Approach
Most agency relationships fail before the work starts. A scope is priced against a symptom, and nine months later everyone is arguing about attribution. This is how we avoid that: what we do first, what we report, and what we will not promise.
Operating principles
We do not quote a scope on a first call. Every engagement opens with a paid or unpaid diagnostic that establishes what is actually constraining growth, because the presenting problem and the real one are usually different, and a proposal written before that distinction is drawn is a guess with a price on it.
Rankings, impressions and citation counts are diagnostic instruments, not results. Reporting states what changed in qualified pipeline, cost per acquisition or revenue, and shows the search and AI-visibility movement underneath it as the explanation rather than the headline.
Nobody controls Google's ranking systems or the retrieval layer inside ChatGPT, Gemini, Claude or Perplexity. Any firm guaranteeing a position is either misunderstanding those systems or misrepresenting them. We commit to the work, the standard of that work and the measurement, not to an outcome we do not own.
An audit that ends in a PDF ends. Ours end in tickets: the change, the file or template it belongs in, the reasoning, and the acceptance criterion. Your engineers should be able to act on our output without a translation meeting.
Search, strategy, performance, technology, brand and on-ground delivery are staffed in-house across 45 services. When a diagnosis crosses disciplines, as it usually does, the answer is a wider brief, not a referral and a second vendor relationship for you to manage.
Some engagements should be smaller than the client expects, and some should not start at all. We have told prospective clients that their constraint was pricing, sales follow-up or product, and that spending on acquisition first would waste the budget. That conversation costs us revenue and is still the correct one.
Engagement model
The order matters more than the contents. Running phase three before phase one is the single most common reason growth spend produces nothing measurable.
Two to four weeks. We audit the technical foundation, the demand landscape, the competitive set and the current AI-visibility position, how often assistants name you, and who they name instead. We interview your sales team, because they know why deals stall and analytics does not.
Output: A written diagnosis naming the constraint, with evidence.
Every candidate initiative is sized against expected impact, time to signal and implementation cost, then sequenced. Work that cannot be measured does not enter the plan. Work that depends on something else being fixed first is ordered behind it rather than run in parallel and blamed later.
Output: A sequenced roadmap with owners, effort and the metric each item moves.
Execution against the roadmap: technical remediation, entity and content architecture, campaign build, product or platform work. We write the specifications and either implement them ourselves or hand them to your team in a form that needs no interpretation.
Output: Shipped changes, documented, with the reasoning preserved.
Instrumentation comes before the work it measures, not after. We baseline organic performance, assisted conversions, cost per acquisition and share of AI citations, then report against that baseline monthly, including the months where the number went the wrong way and why.
Output: A monthly report a CFO can read without a translator.
Search and AI visibility compound; campaigns do not. Once the foundation holds, the work shifts to widening the surface that earns citations and to defending positions competitors are attacking. This is the phase most engagements are cancelled during, and the phase that produces most of the return.
Output: A quarterly re-diagnosis, and an honest recommendation on continuing.
Flagship practice
Search is where our expertise runs deepest, and the part of the field that has changed most. Treating generative engines as another ranking surface is the mistake we spend most of our diagnostic time undoing.
Ranking well means being placed above competitors on a results page a person then scans. Being cited by an assistant means being selected, from a retrieval index, by a model, as the source worth quoting, and then surviving summarisation intact. The second requires that your claims are self-contained, attributable and consistent across every surface the model has seen you on.
A generative engine resolves your brand to an entity before it decides whether to mention you. That resolution depends on structured data, a stable identifier, and corroboration from sources the model already trusts. We build that layer explicitly rather than hoping it emerges from content volume.
We run recurring prompt panels across ChatGPT, Gemini, Claude and Perplexity for the questions your buyers actually ask, record who is named, and track the delta over time. Without that baseline, any claim about improving AI visibility, ours included, is unfalsifiable.
Boundaries
Published because a list like this is easy to write and expensive to keep. Hold us to it.
Guaranteed rankings, guaranteed citation, or guaranteed lead volume.
Private blog networks, purchased links, or any tactic whose defence is that it has not been caught yet.
Publishing unedited generated content at volume. It gets indexed, then it gets discounted, and the domain carries the cost.
Fabricated reviews, invented case study numbers, or testimonials we cannot attribute to a named, contactable client.
Long lock-ins as a retention strategy. If the work stops being worth it, the contract should not be what keeps you.
Questions
Start here
Tell us what you are trying to move and we will tell you whether we are the right firm for it, including when the answer is no.
