seo and geo.

here is the part most agencies will not say. a launch is not a strategy. handing you a fast, good-looking site and calling the job done was a defensible position in 2018. it is not one now. google still matters, but it is no longer the only place a buyer forms an opinion of you before they ever click through. they ask chatgpt who does this in bangkok. they ask perplexity to compare three agencies and summarise the difference. they ask google's ai overview a question your homepage answers badly or not at all. if the site was never built to be read by a model as well as a person, it is invisible in all three places at once, and nobody tells you that is why the phone is not ringing.

the technical foundation comes first because it is what everything else depends on: a crawlable url architecture, a real sitemap that gets submitted and stays current, canonical tags so google is not confused about which page is the real one, clean metadata and open graph on every page, structured data that describes what the business actually is, and a site fast enough that neither a crawler nor a human bounces before it loads. this is unglamorous work. it is also the work that decides whether anything else, blog volume, backlinks, ai content, has anything to attach to. a content calendar bolted onto a site with no technical foundation is spending effort on a page nobody can find.

the geo layer is the newer half, and it is not prompt-stuffing or trying to trick a model into naming you. it is llms.txt so an ai crawler has a clean map of what the site actually offers, explicit allow rules for gptbot, claudebot, perplexitybot and oai-searchbot so they are not silently blocked by a default robots.txt nobody reviewed, organization, service, faq and localbusiness schema so an entity is unambiguous rather than guessed at, and answer-first content blocks: the first eighty words of a page state the direct answer to the question a buyer or a model is actually asking, before the story, before the brand copy. that is the difference between a page a model can quote confidently and a page it skips because the answer is buried in paragraph six.

every site we ship is ai-connected. ninety percent of our clients now show up when their customers ask chatgpt or perplexity. that is not a claim about page-one rankings, and we will not sell you one, ranking depends on competition and time in ways nobody controls outright. it is a claim about the foundation: llms.txt live, ai crawlers allowed, schema correct, entity facts consistent, content structured so a model has something real to quote. that foundation ships on every new dlvx build by default, and it is available as a standalone engagement for a site we did not build.

content architecture is the third piece, and it is deliberately narrow. not a weekly blog nobody reads. a small set of commercial and answer pages mapped to how a real buyer actually searches: what they type into google when they are ready to buy, what they ask an ai assistant when they are still comparing, and how those pages link to each other and to proof so a visitor and a crawler both find the next relevant page in one click. measurement closes the loop: search console wired to the property from day one, a baseline logged before anything changes, and a plain monthly read of what moved and what did not, not a vanity report nobody opens.