Answers, written to be quoted.

Three things, concretely. We optimize for answer engines as a first-class channel rather than an afterthought, with a fixed prompt set and share-of-answer reporting. We build sites so the words are in the page itself and every fact is labelled and linked into one description of the business, because the programs that read sites for AI do not run the code a browser runs. And we use AI heavily inside our own production, for research, drafting, variant generation and QA, which is why our output per dollar does not match a traditional retainer. What we do not do is let a model make the final call on a claim, a price or a design.

AEO, Answer Engine Optimization, is the broadest term: appearing in direct answers on any answer-capable surface, including voice assistants and featured snippets. GEO, Generative Engine Optimization, is the generative subset: being cited inside an AI-written response. LLMO, Large Language Model Optimization, is the same discipline named from the technology side, focused on how a model retrieves and attributes content. Their tactics overlap by roughly 80%, and we run them as one program rather than three retainers.

Yes. Google AI Overviews draws from the Google index, so a page that cannot be crawled and indexed cannot be cited. Classic organic also still carries the majority of trackable non-paid traffic in most categories. SEO now funds two outcomes rather than one, which makes it a better investment than it was, not a worse one.

Share of answer usually starts moving between six and ten weeks, with a meaningful shift by month four to six. Paid media shows signal within two to three weeks once the conversion tracking is trustworthy. A site build takes four to twelve weeks and its effect on conversion is visible immediately. We ask for a six month minimum on the organic and AI visibility retainer because third-party corroboration, the slowest lever, is also the one that matters most to ChatGPT.

No, and nobody can. Model outputs vary by prompt, session, region and model version, and no provider sells placement. What we commit to is moving share of answer against a fixed prompt set, showing you the monthly transcripts, and telling you which competitor holds the slot when we do not.

Yes, for research, drafting, variant generation and quality checks, and we are open about it. A human edits every piece and owns every factual claim before it ships. On paid social and display we apply the platform disclosures now required: Meta’s "Made with AI" label on photorealistic imagery and Google’s "AI Generated" label on primary AI-created elements.

You do. Code lives in an account you own, ad accounts and analytics stay in your name, automations are documented and handed over, and there is no license or hosting you are obliged to buy from us. If we part ways, everything stays with you and still runs.

Mostly teams between roughly five and two hundred people who already have demand and want to compound it. Below about $3,000 a month of paid spend we will tell you to put the money into the site and organic visibility first rather than take the retainer.

Frequently, and it is usually the best value arrangement. We run the technical and architectural layers, your team keeps the voice and the domain knowledge, and we handle briefs, editing, schema and publication mechanics.

Fixed project fees for builds, monthly retainers for ongoing programs, and a percentage of spend with a minimum for paid media. Indicative starting numbers are published on the pricing page. The real number comes after the audit, because an account with a broken conversion signal costs more to fix than one that is merely underperforming.

The answer audit and the technical audit, both delivered with raw evidence rather than a summary. Then the architecture work: entity model, cluster map, schema layer and measurement plan. You should be able to see exactly what we are doing and why by day thirty, before most of the money is spent.

We do not buy links, spin content, run undisclosed AI creative where a platform requires disclosure, or promise placement in a model output. We will not build an automation for a task where a wrong answer is expensive and invisible without a human approval step in it, and we will say no to a retainer we do not think will pay for itself.