The client
A European industrial manufacturer of battery management systems (BMS): the electronics that monitor and protect the batteries of heavy electric vehicles, off-road machinery and energy storage systems.
Its product holds an automotive functional safety certification, assessed by an independent body. A strong selling point, but a very technical one.
It sells to engineers. Teams that compare suppliers for months, read datasheets and, more and more, ask AI assistants. Its priority markets are Germany, France and Spain.
cronuts.digital came in as a strategic partner to design and launch the new website and run its SEO and GEO strategy.
The challenge
The old website was a WordPress site with only 8 pages visible on Google. Only people who already knew the brand found it. Practically no product-related search showed the website.
Behind that sat four problems that were hard to see.
- Closed to AI. Its robots.txt blocked every crawler, including those of ChatGPT, Claude and Perplexity. AI assistants could not read it.
- No measurement. The inherited analytics tag had not worked for more than a year and a half. Nobody knew who arrived or what they did.
- A category AI did not understand. When asked for suppliers of its certified product, Google’s AI answered with chipmakers and large automotive suppliers. It did not cite a single manufacturer in its category.
- An ambiguous name. Its acronym matches other well-known brands and products. Search engines and AI models needed to know who was who.
How we did it
An agentic website works for three audiences at once: people, search engines and AI agents. We built it in four layers.
Layer 1. Fast, light and organised by industry
A new website with a static architecture: 28 pages, three technical pillar pages and five industry solutions. Fast from day one: 98/100 on mobile PageSpeed, 99/100 on desktop and 100/100 for accessibility at launch.
Layer 2. Technical content built to be cited
Before writing, we analysed what Google and its AI answered for the key searches. The finding: no manufacturer in its category owned those answers. That gap set the content plan.
The articles speak engineer to engineer, signed by a senior engineer at the client. Each one opens with the key takeaways and backs every claim with a figure or a standard. That is the format search engines and AI models extract and cite.
Layer 3. The agentic layer
This is the difference from a standard corporate website. The website introduces itself to machines:
- Open to AI: a robots.txt that explicitly lets in the crawlers of ChatGPT, Claude and Perplexity.
- llms.txt: a summary of the company, its product and its credentials in a format designed for language models, which also makes clear what the brand is and what it is not.
- Structured data for the company, product, articles and FAQs, so search engines and models understand who makes what.
- Forms that never lose an enquiry: each submission is saved on the server first, then in a lead log, and only then is the email alert sent. A custom panel with two-factor login handles leads and article publishing.
Layer 4. Measurement from scratch
We set up Google Analytics 4 and Tag Manager with consent, Search Console with its AI features report, and Bing Webmaster Tools.
And we cleaned the data. An automated search, without a single click in its history, inflated the old website’s impressions: in September it was 70% of the total. We removed it before comparing. Every month we count only real enquiries, with no internal tests or spam.
Results
The new website went live on 21 September 2026. October was its first full month, and we compared it with the old website.
Google now links it to its product
With the old website, almost every search that showed it was a variant of the name. In October, 53 distinct non-branded searches already find it.
The technical search for its certification is already on page one. And the website now competes for the industry’s generic search, “battery management system”, with 1,900 searches a month.
One detail worth reading carefully: the average position gets worse. It is not a drop. New generic searches come in from low positions, and that makes the average worse.
AI already recommends it as a supplier
Appearances in AI Overviews and Google’s AI Mode went from 3.8 a day with the old website to 14.1 a day in October: 436 in the month, across 22 different pages. Germany, its priority market, is already the third country.
We checked it live on 11 November 2026, in English and from Germany. When asked for suppliers of its certified product, Google’s AI Mode names the client as the first option and uses one of its articles as a source. In September, the same search answered with chipmakers and large automotive suppliers.
When asked for European suppliers in the same category, the client comes second of four, between its two direct market references.
The articles start to climb
- 4.7x blog impressions in one month: from 161 to 761. The blog is already 38% of the website’s visibility.
- A new guide to BMS earned 263 impressions in 16 days and started competing for the industry’s generic search.
- The article on the certification requirements went from 90 to 297 impressions, at position 6.3.
- 4 in 10 appearances in Google’s AI already lead to a blog article.
From visibility to enquiries
October brought the first qualified sales enquiry through the website: a European energy storage company interested in BMS development and production. One of its target industries.
And people arriving from Google read: 7 min 18 s average visit length and 71% of visits that engage. After the home page, the most viewed pages are product, blog and certifications: the path of someone evaluating a supplier.
Five lessons for an industrial agentic website
- Open the door to AI. An inherited robots.txt can lock out ChatGPT, Claude and Perplexity without anyone knowing. It is the first thing to check.
- Fill the gap AI cannot answer. If AI cites companies that do not compete with you, a better source is missing. Write it.
- Tell machines who you are. With an ambiguous name, llms.txt and structured data stop you being confused with another brand.
- Clean the data before comparing. An automated search was 70% of impressions. Without removing it, the best month would have looked like a drop.
- Look beyond average position. When generic searches come in, the average gets worse even as visibility grows.
Is your industrial website ready for AI?
Request a free digital audit. We review how Google and AI assistants see you and tell you, with data, what is missing.
And if you want to take AI beyond your website, see how we build AI agents for business.
Frequently asked questions
What is an agentic website?
An agentic website is built for three audiences: people, search engines and AI agents. Beyond design and content, it publishes its information in formats models understand, such as llms.txt and structured data, lets AI crawlers in and can expose actions to agents through protocols such as MCP.
Why does an industrial B2B company need an agentic website?
Because its buyers are engineers who compare suppliers for months and increasingly ask AI assistants. If AI cannot read the website or does not understand what the company makes, it recommends others. In this case, Google's AI went from citing no manufacturer in the category to naming the client as the first option.
What is the difference between SEO and GEO?
SEO works on a website's position in Google's classic results. GEO (Generative Engine Optimization) works on getting it cited in the answers of ChatGPT, Gemini, Copilot or Google's AI Mode. In this case both were worked on together, with the same technical content.
How long does a new website take to show results on Google?
It depends on the industry and the starting point. In this case, the first full month brought 2,009 real impressions against an average of 786 with the old website, 53 non-branded searches against 7, and 3.7 times more daily appearances in Google's AI.
How do you measure a website's visibility in AI?
With several sources at once: Search Console's AI features report for AI Overviews and Google's AI Mode, Bing Webmaster Tools for Copilot, Google Analytics 4 for visits, and live tests with real questions. Before comparing, it pays to remove automated searches that inflate impressions.
