Try this: open an AI assistant and ask: “Who offers [what you do]?” Then read the answer as if you were a stranger.
Are you in it? How are you described? And where did the AI get that from?
The question isn't just a game. When people hand the search over to an AI and only decide afterwards, much has already been decided before anyone clicks.
From ranking to being cited
For twenty years, visibility meant a position in a list of links. With AI answers, it's about being named as a source, right in the middle of a summarised answer. The industry calls it GEO, AEO or something else, the terms keep changing. It always means the same thing: a system pulls together a picture of you from many sources and passes it on.
An honest note first: most of what's written about this comes from companies that sell exactly this optimisation. I stick to what several sources repeat and what makes sense even without them.
Four signals that keep coming up
Proof from others. Mentions, reviews, articles about you seem to count more than what you say about yourself.
Consistency. If your website says one thing, your profile another and a directory a third, the picture blurs.
Clarity. What exactly do you offer, for whom, at what price? The easier that is to read, the easier it gets passed on.
Real expertise over time. A web of connected work reads as skill. A single article doesn't.
None of this is mysterious. It's roughly what a careful person would check before recommending you.
Your trail is its context
A speaker at the Just Product conference described how to get good work out of AI agents: give context, set boundaries, check the result. She meant your own agents. But the same logic applies when someone else's agent researches you. Your public trail is its context. If it's thin, scattered or contradictory, it works with whatever it finds.
Visibility as proof
For a long time, my work was done for me as soon as the project was delivered. The decisions, the alternatives, what I learned along the way, I never showed. Today I see it differently. Showing the process is less self-promotion than evidence. Here's how:
- Name the decision you made.
- Show the alternative you didn't take.
- Lay open the tension or the compromise.
- Share what you learned.
Do that regularly and you leave traces that a human and a machine can read as proof.
Done but invisible is the same as non-existent to an AI.
What we don't know
Quite a lot. There's no standard for how agents recommend or buy. Classic search signals seem to keep counting. And the surveys on the trust gap show that for many, a human still decides in the end. I wouldn't rebuild a business on assumptions about agents.
Three basics instead of panic
- One sentence. What do you offer, for whom? So clear that every AI and every human repeats it the same way.
- One trail. Show decisions and learning, not just results.
- One outside voice. At least one person or publication that talks about you.
And then run the test again in a month.