Technical SEO and GEO: Found by Google, Cited by the AIs
Crawlable for Google, quotable for ChatGPT and Perplexity: one discipline, from hreflang to llms.txt.
A React or Next.js site can be fast, beautiful, and invisible. The framework ships an empty document and fills it with JavaScript; the crawler that decides whether you exist may never run that JavaScript. Most of the ranking problems we are called about are not about content at all.
What we actually fix
- Pages a crawler can read without executing your app: prerendered or server-rendered HTML with the real text, not a shell.
- One URL per language, each canonicalising to itself, paired through hreflang in the head and in the sitemap. Two languages on one URL means one of them does not exist.
- Structured data that earns rich results: article, FAQ, breadcrumbs, service and organisation, kept in step with what the page actually says.
- Internal links that exist in the HTML, not only after hydration. A page nothing links to is a page nothing sends authority to.
- Core Web Vitals where they are actually lost: third-party scripts, unoptimised hero images, fonts that block the first paint.
- The measurement that tells you whether any of it worked, rather than a monthly report nobody reads.
The order we work in
- Indexing first. There is no point tuning a title on a page Google cannot fetch, render or reach.
- Then structure: URLs, languages, canonicals, sitemap, the hubs that group your topics.
- Then the content you already have. Updating pages that almost rank beats publishing new ones that nobody will find.
- Then the pages that are missing entirely, usually the ones your buyers search before they buy.
- Then speed, because it moves conversion as much as ranking, and it is measurable the same week.
What we will not do
- Buy links. It works until it does not, and the recovery costs more than the traffic was worth.
- Publish machine-written pages at volume. Answer engines and readers are converging on the same standard: say something true that is not already said better elsewhere.
- Build city pages for cities you do not serve, or service pages for services you do not sell.
- Report on impressions when you asked about revenue.
GEO: the same work, aimed at the engines that answer
A growing share of searches never produces a click: the answer is assembled by ChatGPT, Perplexity, Copilot or an AI overview, from sources it chooses. Industry studies put the majority of searches at zero clicks, and a first position loses most of its usual traffic when a generated answer sits above it. Generative engine optimization is the discipline of being one of the sources that answer is built from - cited, named, linked.
- Write claims an AI can lift whole: a number, a definition, a threshold, in one standalone sentence. The research paper that coined the term GEO measured visibility gains of up to forty percent in generated answers from adding quotable statistics, citations and expert quotes.
- Be readable without JavaScript. Most AI crawlers never run your app; if the text is not in the HTML, you do not exist to them.
- Publish an llms.txt: what the company is, what it sells, where the substantive pages are, in one fetch.
- Be in Bing's index, which feeds ChatGPT Search and Copilot, and tell it about new pages through IndexNow instead of waiting to be found.
- Keep one entity story: the same name, the same description, the same facts on the site, the directories and the social profiles. Answer engines weigh consistency the way link graphs weigh authority.
- Date what you update. A model choosing between two sources takes the one that shows its freshness.
The proof is this site
Everything above is running here. Every page is prerendered into real HTML, readable without a line of JavaScript, in English and French, each language on its own URL with the twins declared in the head and the sitemap. Articles carry article, FAQ and breadcrumb schema; the site publishes an llms.txt and pings IndexNow on every deploy; fonts and cover images are served from this domain. You can view the source and check all of it in a minute, which is more than most agency sites survive.
- 2-4 weeks for the indexing and structure pass
- 90 days before content work is fairly judged
- 0 positions we guarantee
Technical SEO is not a growth hack. It is making sure the thing you already built can be found by the people already looking for it.
Frequently asked questions
How long before we see results?
Indexing fixes show within days: pages that were invisible start appearing. Speed work shows the same week. Content and authority take about ninety days to judge fairly, because that is roughly how long it takes for a page to settle into its real position. Anyone promising ranked results in three weeks is either buying links or describing a keyword nobody searches.
Do we have to rebuild the site?
Almost never. Most of what blocks a modern site is fixable inside the framework it already uses: prerendering the routes, correcting canonical and hreflang, adding the schema, unblocking the render. We only recommend a rebuild when the platform itself cannot produce crawlable HTML, and we will show you the specific limitation rather than assert it.
Is GEO a separate service from SEO?
No, and be wary of anyone selling it as one. The foundations are identical: crawlable text, structured data, clear claims, one consistent entity. What changes is the output you measure - citations and mentions in generated answers, traffic referred by AI platforms, coverage in Bing - and we report on those alongside the classic rankings. One engagement covers both, because it is one discipline.
Is AI-written content a problem?
The volume is the problem, not the tool. A page that says something true, specific and not already said better elsewhere works regardless of how it was drafted. A hundred pages assembled from the same average of everyone else's writing dilute the site that publishes them, and increasingly they simply do not get cited by the answer engines your buyers now ask.
How do we get our company mentioned by ChatGPT?
Make the site readable without JavaScript, allow the AI crawlers in robots.txt, publish an llms.txt that states what you do in one fetch, get into Bing's index through IndexNow, and phrase your key facts as standalone sentences an engine can lift whole. Those five steps are the audit we run first; most sites fail at least two of them, which is why their competitors get named instead.