AI search (GEO) growth, run for you.

The question that decides the channel

Open an assistant and ask it what to use for the problem your product solves. Whatever it names, that is your current position in this channel. There is no dashboard behind it and no place to submit a listing.

That single test is why founders take GEO seriously faster than they take SEO seriously. It is immediate, it is specific, and the competitor named instead of you is usually one you have already lost a deal to.

Why it matters now

Two figures are worth citing, both from named sources, because most of what circulates about AI search volume is vendor-generated.

Pew Research Center found that users clicked through to a result in 8 percent of visits where a Google AI summary was present, compared with 15 percent of visits without one. SparkToro's zero-click analysis put the share of searches ending without a click at 68.01 percent.

Neither number tells you that search is dead. What they tell you is that the answer increasingly gets delivered in the interface, and that being the thing named inside it is a different asset from being the tenth blue link.

Where the answer actually comes from

An AI answer draws on two things: what the model absorbed during training, and what it retrieves at question time. You cannot edit the first. You can influence what exists to be retrieved, and what the next training pass sees.

Look at the right-hand column. Almost none of it is a publishing task. Most of it needs a person who is an accepted participant somewhere, or who is willing to file a correction with a third party and follow up.

How we run it

The AI layer builds the prompt set from how your buyers actually describe their problem, runs it across the assistants your market uses, and records what gets named and what the answer appears to be leaning on. It also drafts the pages and the corrections.

An operator does the placement. They participate in the communities that get retrieved, as themselves, disclosing the relationship, which is the same discipline we apply on Reddit and for the same reason. They publish and maintain the comparison and alternatives pages. They keep third-party listings consistent, because inconsistent descriptions of what you do produce vague answers. And when an assistant states something wrong about your product, they trace it to the source and get the source corrected, which is usually a person to email rather than a form to fill in.

Measurement, honestly

This channel has a genuine measurement problem and we would rather name it than paper over it.

The same prompt asked twice can give different answers. Assistants personalise, retrieve differently, and change models without notice. So a screenshot proves nothing, and any vendor showing you a single flattering answer is showing you noise.

We sample instead. A fixed prompt set, run on a schedule, reported as share of runs where you are named, alongside which competitors appear and how often. That number moves slowly, and it moves for reasons you can trace back to a specific source we placed.

Expect six to ten weeks before it moves at all, because the placements that feed it are community and third-party work that cannot be rushed without becoming the thing we refuse to do.

What we will not do

We will not create fake discussion, fake reviews, or sock-puppet threads to feed a model. It is dishonest, it is detectable by the communities involved, and losing those accounts costs you the channel permanently.

We will not claim to influence a model's training data directly. Nobody can. We influence what exists in public about you, which is the only real input.