Being cited by AI isn't enough: the source trap
A B2B brand just hit #1 in ChatGPT and Google's AI Overviews — for its entire category. A win? Almost. Because the AI cites it… by linking to a dead third-party page. Not its own site. Here's the GEO lesson almost nobody sees.
AI visibility is two jobs — not one
When people talk about "AI visibility," they picture a single thing: being named in the answer. Showing up when a buyer asks ChatGPT or Perplexity "what's the best solution for X." That's genuinely essential — and it's half the job. Only half.
The second job is being the source the AI cites to back up your mention. Generative engines no longer just write a name: they attach each claim to a link, a source, a page. So the real question becomes: when the AI cites you, where does it actually send the reader?
The case that says it all
Take a real example (brand anonymized). A young B2B SaaS platform, nearly invisible in AI answers a few weeks ago, is today cited at the top of its category across ChatGPT, Google AI Mode and AI Overviews — ahead of far bigger legacy players. On paper, a total success.
Except the link the AI shows doesn't point to their site. It points to an old incubator page that's now dead. The engine "learned" their name through that third-party source and keeps citing it — even though the page no longer even loads.
Why layer 2 gets ignored — and what it costs
Nobody thinks about the source because being named already feels like winning. Yet a citation that points anywhere but your own site is expensive:
- The click and the traffic: the interested prospect lands on a third-party (or dead) page, not on your site — so no visit, no conversion.
- Control of the narrative: a dated press article, a directory or an incubator tells who you are — sometimes with stale information.
- Durability: your visibility rests on a page you don't control. The day it changes or disappears, your AI presence wobbles with it.
Become your own canonical source
The fix is to make your own domain the reference the AI prefers to cite. Concretely:
- Structured data (JSON-LD): describe your company, products and services in a format the engines read without ambiguity.
- An llms.txt file: point the models to your most important, authoritative pages on your topics.
- A canonical, up-to-date page per key topic: a clear, complete, recent answer — better written than the third-party source the AI cites today.
- Citation consistency: the same information (name, activity, differentiators) repeated coherently across the web, so the engine consolidates everything toward you.
The goal: give the AI an obvious reason to switch its citation from the third-party page to yours.
Site rebuilds: the riskiest moment — and the best opportunity
If you're rebuilding or migrating your site right now, tread carefully. It's the riskiest window: changing URLs, structure, or losing old pages can wipe out the few citations the AI already attributes to you. But it's also the best opportunity: you can bake in structured data, llms.txt and canonical pages from day one — instead of retrofitting everything later. A well-planned migration turns a hazard into an accelerator.
In short
Being named by AI clears the first gate. But if the link points to a dead or third-party page, you stay invisible where it counts: on your own site. The two jobs go together — get named, and be the cited source.
The AI already knows who you are. The question is whether it knows where to send people.
Where does AI send your future customers?
We test what ChatGPT, Gemini and Perplexity say about your market — and, crucially, which sources they cite. Free report within 48h.
Get my free audit →Also read: Why ChatGPT doesn't recommend your business (and how to change that)