One year of seo-kreativ.de: what one data set reveals about SEO

The most important number of my first year is not in Google Search Console. It is in Bing: according to Bing Webmaster Tools, my pages were cited 8,879 times in AI answers within half a year.

At Google, the same year looks like this: impressions up 85-fold, average position from 60.5 to 9.5 - and the click-through rate cut in half anyway. Shown 2.14 million times. Clicked 6,687 times. In July 2025 it was 33 clicks in the entire month.

For twelve months I celebrated the wrong curve. My work had been landing for a long time - just not as visitors, but as a source. How I noticed that, what the detour cost me, and why to this day I cannot prove whether Google cites me: that is the rest of this article.

One thing up front, so the expectation is right: this is not a traffic play. “seo-kreativ.de development” has zero search volume, nobody will find this article through Google, and that is on purpose. It is the data set I usually only put on the table in client meetings: a real, young domain, observed for a year, with all the numbers in the open. Including the uncomfortable ones.

I have been building this blog in public since spring 2025. 90 German posts, each mirrored into English, makes 180 articles. The Search Console data runs from 1 July 2025 to the end of June 2026, the link and ranking numbers come from Ahrefs and DataForSEO, the AI numbers from Bing Webmaster Tools. None of it is estimated.

And because I do not want to fool myself, one rule applies: I label every statement. Measured means the number appears that way in a tool - with that tool’s measurement logic and its uncertainties, not as absolute truth. Timing means two things coincide in time, nothing more. Assumption means I believe it but cannot prove it. That sounds pedantic. It is the only protection against turning your own numbers into a hero story.

Because I will continue this every year, this is year 1 - same URL, next year the next curve. Let us start with the one that almost fooled me.

In September 2025 my click-through rate jumped to 0.92 %, and clicks went from 57 to 486. For a few weeks I thought I had figured it out. It was the best CTR month of the entire year. From then on it went down, month after month - not because my posts got worse, but because the search results page changed. The headline numbers of the year, measured in Google Search Console:

That is the core of the whole year in four rows. The position gets radically better, reach explodes - and the click-through rate falls anyway. Across the full year that adds up to roughly 2.14 million impressions and 6,687 clicks, an average click-through rate of 0.31 % (measured).

For months I stared at the first row and felt good. I took the second one far too lightly. I was seen 2.14 million times. I was clicked 6,687 times. For every thousand people Google showed my page to, three opened it.

This is exactly what the industry often calls the Great Decoupling - not an official Google term, but an industry narrative for a real pattern: impressions and clicks come apart. Ranking progress that used to translate directly into visitors now drains away into AI Overviews, zero-click SERPs and a results page that answers more and more by itself. I do not need to quote a study for this. I can see it in my own CTR curve, which fell from 0.92 % in September 2025 to 0.19 % in June 2026 while my position improved at the same time. Position up, click-through rate down: that is not a ranking problem, that is the new SERP. In a concrete case I measured what was behind the decoupling.

Four curves, one time axis. The blue one shoots up, the green one crawls, the red one falls, the yellow one improves. Anyone who only sees the blue one is celebrating a year that did not happen that way.

With backlinks I am deliberately careful. Ahrefs (free) counts 343 backlinks - but only 163 of them are dofollow, 182 are nofollow and 3 are UGC (measured in Ahrefs). So more than half of my links pass no direct ranking signal at all. And even the dofollow number still flatters; I will break down why further below.

A look beyond Google supports the picture. On Bing I collected 396 clicks from 23,674 impressions in the same year (measured in Bing Webmaster Tools) - tiny in volume, but with a click-through rate of 1.67 %, more than five times my Google CTR. Put differently: per 100 impressions, Bing sends me five times as many people as Google does. Bing is small for me - but on Bing people click. Same content, same person, just a results page that (still) eats less of it itself. Decoupling is not a law of nature, it is a property of the particular SERP.

When I saw the May curve, I wanted it to be the core update. Impressions from 301k to 451k, right inside the rollout window. That would have made a nice story: “I set the right signals, the update rewarded me.” I did not write it, and in a moment I will explain why I was not allowed to.

Four larger Google updates rolled out during the observation year, all officially confirmed via the Google Search Status Dashboard:

And now the part nobody else writes. I could attribute each of these jumps to an update. That would simply be dishonest. My domain is growing during the same period anyway - new posts, more indexed pages, more topics. When impressions shoot up in May 2026, that coincides with the core update (timing), but it coincides just as much with my own growth curve. Separating the two would require a control group I do not have.

If you want to know how I read volatility at all without panicking: I wrote a separate post on interpreting the Semrush Sensor. Short version: a spike in the sensor is a weather report, not a verdict on your site.

If you ask “which post performed best?”, the honest answer is: depends which number you mean. Four real examples from the year (measured):

Look at that spread. My TL;DR page pulls 471 clicks in its English version - but out of 357,824 impressions. That is a generic definition query (“tl dr meaning”) where I do sit on page 1, but which the SERP answers almost entirely by itself. Lots of reach, hardly any yield.

The irony only hit me while analysing the data. For the query “tl;dr” alone I was shown 99,551 times - 0.05 % clicked. So I rank for an abbreviation that means “too long, didn’t read”, and that is precisely what happens. By the way, my best click-through rate by far belongs to an entirely different query: “seo kreativ”, at 23.17 % (measured). The only search query where the SERP cannot replace me is me.

The opposite pole: my German post on the num=100 parameter and the core update news pull 2.5 to 3.5 % click-through rate - from a fraction of the impressions. News queries get clicked because the answer has to be fresh and the SERP does not (yet) deliver it itself. And then the data point that made me swallow: my Core Web Vitals post sits at position 7.6 and pulled exactly two clicks out of 37,377 impressions. Two. Not two thousand. Position 7 on page 1 sounds like everything you want in SEO - and in this case it was a full year of work for two people.

The consequence for editorial planning: reach posts, click posts and authority posts are not the same thing. Optimise only for impressions and you produce vanity. Chase only CTR and you miss the topics that establish you as a source long term.

At this point in the year I was fairly sure I was losing. Better rankings, more reach, hardly any clicks - and the question that decides everything seemed unanswerable: am I even mentioned in AI answers? For this review I tested 13 of my core topics live. The result has two halves, and I want to show both.

The good half (measured in a live SERP check on 9 July 2026): for 8 of 13 topics seo-kreativ.de appears near the top organically, several times at position 1 - for the Semrush Sensor, for the December core update, for Google’s user agent for AI agents and for the commodity content self-test. For several of them I sit ahead of established players. For a one-year-old one-person domain in these topics, that is the real surprise of the year.

The honest half, part 1 - Google: ranking organically and being cited in an AI answer are two different things. On the Google side I cannot prove the second one. Since June 2026 Google does have a dedicated AI report in Search Console, but it currently shows only impressions in AI Overviews and AI Mode - no clicks, no CTR, and certainly not whether and how you are cited in the answer (according to Google Search Central, June 2026). On top of that it is rolling out to only a subset of accounts; as of July 2026 it has not arrived in mine at all. Google develops these reports iteratively, so this can change at any time. A SERP tool, meanwhile, shows me the organic list, not the citation layer. So I have the foundation that AI Overviews draw from - whether my name appears in the generated answer remains a black box at Google.

Part 2 - and then the turn: the answer was there the whole time. Just not at Google. Bing Webmaster Tools reports directly how often Bing counts my pages as a citation in AI answers - first-party, not a third-party extrapolation. According to that report it is 8,879 citations in six months, grown from 637 in January to 3,510 in June, with a daily peak of 260 and around 12 cited pages per day on average (measured in Bing Webmaster Tools).

I spent a year counting clicks and missed the fact that my posts are being read for a long time already - by machines that pass them on. That is the number that turned my picture of this year around.

One detail I deliberately do not over-interpret: my strongest day was 260 citations on 22 June 2026. That is the same day my URL length study with a DOI was published on Zenodo. Timing, not proof - I have no control group and no causality. But it is a piece of timing I will remember.

A brief look at Ahrefs through a company account points in the same direction - there, too, first mentions in AI answers show up. But I am deliberately reporting that only quietly: third-party data, different methodology, no first-party evidence like at Bing.

Why do I rank so stably in the niche anyway? My assumption (and nothing more): fresh news topics with a clear structure, a consistent author entity, and topics big enough to have demand but too specific for the tool giants. That is not magic, that is a gap. Why pure ranking without authority is not enough is something I took apart in more detail in why good content alone does not rank.

Now the promised breakdown. Those 163 dofollow links look like 163 convinced websites. They are not. A large share of the highly rated ones among them are *.wordpress.org subdomains: de.wordpress.org, fr.wordpress.org, es.wordpress.org and so on. That is one link - my plugin entry in the official directory - mirrored across dozens of language versions. It counts as many links in the statistics, but in reality it is one source (verified in the link profile).

Scrolling through the list, it got almost funny. By tool logic my “strongest” referring domains are the Frisian, the Basque, the Zulu and the Khmer versions of wordpress.org. Dream metrics, all of them. I did nothing for a single one of those links except publish a plugin years ago. Anyone who sorts their link profile by domain rating and then feels proud should take a look at who is actually linking.

And then the part I am genuinely proud of. Once you subtract the plugin effect and the spam noise, a few links remain that nobody ordered. The nicest of them: the HubSpot blog links to my write-up of a Google search statistic as a source (according to DataForSEO). You cannot buy an editorial link like that, you have to earn it by writing - it is the real reward for the work that goes into an article. A few more specialist sites were added over the year, but I am deliberately not hanging them up here as a trophy wall. Fitting the picture: Ahrefs puts my domain rating at roughly 37 after one year, starting from 0 - a snapshot from a brief look through a company account, to be read only as a rough direction, not as an exact curve.

What remains beyond that are the few anchors I set deliberately myself, rather than begged for:

• The WordPress plugin “External Links Overview” with more than 800 active installations in the official wordpress.org directory - and that plugin entry is exactly the source of all those language subdomain links above (verifiable on wordpress.org).
• Open source code on GitHub at github.com/seo-kreativ, among others the commodity checker - verifiable and linkable (DataForSEO has recorded the link since June 2026).
• The SEO-Kreativ handbook (tredition, 2026, ISBN 978-3-384-84004-2) - a book with an ISBN is an entity anchor that no link exchange can replace.
• A URL length study on Zenodo, my own analysis of more than 27,000 German SERPs with its own DOI (10.5281/zenodo.20798700) - citable like an academic source (Zenodo, June 2026).
• The Grokipedia backlink test: DataForSEO has recorded a Grokipedia link to the domain since May 2026 (according to DataForSEO, not independently verifiable). Whether such a link is worth anything is something I dissected in my case study on Grokipedia backlinks - more interesting as a signal than as link power.

The point behind all of this: a book with an ISBN, a study with a DOI, a plugin in the official directory and open code on GitHub are not “backlinks” in the classic sense. They are entity anchors - evidence that behind the domain stands a real, verifiable person with real output. Exactly the kind of signal a young brand needs so that search engines and AI systems recognise it as an entity in its own right. How I get links for small domains at all is in building backlinks for small domains.

So far a lot of this sounded like it went to plan. It did not. Three things this year cost me time or fooled me, and if I leave them out, this review is an advertising brochure.

The English rollout is a double-edged sword. The English versions now account for around 92 % of my impressions - in June 2026 that was 533k English against 48k German impressions. But the English click-through rate sits at 0.16 %, the German one at 0.53 %. In German I reach a small, engaged niche; in English I collect enormous, AI-Overview-exposed reach with a meagre yield. In absolute terms English brings more clicks (826 against 253 in June), but per impression it delivers a third of the efficiency. Whether the doubled maintenance effort is worth it is still open for me - honestly, not rhetorically.

Part of the link growth is noise. In June 2026 a whole batch of new referring domains suddenly appeared in the backlink tools. That is not a success, those are scraper and spam networks that harvest content automatically and link to it. Anyone celebrating such a spike as a “link explosion” has not looked at the domains. I do not count them towards my authority - on the contrary, I watch things like that so they do not tip over.

My favourite find from that sediment: on the website of a Spanish secondary school there is a link to my TL;DR post. The anchor text is an advertisement for a Telegram channel that sells backlinks. Somebody dropped spam there and built in my URL along the way. I have nothing to do with it, I cannot prevent it, and in every backlink tool it counts as a “new referring link”.

Vanity impressions on generic terms. On “tl;dr” alone I have 99,551 impressions at a 0.05 % click-through rate. It ranks, it looks great in the statistics and it delivers practically nothing. Terms like that inflate the impression curve without bringing a single meaningful visitor. Nice for the ego, worthless for the business.

I am deliberately holding back on numbers here, for a simple reason: the business value of this project does not hang on the click count. If it did, a blog that pulls 6,687 clicks a year would be a bad investment. The value lies elsewhere.

The one hard, publicly visible signal is my plugin “External Links Overview” in the WordPress directory with more than 800 active installations. That is not a marketing figure, those are people actively using a tool of mine. Consulting enquiries and sales of the handbook also run through the blog - I am deliberately mentioning those only directionally, not as a metric. The blog is the place where someone sees how I think before they get in touch. That is exactly what it is for.

Put differently: the ROI of the article you are reading right now is not its clicks. It is the two or three people who read it and think “he knows what he is talking about”. That appears in no Search Console. And it is the only reason why, after twelve months and 6,687 clicks, I have not stopped.

What I take away from twelve months, without the calendar wisdom:

• Entity anchors beat reach. The book, the study, the plugin and the open code did more for my visibility as a source than any viral impression peak. They compound; reach evaporates.
• Owning a niche beats playing everywhere. Where I hold position 1, the topics are specific and fresh - not the fiercely contested evergreens like “javascript seo” or “google discover”, where Google docs and the tool giants push me out (measured in the live SERP check).
• Impressions are a trap. I looked at the beautiful, steep impression curve for too long. The flat click curve next to it is the more honest metric.
• Attribution is harder than anyone admits. On a growing domain you can barely separate cause from time. Anyone telling you otherwise is selling you timing as causality.
• Refresh before new. My strongest pages are maintained, updated posts, not one-off publications. How I find and clean up dead content is in content audit and pruning.

Concretely, for year 2 that means: less chasing impressions, more entity, an honest review of the English rollout, and steering the internal link equity more deliberately towards the pages that can actually make it into the click zone.

The curve that almost fooled me was the impression curve. It climbed 85-fold, it looked like victory, and for a year it answered the wrong question for me. The right question was never “how often am I shown”, but “what am I the source for”.

If I boil the year down to one sentence: search rewards visibility generously and clicks hardly at all, and the younger the domain, the more tempting it is to talk that picture up. I have tried to do the opposite - to label every number, to call every weak effect weak, to hold every impression peak against the flat click curve beside it.

The most valuable thing I take from twelve months is not the 85-fold reach. It is the realisation of how easily you deceive yourself with your own data - and that the only signals that truly compounded lie off-site: a book, a study, a plugin, open code, and specialist sites that link of their own accord. That Bing counts my pages thousands of times in AI answers shows where this substance travels: into the AI answers, where the pure chase for clicks was lost long ago. Same URL next year, and we will see whether the gap widens further.