SEO and GEO for a small-scale crisis care home for young people

In short

Credo Jeugdzorg is a small-scale crisis care home for young people aged 10 to 18 in Bennebroek, in the Dutch province of North Holland. The goal was to be found by referrers and local authorities, in Google and in AI answers. I first measured how Google and AI systems saw the organisation, then put the business details, frequently asked questions and internal links in place, and measured again after four weeks with the exact same questions. Now 5 of the 10 pages are in Google instead of 1, and Perplexity no longer confuses the organisation with another one that has an almost identical name.

Key facts

About the client

Credo Jeugdzorg takes in young people who can no longer live at home or need somewhere else for a while. Each young person has their own living space, with round-the-clock care and one-to-one support from SKJ-registered professionals (SKJ is the Dutch register for youth care workers). Referrals come from local authorities and professionals, not from parents. Youth care is YMYL (topics where a wrong answer can cause real harm), so every sentence has to be right.

The question

Credo wanted to be found by professionals who urgently need a place for a young person. They search by situation and region, such as “crisis care young people Haarlem”, and increasingly they ask an AI assistant instead. So the brief covered SEO (being found in Google) and GEO (being mentioned in answers from ChatGPT, Perplexity and Google AI Overviews).

Starting point

I ran a baseline on 24 and 26 August 2026. The site is built in Lovable, a tool that generates a React app. Without JavaScript you get an almost empty page of about 2,235 bytes, with the same title on every route. A Google site search returned 1 of the 10 pages, only the homepage.

Across thirteen fixed questions in Perplexity, Credo was mentioned 3 times and its own site was the source 0 times. Asked “is Credo Jeugdzorg reliable”, Perplexity answered about a different organisation with an almost identical name. With the village name “Bennebroek” added, the answer was correct. The place name made the difference.

What I did

Measure first, then publish

I ran the baseline when my changes were ready but not yet live, so the starting point was clean. The thirteen questions are fixed, because different questions make comparison meaningless.

Making the organisation recognisable

I added structured data (schema, code that tells search engines who you are): a LocalBusiness with the Bennebroek address, Chamber of Commerce number and AGB healthcare code, linked to the LinkedIn company page. AI systems were already using that page as a source. The address and registration numbers separate Credo from the organisation with the similar name. I also aligned the canonical (the preferred URL of a page) and completed the blog post data.

Answering what referrers ask

The Crisis care and Referral pages each got a block of six frequently asked questions with FAQPage schema: who can refer, how fast, for which ages, what happens after placement. The client approved every question. Ten internal links now point to those two pages.

Deliberately left out: a number of available places or an availability pop-up. That number goes out of date within days, and a pop-up gets in the way on mobile and is not read by AI systems.

Making it measurable

I set up Search Console, submitted the sitemap and requested indexing for six pages.

How I checked it

With Lovable, the editor proves nothing. After every publish I checked the live site to see whether the change was really there. For the FAQ blocks I checked that the questions are visible and that the schema matches them.

What Lovable serves to a search engine cannot be tested from outside. A plain fetch (loading the page without a browser) still returned the empty page on 3 September. The proof came indirectly: Google’s AI Overview for the brand name quotes sentences that only exist in the fully loaded page. So Google does see the site. For other AI crawlers that has not been shown.

I measure indexing with a site search. The URL Inspection tool in Search Console lagged behind: a page that already had impressions was still listed there as not indexed.

For AI I ask the same thirteen questions word for word each time: Perplexity logged out, ChatGPT in a temporary chat without personalisation, Google with language and country set to the Netherlands. For each question I record whether Credo is mentioned and which source the system uses.

Results

Re-measurement on 23 September 2026, four weeks after launch, against the baseline from late August 2026:

Search Console, 26 August to 20 September 2026: for “credo jeugdzorg” the site ranks at an average position of 1.0.

Not there yet: for searches without the name, the site sits between position 35 and 75. New mentions on general questions in Perplexity come through an external profile, not the site itself. The Referral page is not in Google yet.

What comes next

I keep working on the sources AI systems use right now. The Referral page also needs to get indexed. Around 18 November 2026 I will measure again, with the same questions and setup.

Frequently asked questions

Can a Lovable site be found in Google?

In this case, yes: Google sees the full page, even though the raw HTML is empty. Check it on the live site and with a site search, not in the editor.

How do you measure whether an AI assistant mentions you?

With fixed questions your clients actually ask, word for word the same each time. For each question I record whether you are mentioned and through which source. Here that was thirteen questions, before launch and after four weeks.

Why work on profiles outside your own site?

Because the measurement showed where AI systems get their information. For Credo, the new mentions came through an external profile. That makes it the quickest win.