead.
They built the category. Buyers already trust them, analysts already cite them. Ask AI to recommend a vendor, and it names a competitor with a fraction of their history. The audit found why. Months after the fix, we tested again.
The homepage said one thing. The public record still said another. AI trusted the record.
Two surfaces. AI weighs them differently.
A homepage fix cannot repair a broken record. The headline is the layer AI reads live, at the moment someone searches. The recommendation question runs through a different layer, the one that decides who deserves the sale before a buyer ever asks.
Most companies fix the surface they can see, call it done, and never learn the recommendation question runs somewhere else entirely. Which surface gets fixed first determines whether the fix means anything to a buyer.
Two clocks, running at different speeds.
Clock 1: What AI looks up. AI reads the current web at the moment someone asks. A new headline shows up in results within weeks. This is the clock the homepage fix reached.
Clock 2: What AI remembers. Underneath live lookups sits a record other systems treat as the file of who a company is. It’s the one AI checks when the question is who to recommend. It updates on a slow cycle, months, not weeks. A headline change cannot reach it.
The gap between the two clocks is the whole story here. The homepage moved fast because it lives in the first clock. The recommendation stayed stuck because it lives in the second, and nobody had touched it in years.
The category ranking came from a surface AI reads live. The recommendation loss reflects what a homepage cannot reach. An outdated file still sits in the record AI trusts most when the question is who to recommend.
AI rewards the surface you fixed. It does not automatically extend that fix to the surfaces you never touched. A homepage rewrite is real progress, and a partial one. It is easy to mistake the first for the second.
The question to ask before calling a fix complete: does AI’s answer change when you ask it to recommend a vendor? Not just describe your category. If the category question moved and the recommendation question didn’t, there is a second surface nobody has fixed yet.
This isn’t an isolated case. See the Industrial B2B Index snapshot: 120 vendors, 84% invisible →
Is this happening to you?
The AI Visibility Snapshot runs your company through the same diagnostic. Five buyer-intent queries across three AI platforms, scored against all seven Stack layers. You receive a written finding within 48 hours: which layer is breaking first, what AI is extracting today, and what it is costing you.