The question arrives in almost every first conversation, usually with an edge on it. Should you really optimize what AI says about you? Isn't that gaming the machine?
It deserves a straight answer, not a defensive one. Here is ours.
The worry is legitimate
There is a version of this discipline that should worry everyone. Hidden text in pages instructing models what to say about a brand. Review rings manufacturing consensus that never existed. Slop farms publishing synthetic articles at industrial volume so a claim appears independently corroborated. All of it exists, all of it gets pitched to marketing leaders as strategy, and all of it is an attempt to make engines assert things that are not true.
If that were what AI search optimization meant, the right move would be refusal. It is not what it means, and the difference is the entire point of this post.
Why manipulation loses
We have worked through five migrations since the first AdWords auctions, 26 years of them: search, SEO as a service, social, mobile, and now AI answers. Every migration had its spam wave. Link farms. Keyword stuffing. Follower factories. Each one worked for a while, each one was systematically crushed, and the brands built on them were crushed alongside.
The pattern is not sentimental. It is structural. An answer engine's product is being right. Every commercial incentive its makers have pushes toward better source evaluation, better provenance, better detection of manufactured consensus. Betting on manipulation is betting that the technology stops improving, and that bet has lost in every migration we have watched.
So prompt-stuffing and slop farms will fail, and they deserve to. The answer layer is becoming the place buyers decide. Polluting it is not cleverness. It is borrowing against your own name, at interest.
What optimization actually is
Our stance is simple. AEO done right is making true things legible. Evidence, structure, provenance.
Your analyst recognition is real: put it on an open page instead of behind a form. Your customer results are real: publish how they happened, with the customer's consent, in text a machine can read. Your product does what you claim: say so consistently everywhere your company is described, so the facts agree with each other. Mark up your pages so they are unambiguous. Sign your work so its origin can be checked.
Nothing in that list persuades a model of anything false. It removes the friction between a true fact and a system trying to verify it. The truth was always your best asset. Legibility is the decision to stop hiding it.
This is how we practice AI search optimization, and it is why the work looks less like a trick and more like operations: audits, structure, corroboration, honest readouts against success criteria we set in advance, and kill lines for any tactic that stops earning its keep.
What we refuse to do
Lines we do not cross, stated plainly so you can hold us to them.
No fabricated proof. No invented case studies, no synthetic testimonials, no numbers that did not happen. If the evidence does not exist yet, the work is to earn it, not to type it.
No deceptive schema. Markup describes what the page actually says and what the company actually is. No awards you did not win. No review ratings that do not exist.
No astroturf. No fake reviews, no sockpuppet threads, no undisclosed placements dressed up as editorial. Our own 5.0 on Clutch matters to us precisely because our clients wrote it and we cannot.
We would rather show you a flat readout than a flattering fiction. An evidence-first firm that fakes evidence has nothing left to sell.
Provenance is the trendline
Watch where the engines themselves are heading: more citations shown to users, more weight on sources that can be traced, more scrutiny of consensus that appears from nowhere. We will not pretend to know the exact mechanics inside any model, and anyone who claims to is guessing. But the direction is consistent across every major engine, and it points one way: toward provenance.
Which means the only durable strategy is also the honest one. Be genuinely citable. Keep the receipts where machines and skeptics can check them. Disclose what you do. The gap between what AI says about you and what is true of you is closing from both ends, and you want to be standing on the true side when it closes.
The version worth building
Here is the conviction underneath our whole practice. The answer layer is starting to mint category leaders, the way search and social minted them before. The open question is what kind.
We think the leaders minted in AI answers should be the companies with the receipts: the real recognition, the consistent facts, the customers willing to say so in public. Not the loudest. Not the best funded. The most verifiable. That would be a better outcome for buyers, for the engines, and for every company that has actually done the work.
So, should you optimize for what AI says about you? Ask a cleaner question first. If the engines told the whole truth about your category tomorrow, would you win?
If yes, optimization is simply making sure they can. If no, the problem was never the machine.
Either way, the honest starting point is knowing what they say right now. Ask us about the AI Search Diagnostic, and talk to Danton about the true things worth making legible.