For anyone who invested seriously in SEO over the past few years, 2026 feels like a cold shower. Our clients see their organic traffic fall by 30 to 60 percent. Not because their content got worse. Not because they were penalised by Google. But because the way people search is fundamentally changing.
We want to be transparent about that, not least because we have to solve it on our own site. At the bottom of this post is a link to our live dashboard where you can see in real time which AI bots read our pages. First, let us set out the problem.
What is happening
Someone with chronic back pain used to type “fysiotherapeut Amsterdam beste” into Google, scrolled through three pages, clicked on five sites and chose one. That same person now asks ChatGPT: “I have had lower back pain for a year that will not go away. Which specialist do I need and what should I look out for when choosing?” ChatGPT gives a four-paragraph answer. No clicks. No scrolling. No comparing.
That answer is based on information ChatGPT fetched from somewhere. Sometimes a source is cited, often not. And that is where the shift sits that you see in your Google Analytics as a halved chart.
The same goes for the expat wondering how, as an English software engineer, to get a mortgage in the Netherlands. Or the owner of a small webshop looking into VAT OSS for sales within the EU. Or the parent who wants to know whether their child has ADHD or is simply lively. All questions that used to lead to a long search via Google and a handful of visited sites. Now it leads to one conversation with an AI tool, and at most one click at the end, if the answer is not enough.
The paradoxical answer
The obvious response is: more content, better SEO, higher rankings. But that is mopping with the tap running. The question is not how you rank higher in Google. The question is how you end up in the answer from ChatGPT, Claude, Perplexity and Google AI Overviews.
That is a fundamentally different discipline. Google's algorithm weighs backlinks, keywords and user behaviour. AI tools work differently. They read sites, summarise information, and choose which sources to cite. Whether your site is among them depends on two things.
One: can AI crawlers read your site at all? Surprisingly, for many modern sites the answer is: no. Sites built with React or Vue without server-side rendering often serve a crawler a nearly empty HTML page. The content only appears after the browser has run JavaScript. Browsers do that, AI crawlers often do not. Result: the bot sees nothing of your expertise, and chooses another.
Two: do you offer the content in a way AI tools can process? Short, clear paragraphs. Concrete examples. Questions answered literally in your content, not implicitly touched on. AI tools look for written answers they can cite one to one.
Check your own site
Free, no email required.
What we did ourselves
We ran into this problem on our own site. Hiveminds.nl ran until 12 May as a Single Page Application, nice for visitors, a blind spot for AI crawlers. On 12 May we rebuilt the site to server-side rendering. From that moment every page serves a complete HTML version to every visitor, human or bot.
To make the impact measurable, we measured a baseline before the switch. We ran the last pre-SSR build locally and, with a GPTBot user-agent, requested the same pages we now serve live. The difference is dramatic.
Before the switch, our site served GPTBot an HTML shell of 4,608 bytes: meta tags in the <head>, and an empty <div id="root"></div> in the body. Zero headings, zero paragraphs, zero visible content. After the switch the same bot receives full, rendered HTML, almost ten times as much on the homepage, and up to sixteen times as much on content-heavy pages.
What GPTBot actually receives
HTML response per route, before and after the SSR switch (12 May 2026)
Method: curl -s -A "GPTBot" [url] | wc -c. Pre-SSR measurement on a locally run build of commit 166d63c (the last commit before the Vike migration). The pre-SSR response was identical on all routes: <div id="root"></div> without content.
At the same time we built a dashboard that shows live which bots visit our site and how deep they crawl. Not because it is a nice product. Because we wanted to measure whether the change works. You can view it at hiveminds.nl/lens/live. The vertical line on 12 May marks the moment of the switch. In a few days you will see whether GPTBot, ClaudeBot and PerplexityBot visit more often and read more pages.
This is exactly the kind of measurement you should be able to do for your own site. Not in a month. Not as a hypothesis. But with your own figures.
In concrete terms: where do you start?
If you notice your organic traffic falling and you suspect AI search has something to do with it, there are three steps you can take today.
Step one is to measure where you stand. Which AI tools already mention your brand? Which mention your competitors but not you? Which questions directly relevant to your offering do ChatGPT and Perplexity answer without citing you? These are questions you cannot answer via Google Analytics, that data is not in your traffic statistics. That is what we built Lens for, our free AI search snapshot product. You enter a URL, we ask ChatGPT, Claude and Google AI in your field what they know about your business and where you stand against the competition. You get the report within a few minutes. At hiveminds.nl/lens you can request it.
Step two is to have your site checked technically. Do you serve server-side rendering or static HTML? Are your most important pages actually read by AI bots, or do they see an empty carcass? This is a one-off technical check you can often fix in a day.
Step three is to rewrite your content with AI citation in mind. Not all content, start with your 10 most important pages. Make sure the most-asked questions in your field are answered literally in the content. Short paragraphs. Concrete examples. Figures where possible. No marketing language that means nothing.
The deeper shift
What is happening here is bigger than an algorithm update. It is a shift from search-and-compare to ask-and-answer. For niche experts, coaches, specialists and small service providers that is good news once you get it. Because AI tools reward specific expertise in a way Google never did. Anyone who writes in-depth, accurate, structured content on a defined topic has a better chance of being found in 2026 than in 2024.
But then you do have to be there. And that is where it stands or falls. An excellent physiotherapist who never let AI bots crawl their site is invisible to ChatGPT users. A mediocre competitor with a well-structured, readable site does come forward. That is not justice, that is how the system works.
So the question is not whether you should do something. The question is how quickly you come into view before your competitors do.
Originally published on hiveminds.nl
Want to know where your business stands in AI search? Request a free Lens snapshot. We ask ChatGPT, Claude and Google AI whether they know you and how they position you against your competitors.
Start a free Snapshot