- Home
- About
We rebuilt the agency around how people ask now.
Search did not disappear. It moved inside an answer. That single change broke the assumptions under most agency retainers, and rather than bolt a chatbot onto the old ones, we started again from what a buyer does in 2026.
Two readers, one page.
Everything we do follows from this. A page now has two audiences with different needs, and most sites serve only one of them.
The first reader is the person deciding whether to buy from you, and they arrive later than they used to, already briefed, already holding a shortlist a model gave them. They have less patience and a narrower question. The page has to answer it fast and be worth staying on.
The second reader is a machine. AI crawlers now generate roughly 28% of the request volume Googlebot does, and GPTBot and ClaudeBot alone account for hundreds of millions of requests a month. They do not execute your JavaScript the way a browser does, they resolve entities rather than pages, and they lift short factual sentences rather than paragraphs.
Serving one of those readers well and the other badly is the most expensive mistake in digital right now. A beautiful site that a model cannot read gets left out of the answer. A perfectly structured site with nothing persuasive on it gets cited and then loses the sale. We were built to do both, in the same pass, by the same team.
The second thing that changed is production economics. AI is good at research, clustering, drafting, variant generation and quality checking, which is why a small team can now produce what used to need a large one. It is unreliable at knowing what is true about your business. So we use it hard on the first list and never on the second, and a person signs off every claim that ships.
Five things we hold to.
Not values on a wall. These are the rules we lose arguments over, internally and occasionally with clients.
Evidence, then opinion
01Every recommendation arrives with the transcript, the crawl, the account data or the test result behind it. Where we are guessing we say so, and we say how we plan to find out. An agency opinion with nothing under it is the most expensive thing you can buy.
AI drafts, a human owns
02Models fabricate specifics with complete confidence. A person checks every figure, every product claim and every competitor comparison before publication, and cites the source where one exists. This is the rule that costs us the most time and the one we will not move on.
Never fabricate proof
03No invented testimonials, no borrowed logos, no metrics rounded up until they look better. Our case study names are changed under NDA and the figures are not. If we cannot show something, we say we cannot show it.
You own everything
04Code in an account you own, ad accounts and analytics in your name, everything documented and handed over. There is no license, no proprietary platform and no hosting you are obliged to buy. If we part ways, it all still runs.
We will talk you out of things
05Off a channel that will not pay for itself, off a migration you do not need, off an automation for a task where a wrong answer is expensive and invisible. Turning down a retainer we do not believe in is cheaper for us than delivering it.
What we will not claim.
We cannot guarantee a model will recommend you. Outputs vary by prompt, session, region and model version, and no provider sells placement. What we can do is move share of answer against a fixed prompt set, show you the transcripts every month, and name the competitor holding the slot when we do not.
llms.txt is not a strategy. As of early 2026 no major provider has publicly committed to reading it in production and Google has said it does not support it. We ship one because it costs an hour. Anyone selling it as the centerpiece of an AI visibility program is selling you the cheapest part.
AI-generated creative is not universally better. It has been measured outperforming human creative for ecommerce products under roughly $100 average order value and underperforming above it. We treat that as a production decision rather than a belief, and we apply the platform disclosures now required on both Meta and Google.
Some of this is still young. Autonomous campaign-building agents are newer than their marketing suggests, and parts of the lifecycle AI stack are still in restricted release. We will pilot those alongside a program that works without them, rather than build your quarter on a roadmap.
Find out where AI sends your customers today.
We run your brand through the answer engines your customers use, then send you the transcript of what they say about you. Free, and yours to keep whether or not we work together.