Problem:
Buyers now hear about companies from ChatGPT, Claude, and Perplexity before they ever type a domain into a browser. Those engines only mention sites they can find, parse, and trust enough to cite. A business that is invisible to agents never enters the shortlist, no matter how well the human site converts once someone arrives.
Most marketing sites were built for people and for classic crawlers. They lean on JavaScript-rendered copy, thin service blurbs, missing canonicals, and no machine-readable account of what the company actually does. Answer engines skip that kind of page, or they invent a summary from a third-party directory that you do not control.
The gap is not a keyword list. It is whether an agent can discover the offer, understand the scope, and take a next step — usually toward Contact — without guessing. That is the work of Answer Engine Optimization, and it is a different brief from dressing up last year's SEO checklist.
Approach:
Carbonology prepares the website for agent discovery: structure, content, and signals so answer engines can parse what the business does and cite it. The pass starts with what a model can actually retrieve. We look at robots and content signals, sitemaps against canonicals, titles and descriptions, on-page depth, and whether the important claims live in HTML rather than in a canvas the crawler never sees.
Then we write the pages that agents need in order to describe you accurately. Service URLs get real problem, approach, and deliverable copy instead of a hundred-word brochure. Organization and Service schema name the provider, the offer, and how to reach you. Internal links connect the essay that makes the argument to the page that sells the work, so a citation can resolve to a page that is ready to be cited.
The same practice covers the discovery layer that sits beside the public HTML: llms.txt, agent notes, well-known catalogs, and markdown alternatives where they help a model read the site as a site rather than as a pile of routes. We do not invent testimonials or inflate credentials. The job is to make the true offer legible. The longer case for why this matters now is in Most internet traffic is no longer human.
Deliverables:
- An agent-readiness pass on the live site: what crawlers and answer engines can see, what they skip, and which gaps actually block a citation.
- On-page rewrites for the money URLs — services, work, contact — so each page can stand alone as a source.
- Technical signals: titles, unique descriptions, canonicals, sitemap alignment, and redirects that do not leak equity through a temporary hop.
- JSON-LD for Organization, Service, and related page types, matched to visible copy so nothing is cloaked.
- Optional agent-facing files (llms.txt, catalogs, markdown negotiation) when they help discovery without confusing search crawlers.
- A written set of next steps you can run, or a build we implement on the existing Next.js site.
Who it's for:
This is for operators whose buyers already start in ChatGPT, Claude, or Perplexity, and for teams that have done competent SEO but still do not appear when someone asks an assistant for a recommendation. It is also for studios and product companies that want the public site to be a source, not a brochure that an agent has to reconstruct from elsewhere.
It is not a promise that every prompt will name you. Citation follows from being findable, clear, and worth repeating. If that is the problem, contact us and we will tell you what a first pass would actually change.
Related:
Questions:
SEO is still about being retrieved and ranked for people. AEO is about being findable, understandable, and citable by answer engines and agents. The two share technical hygiene — titles, canonicals, sitemaps, real copy — but AEO also asks whether a model can describe the business and send a reader to a page that can close the loop.
No honest practice can guarantee a mention. Answer engines cite sources they can read and trust. We make the site a better source. Whether a given prompt names you still depends on the category, the competition, and how the model is retrieving that week.
Usually not. Most of the work is structure, copy, schema, and signals on the site you already have. A rebuild only comes up when the current stack cannot emit the HTML or the routes the offer needs.
The live URL, access to the repo or CMS if you want us to implement, and a plain account of what you sell and who should find it. A first note can be as short as that. Write us from Contact.