SEO in Vejle: technical SEO, indexing and GEO

SEO is two trades that share a name. One is content and links: keyword research, writing, digital PR, outreach. The other is the engineering underneath it: whether pages load fast enough to be worth ranking, whether Google can crawl and index them at all, whether the markup describes what is on the page, and whether the site tells search engines a consistent story about which URL is the real one. This offer is mainly the second trade.

That is worth saying clearly, because the two get sold together and the technical half is usually the neglected one. A site can have excellent content and still be invisible if its canonical tags contradict each other, if half the pages are excluded by a stray directive, or if the mobile page takes six seconds. Once that is fixed, the content has a fair chance of being found.

Locally I go one step further: keyword research into the searches your customers actually use, up to six pages rewritten for service and town, and the Google profile set up properly. If what you need is a large content calendar and link-building, I am not the right supplier and I will say so in the first email. There are agencies in the region who do it well, and I am happy to name one.

GEO stands for generative engine optimisation, and it is about being visible in an answer rather than in a list of results. ChatGPT, Claude, Perplexity, Copilot and Google’s AI overviews fetch pages the way a search crawler does, but most of them do not run JavaScript, and they answer from the HTML they were handed. A site that assembles its content in the browser is empty to them, even though it looks right to you.

There are five things to do in the code. The pages have to be server-rendered, so the content is there without JavaScript. Robots.txt has to allow the assistant crawlers deliberately. A named group in that file overrides the wildcard rules for that one bot, which makes it easy to shut them out by accident. Structured data has to carry the real figures, prices and services, because an amount with markup around it is easier to quote correctly than an amount in a heading. An /llms.txt can state what the business is, what it costs and what it does not sell. And the crawlers that actually fetch something can be logged, so the question of whether any assistant has ever been past is answered from a log. The site you are reading this on runs all five parts.

Two caveats apply. Nobody can sell you a guaranteed place in an AI answer, and I do not promise one. And the part that depends on sources other than you confirming who you are (Danish company registries, a Google Business Profile, mentions elsewhere) sits outside the code. I write down what is missing, but they are your own accounts to open.

These are the findings that keep coming up, roughly in order of how much they matter. Pages that return 200 with no real content, which Google files as a soft 404 and quietly drops. Canonical tags pointing somewhere other than the page you want indexed. Redirect chains where one hop would do, leaking a little signal at each step. A sitemap listing URLs that redirect, or omitting the pages that actually sell. Images served far larger than they are displayed. A robots.txt or a meta directive doing something nobody intended. That last one costs the most and is the hardest to spot.

Then the speed side, which on most Danish business sites comes down to the same few things: JavaScript loaded on every page for a feature used on one, third-party scripts that block rendering, no caching in front of pages that never change, and fonts that arrive late enough to shift the layout under the reader.

On the AI side a different set of findings recurs. The content exists only inside a JavaScript bundle, so what an assistant is handed is an empty shell. A bot filter at the host or in front of the CDN turns GPTBot and PerplexityBot away long before robots.txt is read, and nobody in the building asked for that. Prices, services and opening hours live in running text or inside a PDF, where they cannot be extracted with any confidence. The markup carries none of the references to other sources that let a model risk saying who you are. And there is no log, so whether any assistant has ever fetched a page can only be answered with a guess.

These are common problems, and finding them does not require a monthly retainer. Somebody has to sit down with the site, Search Console and a real device, and write down what is actually happening.

The engagement is priced after a first measurement of the site and is delivered within a week. What you get is the baseline numbers, the fixes made in the code rather than added as another plugin, and the same numbers again afterwards.

The measurements are Core Web Vitals from field data where there is enough traffic for it, Lighthouse where there is not, index coverage and impressions from Search Console, and a crawl of the site before and after. Rankings are deliberately not the headline metric: they move for reasons outside this work, they differ per person and per device, and a supplier who reports them selectively can show almost anything. Speed and indexing are things I can change and things you can verify.

GEO is measured three ways, because there is no ranking to read off. A fixed set of questions that a customer would actually ask goes to ChatGPT, Perplexity and Google’s AI answers before the work and again afterwards, and the answers are kept so before and after can be compared directly. Traffic referred from the assistants is separated out in the analytics, so you can see whether anyone arrives that way. And the crawler log shows which assistants fetched what, and when.

I also hand over what I changed and why, written for whoever maintains the site next. If that is your own developer or another agency, the report should be enough for them to carry on without me.

I am based in Vejle and take work across the Triangle Region, from Kolding and Fredericia to Middelfart, and can meet in person when that helps. Clients are typically businesses with a website or shop that is not being found even though the content is fine, and that need somebody to go into the code and the hosting and fix the cause.

The other group is marketing and SEO agencies that have produced a technical report but lack a developer to carry out the recommendations. There I deliver the fixes in the site's code, measure before and after, and leave the reporting and the client relationship with the agency if that is what you prefer.

The free website check at micci.dk/en/website-check gives you a first picture in half a minute, and a technical review gives you a ranked list of what is holding the site back. The price of the fixes themselves depends on how much has to change and which platform the site runs on, and you get it in writing before the work starts.