Rank on Google, and get cited by the assistants
We work both fronts that now matter: your ranking in Google’s organic results, and the odds that ChatGPT, Perplexity or Google AI Mode names your company when a buyer asks. The retainer starts at 500 € per month excluding VAT on a three-month minimum, because below three months no search result is measurable, with any agency. Paid campaigns are added only where they earn their place.
Free quote · no commitment · you own the code and every account
The shift
Ranking first is no longer the same as being found
A generated answer now sits above the results, and a growing share of buyers never reach a blue link at all. Your position did not move. The click did.
Someone asks an assistant to recommend a company like yours. It names three. You are not among them, and no analytics tool anywhere recorded the moment it happened, which is exactly what makes this problem so easy to postpone for another quarter. The loss is real and it leaves no evidence you can put on a slide.
There is a mechanical reason it happens more often than it should. Vercel measured how AI crawlers behave and published the results on December 17, 2024, in a study called “The rise of the AI crawler”: none of the major AI crawlers currently render JavaScript. They do fetch your JavaScript files. They simply never execute them. So a site assembled inside the browser hands them an empty shell, however good the writing sitting inside it is. Google’s own developer documentation, updated March 4, 2026, points the same way: “Keep in mind that server-side or pre-rendering is still a great idea because it makes your website faster for users and crawlers, and not all bots can run JavaScript.”
We attach a caveat to that measurement every time we cite it, and you should hold us to it. Vercel ran the test in December 2024. Nobody has publicly reproduced it since, and the later articles that state it as current fact are all restating that one measurement. It remains the best evidence available. It is not a permanent law of the web, which is precisely why we run the check on your site instead of quoting a study at you.
We published the whole protocol: six steps, roughly ten minutes, every source linked and dated. Run it yourself before you talk to us.
What the retainer covers
Four workstreams, run in parallel
They feed each other. Content nobody can crawl is wasted, and a crawlable site with nothing worth quoting on it gets crawled and ignored.
Pages that arrive complete
Server-rendered HTML, page structure, structured data, indexing, and load times under 2.5 seconds on a mid-range phone over 4G. A crawler that never runs JavaScript sees whatever your server sent, and nothing else.
Writing built to be quoted
Articles that answer what your buyers actually type, with the answer first, every figure dated, and paragraphs that still make sense once they are pulled out of the page. That last property is what makes a passage citable at all.
What the rest of the web says about you
Listings, industry directories, LinkedIn, YouTube, press. Assistants lean heavily on what other people say about a company and only partly on what the company says about itself, so this is slow work that nothing on your own site replaces.
Measurement, and the decision to stop
Search Console, Bing Webmaster Tools, and a fixed set of buyer questions asked every month in ChatGPT and Perplexity. Paid campaigns go in only where they earn their place, usually a launch or a market too crowded to wait out. Anything producing nothing gets cut.
First ninety days
What happens before you have any right to expect results
Three months is the floor below which nothing is measurable, so here is what fills it.
- The raw HTML your server sends, read on every template rather than on the home page alone, exactly as a crawler that runs no JavaScript receives it
- Your robots.txt checked line by line for GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot, because those settings are independent of each other and most sites got at least one of them wrong by accident
- Server-side rendering or prerendering put in place wherever pages are currently assembled in the browser. Your design, your copy and your URLs all survive that change
- Load times brought under 2.5 seconds on a mid-range phone over 4G, measured on the handset rather than on a laptop plugged into fiber
- Search Console and Bing Webmaster Tools connected, including the generative-AI performance report Google opened on June 3, 2026, if your property is inside the subset it reached
- A fixed list of buyer questions asked every month in ChatGPT, Perplexity and Google AI Mode, where we record the cited sources rather than the answers
- Your off-site footprint pulled into agreement: listings, profiles, reviews and mentions that all say the same thing about you
- Two to four pages or articles, chosen on queries you have a real chance of winning rather than on search volume
How it runs
The same cycle, every month
Four steps that repeat. The fourth is the one that keeps the other three honest.
Audit and order of work
Technical state, editorial state, and what the rest of the web currently says about you. One pass. You get the list ordered by what each item costs you rather than by how fast we could fix it, and the audit is free and yours either way.
Foundations
Rendering, speed, page structure, structured data, both search consoles connected. Boring work. Nothing above it holds without it, and it is the half of the job most agencies quietly skip.
Monthly production
New content, existing pages rewritten, off-site presence built out. A pace you can hold for a year beats one big push followed by six months of silence, and search rewards the first far more than the second.
Report, and the cuts
Every month you get positions, traffic, and which prompts cited you. We also name the workstreams of our own that produced nothing, and we stop those. Reporting that never contains bad news is not reporting.
Honestly
What nobody can promise you, including us
This is the service where results take the longest and are hardest to attribute. You should know that before you sign rather than three months in.
Start with the hardest part. Nobody outside the model companies knows what makes an assistant pick one source over another. The weighting is not published, the behavior shifts between releases, and there is no report anywhere that tells you why you were left out of an answer. Anyone promising you a position in ChatGPT is therefore selling you something they cannot deliver. That includes us, on any day we say it, and you should walk out of the call if we do.
Attribution is the second problem, and it is nearly as bad. Google opened a generative-AI performance report inside Search Console on June 3, 2026, which is real progress, and it reached a subset of site owners with impressions as the only documented metric. Outside Google there is no report at all. What you get from us instead is a fixed set of prompts asked every month, the cited sources written down, and a plain note in the months when the list did not change. It is manual work. It is also the best measurement that currently exists, and we would rather tell you that than dress it up.
Then the timing. In our experience the first position movements appear in the second or third month and anything meaningful lands between the fourth and the sixth, while the part that matters most, what other people publish about you, moves slower than anything you control on your own domain. Three months is the minimum engagement for one reason only: below it you are reading noise and paying us to interpret it.
One last thing. If your real problem is an unclear offer or a price nobody understands, search work will bring more people to the same page and change nothing at the bottom of the funnel. We will say so on the first call. It costs us the retainer and it saves you three months.
FAQ
What people ask before signing
Where to go next
- Digital auditStart here. We read what a crawler actually receives from your site and send it back within a day.
- Website designWhen the site itself is the ceiling, this is how we build the replacement.
- React developmentClient-rendered pages are a rendering problem. This is the team that fixes them.
- PricingEvery range we charge, before you talk to anyone here.
Find out what the crawlers are getting
The free audit reads the raw HTML your server sends, times your pages on a phone, and asks the assistants what they already know about you. No access to your site required, and you keep the document whatever you decide.
No commitment · we reply within one business day