Getting your website found on Google: from 0 to 27.6 times

You have a website. It cost money — maybe a lot of money. Someone built it for you, it looks good, and you were happy when it was done.

And then: nothing happened.

No enquiries. No visibility. Maybe you even booked an SEO agency on top, which sent you a monthly report you could not really read. Maybe you had none and quietly asked yourself what you were doing wrong.

At some point you started putting your energy elsewhere. Instagram. LinkedIn. YouTube. At least something happens there, you see numbers there, someone reacts.

And your website has stood there ever since like an expensive piece of furniture people walk past.

It was exactly the same for me. What changed it was no new design and no new agency. It was a thought I arrived at far too late.

The actual problem: you never saw your numbers

The reason your website feels like a riddle is not that you understand too little about it. It is more banal.

Nobody ever showed you what actually happens.

If an agency worked for you, it had the numbers — you had the report. And when you wanted to change something, you asked, waited and paid. If nobody worked for you, then simply nobody had them.

Either way you landed in the same place: you make decisions about your website by feel. Which text is better. Which topic is worth it. Whether a page is found at all. All conjecture.

And you cannot work with conjectures. You can only have them.

That, by the way, is also the honest reason for the flight to social media: there you see immediately whether something lands. Not because Instagram is better than your website — but because it gives you feedback. Your website gives you none, so you lose interest in it. Completely understandable.

The first step: I did not pick my design

In July I rebuilt my website. Not with an agency and not in WordPress, but with Claude Code.

And the very first step was already different from last time. Instead of asking myself which design I liked, I had Claude analyse all my existing landing pages — and we took what had demonstrably worked best.

Not the prettiest. The one with the best numbers.

That sounds unspectacular, but it is the moment the principle shows itself for the first time: until now I have always paid for taste in design — mine or someone else’s. This time an evaluation decided.

The thought that turned everything around

While building on, the thought arrived that this whole article is about. It consists of two halves, and most people only do the first.

First half: Claude has to know who I am

Not “write something about my offers”. But properly: we defined me cleanly, once. Who I am, whom I work for, what my dream client looks like and what she is currently stuck on, which products exist and how they connect, what my strategy is. All written down, in one place, in a form a machine can work with.

That is work that at first feels like a detour. But it is the prerequisite for everything that follows — and, as will become clear, incidentally the solution to a problem I did not even know I had.

Second half: Claude has to know how the world sees me

For that I connected two data sources; that was all.

Google Search Console — free, and you have probably never opened it. It shows what Google knows about your site: which search queries you appeared for, what position you stood at, how often someone clicked. And, most valuable to me: which of your pages Google knows but has never visited.

If you do exactly one thing after this article, make it this. It costs nothing.

A source for market data — for me, DataForSEO. It delivers what is searched for and how often, and, far more importantly: the real search results in real time. So not just “this term is searched 720 times a month”, but “and these are the ten pages you would be up against”.

And then Claude got going.

Knowing who I am, it looked up what I am actually searched for and how people find me today. From the overlap of the two — what I offer and what is genuinely asked for — a strategy emerged that has been running ever since, without me having to make it anew each time.

That is the difference. Almost everyone gives the AI only the first half: “Here is my offer, make something pretty out of it.” Then you get a pretty conjecture. And some give only the second: a list of search terms that produces texts that sound like nobody.

Only together does it amount to something: A feeling became something measurable. And what is measurable, I can improve.

And what does that look like day to day?

That is the question I get asked most often — so probably the part you really want to know.

I run my website through Claude Code today. Not through a WordPress menu, not through an agency I write to and then wait. When I want to change something, I say so. When I want to know whether it worked, I ask.

I get regular evaluations of how my site is doing. Not as a PDF report with colourful bars that I click away, but as a concrete list: this happened, this is the reason, these would be the next three steps — sorted by what delivers most.

And for content I simply say: “Write me a blog article.” Claude knows by now what about — because it knows who I am and what is searched for. By now I have even had a Notion board set up where my regular blog posts sit pre-drafted. I decide what goes out.

The difference from before is not that anything is faster. The difference is that I know what I am doing, instead of hoping.

The numbers

I moved on 15 July 2026. Here is what has happened since. Source: Google Search Console, 26 April to 25 July 2026. Real numbers from Google, no one’s estimate.

“Impressions” means: that is how often my site appeared in a results list. Not clicked — just shown.

July sits at 6.8 times June. But the month-on-month comparison blurs what actually happened, because the increase is not spread evenly.

It happens across two days.

• 26 April to 18 July, i.e. 84 days: 12.7 impressions a day
• 19 to 25 July, i.e. 7 days: 350.9 impressions a day

That is 27.6 times. Over 70% of a whole quarter’s visibility falls into seven of 91 days. The move was on 15 July, two new articles followed on the 17th and 18th — and then it tipped.

What this number does not prove

And now the part you otherwise never read in articles like this. You have probably believed someone before who showed you big numbers. So here you also get the numbers that argue against it.

One week is not a trend. A spike can remain a spike. I will measure again at the end of August and write down the result — even if I do not like it.

From the bottom, every factor is big. 12.7 to 350.9 is a huge jump. But 12.7 was a tiny number. Anyone already at 5,000 impressions a day will never experience this factor, no matter how well they work. Compare it with your own starting point, not with mine.

Visibility is not yet a client. Over the same period my click-through rate fell — from 10.56% in May to 1.98% in July. I am shown far more often and clicked proportionally far less. Of 118 clicks in the quarter, 56 were also on searches for my own name. That is, from people who knew me anyway.

That is the sort of number one usually trims away. I am leaving it in — because it leads straight to what I really take from these weeks.

Four things I was wrong about

As soon as the data was there, the first thing it showed me was that I was wrong. Four times. And every one of these misconceptions is one you most probably have right now too.

Misconception 1: “My best text is finished.”

My strongest page is a comparison of different AI models. In the quarter it was shown 1,158 times — a third of my entire visibility.

And a whole group of search queries about it revolved around something that appeared in the text only once. 37 different search queries, 160 impressions — a single click.

Google had understood that the page belongs to the topic and dutifully showed it. People did not click, because nothing promised them they would find an answer here.

What that means for you: Your best pages are rarely finished. They are just no longer current. And you do not notice, as long as nobody shows you what people were searching for when your page appeared. I would have considered the text complete — forever.

Misconception 2: “I have to climb higher.”

That is the sentence almost every SEO agency serves — and mostly it is not the actual problem. When I split my search queries into two groups, this came out:

Four clicks from almost 600 impressions.

My pages already stood on page one. They just gave no reason to click. The short text Google shows in the results list simply convinced nobody.

What that means for you: Before you put money into better rankings, check whether you are even being clicked at your current positions. That is the cheaper lever, and it is a ten-minute change. Without this split I would have kept working on what already worked anyway.

Misconception 3: “If the page is online, it is there.”

In Search Console there is a list with the unwieldy name “Discovered — currently not indexed”. Mine held 13 pages. Google knew their address and had never visited them. They practically did not exist.

I checked four explanations and refuted three of them:

• Too new? No, some were months old.
• Not linked anywhere? No — one is linked from 53 other pages, another sits in the main menu.
• Broken links? No, technically flawless.
• Too thin? Partly, yes.

What remained: Google visits a small website only a limited number of times — and I had wasted those few visits myself. My content overview claimed on every visit that all 48 pages had changed today. If everything looks equally urgent, in the end nothing is urgent.

Two changes: seven thin pages set to “do not show”, leaving 41 of 48 that are worth it. And the change date pulled from the real edit dates.

What that means for you: More pages are not more visibility. Every superfluous page costs you another one. If someone built you a website with forty subpages, twenty of which are filler, those twenty are working against you.

Misconception 4: “What happens in AI answers cannot be influenced.”

More and more people no longer ask Google, but ChatGPT or Claude. There is no ranking there to pride yourself on. There is only: you are named, or you are not named. And if you are named — with what description?

On 17 July, ChatGPT, when asked, still described an outdated version of what I offer. Not wrong in the sense of invented. Just my offer from two years ago. On 27 July it was gone.

The reason for that is precisely the work from the beginning that felt like a detour: Because I once cleanly wrote down who I am, whom I work for and what I offer — and because that description is now the same everywhere — the machines have something unambiguous to grasp. Before, they pieced together from old fragments on the internet what I presumably do.

What that means for you: If ChatGPT had to describe you today — would you know what would come out? That is a question you can answer in thirty seconds, and the answer is often uncomfortable. But it is also changeable, and not through more content, but through clearer content.

And once, the tool itself was off

One more point, because it shows that “having data” does not mean “believing data” — and because probably nobody has told you this so plainly.

For another of my projects, two search terms looked perfect: decent volume, purchase-ready searchers, and the pages in the top spots had strikingly few recommendations from other websites. By the usual logic that means: weak competition, you can get in.

Then I looked at the actual results list. All ten spots belonged to large software corporations — among them IBM, Zendesk and Atlassian. The low number was no open door. It was a measurement error. I struck both suggestions.

What that means for you: If someone sells you a keyword list without showing you the real search results behind it, they are selling you a conjecture with decimal places. Look first, believe after.

What you need for this — and what it costs

So this does not sound like magic:

Google Search Console is free. Setting it up is your first step, independent of everything else here. For your own site it is superior to any purchased estimate — in the same project I saw a provider underestimate my situation by more than tenfold, because its data was three weeks old.

For market data I use DataForSEO*. And here I will be honest with you: This is a tool for developers. It has no pretty interface to click around in. It gets connected to a program — in my case Claude Code — and delivers its data there. You pay per query, not a flat rate; a single run costs cents.

If you are thinking “I will set that up over the weekend”: probably not. That is exactly why the offer sits below.

And, less technical but more important than all of it: the willingness to let the numbers win against your own conjecture. All four misconceptions above were plausible. They were just wrong.

* Advertising: the link to DataForSEO is a referral link. If you create an account through it, I receive a small commission — nothing changes in the price for you. I recommend the service because I use it myself; all the numbers in this article were produced with it.

These numbers are mine. It gets more interesting when you lay your own beside them. That is exactly what we talk about in my free community “KI — aber richtig” — the templates and workflows from my videos live there too. And if you want company instead of doing it alone: the AI Business Community offers monthly live calls, workshops and the MACHZEIT! challenge for exactly that.

Frequently asked questions

I already have a website that cost a lot of money. Was that for nothing?

No. In the vast majority of cases the content is usable and so is the positioning — what is missing is the feedback. You do not know which of your pages works and which is just there. That can be retrofitted without throwing everything away. For me the complete rebuild was the right decision, but it is not for everyone.

What does it mean to “define yourself as an entity”?

It means: once, cleanly write down who you are, whom you work for, what you offer and how it all connects — in a form a machine can read too, and then the same everywhere. Sounds like bureaucracy, but it is the foundation for an AI not having to piece you together. And it is the reason ChatGPT now describes my current offer and no longer my old one.

Do I need programming skills for this?

For the setup you need a guide and some patience, not the ability to program. For the part afterwards you need something entirely different: the willingness to question a number. That is the real competence, and it has nothing to do with technology.

Is this not simply what an SEO agency does?

The difference is not the method but who sees the numbers. With an agency they sit there and you get the report — and for every change you ask, wait and pay. Here they sit with you, and you make the change the same day. I have nothing against agencies. I have something against not knowing my own numbers.

Is this worth it for a very small website?

Especially then. On a small site, Google’s attention is scarce, every superfluous page costs you another one, and a single fitting article moves a great deal proportionally. Both are in the numbers above.

Should I give up social media now?

No. The point is not that social media was wrong — it gave you feedback when your website did not. The point is that your website can do the same as soon as you see its numbers. And unlike Instagram, it belongs to you.

Conclusion

The 27.6 times is a lovely number, and I show it gladly. But it is not what I take from these weeks.

What I take: I believed four things about my own website that were not true. None of these assumptions was stupid. They were all plausible. They were just wrong — and without data, none of them would ever have surfaced. I would have kept working in the wrong direction with full conviction. For years.

And if I had to single out one thing, it is this: the turnaround did not come from me writing better texts. It came from Claude knowing both — who I am and how the world searches for me. Everything else followed from that.

If you recognised yourself in the feeling at the beginning, it is probably not because you know too little about websites. It is because nobody ever showed you your own numbers.

As soon as you see them, the guessing stops. And then your website is no longer an expensive piece of furniture, but something you can turn the dials on.

If you want this for your website: This exact build is what I do with you in the Website Sprint — defining you cleanly once, rebuilding your website as your property instead of rent in a site builder, connecting the search data and setting up the regular evaluations. Two weeks, live, in a small group with a maximum of six to eight places. The first round runs in the second half of August.