Four days old, and second in five countries
We published an article on a Wednesday. By Saturday it was the first organic result in Hungary, Denmark and Portugal. Here is exactly what we did, what it tells us, and the second half of the same dataset — which turns out to be the more interesting one.
What happened. An article we published on 29 July held average position 2.0 in Hungary, Denmark and Portugal by 1 August, 4.0 in Germany and 7.0 in the United States.
Why it worked. Not authority, not budget, not luck. We answered a question that became askable a few weeks ago and that almost nobody had written a proper page for.
The second half. On “seo cluj” — in a city with roughly 1,400 IT companies — we went from an average of 43.3 to 23.0 in a single week. Still climbing, and that is the part we are enjoying most.
The starting line. Our domain has been ours since 2011, which helps. It also sat untouched for about six years, which means the help is gradual rather than instant. Day one of real SEO on it was 25 July.
What actually happened
On 29 July we published a piece on what is and is not live for ChatGPT ads in Europe. It was a Wednesday. On Saturday morning we opened the Search Console API to write something else entirely, and found the article sitting second in three countries.
Here is the whole dataset. Not the flattering rows — all of them.
Four days old, second in three countries
The article went up on 29 July. These are its average positions on 1 August, straight from the Search Console API, split by the two ways people phrase the question.
For “chatgpt ads in europe” we came seventh in the United States, fourth in Germany and second in Hungary — four days after the article went up.
cittago.com · Search Console API, classic Performance report, 26 July – 1 August 2026
| Query | Country | Average position |
|---|---|---|
| chatgpt ads in europe | Hungary | 2.0 |
| chatgpt ads in europe | Germany | 4.0 |
| chatgpt ads in europe | United States | 7.0 |
| chatgpt ads europe | Denmark | 2.0 |
| chatgpt ads europe | Portugal | 2.0 |
| chatgpt ads europe | Germany | 5.7 |
A note on how to read that, because it matters. Google counts the AI Overview block as a result, so an average position of 2.0 means we were the first blue link on the page, directly under the generated answer. On the Hungarian result, the AI Overview summarised the launch date and the regulatory position, and our article was the first thing a reader could click to go deeper.
Two phrasings, six data points, five countries. These are averages across the week, pulled from the API — not a screenshot of one search typed once.
Two kinds of question, two different games
Now for the half of the dataset that we find more interesting, because it is the one with room in it.
In the same seven days, on the same site, here is where we sit on the queries an agency in Cluj would love to own.
The ones we win today, and the ones we are climbing
Same site, same week, same effort behind both. The difference is not the work — it is how old the question is, and how many people have already answered it.
Long, dated, specific. Not many people search these, and almost nobody has written the page that answers them properly. We are third, second and first.
cittago.com · Search Console API, seven-day and 28-day windows to 1 August 2026
| Query | Type | Average position |
|---|---|---|
| cdn pmi acquisto senza contratto annuale | Long, specific | 1.0 |
| chatgpt ads in europe | Narrow, dated | 2.0 |
| google ads 17 august | Narrow, dated | 3.0 |
| seo cluj (last 7 days) | Short, contested | 23.0 |
| seo cluj (last 28 days) | Short, contested | 43.3 |
| google adwords cluj | Short, contested | 62.2 |
| adwords cluj | Short, contested | 76.0 |
Toggle that to the second view and the numbers look rough at first glance: 76, 62, 23. Then look at what they were. “seo cluj” averaged 43.3 across the last 28 days and 23.0 across the last seven. It halved in the week we started, from a standing start, and that was the easy half — the part where a site that had been invisible simply becomes visible.
The hard half begins in a few days, and we are genuinely curious about it. Cluj-Napoca is not an easy town to rank in. It is a university city that has spent a decade being called the Silicon Valley of Eastern Europe, with something on the order of 1,400 IT companies and around 20,000 people working in tech, on a population of roughly 320,000. Hundreds of agencies. Thousands of freelancers, many of them very good and many of them with a decade of head start on this exact keyword.
So no, we do not expect “seo cluj” to be a walk. That is precisely why it is worth doing: it is the honest test of a method. If it works there it works anywhere, and if it does not we will have learned exactly where it stops working, which is worth almost as much.
Narrow, dated, specific questions are winnable in days. Short, contested ones are won over months and years, on the authority the narrow ones build for you. Start where you can win, so that later you can compete where you cannot yet.
To be clear about the stakes: this is not a target we have to hit. Nobody is paying us to rank for our own city. It is a challenge, and a bit of housekeeping — we had not touched our own platform in nearly seven years, and it was starting to feel absurd to run this work for other people and not for ourselves. Whatever we do from here, heavily or lightly, we will do with the same expertise and the same meticulous hygiene we apply to client work. Everything genuine, everything organic. We are fairly confident that pays. The only real variable is finding the time between client projects to stay consistent.
An old domain, eight days of SEO
Some context, because it explains both halves of the chart above.
Cittago has been a digital studio in Cluj-Napoca since 2011. Fifteen years of genuinely liking this trade, watching what changes, and following it hardest whenever a new client dragged us into a niche we had never worked in before. That part is not marketing. It is most of the reason we are still here.
We have also been boutique the entire time, by choice rather than by circumstance. Our busiest year was 2020: eight freelancers working across different projects and two large Cluj agencies as partners. Then came the pandemic and we pivoted. The digital work stayed, mostly for old clients and close friends, kept up more out of interest than ambition.
An old domain, and eight days of SEO on it
Tap a period. The short version: we have been at this since 2011, we have been deliberately small the whole time, and our own website has been the one project nobody was billing for.
We started in 2011 and never wanted to be big. Fifteen years later that is still the plan: a small senior team, which is a business model rather than a stage on the way to something else.
cittago.com · our own history, told plainly
| Period | What it was |
|---|---|
| 2011 | Founded in Cluj-Napoca as a boutique digital studio |
| 2020 | Busiest year: eight freelancers across projects, two large Cluj agencies as partners |
| 2020–2025 | Pivoted; digital work kept for old clients and friends |
| 2026 | Active again: new site 6 July, own SEO and GEO from 25 July |
Somewhere in that stretch our own website stopped being anybody’s job. The domain kept ageing, which is a real asset — it just is not a shortcut. An aged domain that nobody has worked on for six or seven years does not hand you rankings; what it does is compound faster once you finally give it a run of consistency to work with. We are counting on that second part.
Since the spring that has changed. We started taking new projects in earnest again, updated how we work, followed the AI market far more closely than is strictly relaxing, and tested things on client accounts — some of which worked well enough that not running them on ourselves started to feel silly.
So we rebuilt the site. It went live on 6 July 2026. Everything until the last week of that month was foundation rather than ranking work. The SEO and GEO proper started on 25 July. This article is day eight.
What we actually did, in order
None of this is clever. It is the boring version of things everybody already knows, done consistently for a week.
- Picked topics from query data, not from a calendar. A weekly pull of the Search Console API, looking for questions where we already collect impressions and hold no position worth having.
- Chose questions with a date attached. The pattern that keeps working: year, region, platform. “chatgpt ads in europe”, “google ads 17 august”. Specific, current, short-lived.
- Answered the actual question in the first paragraph. The winning article opens by naming which European countries can buy ChatGPT ads today. Zero of them. Nobody has to scroll for that.
- Wrote each language natively. Not translated — rebuilt from an outline, because people type differently in Italian than in English and a translation carries the English phrasing across with it.
- Put a real FAQ on every article, because an assistant can lift a block and cannot lift an essay.
- Linked internally from day one, from every new piece to a service page and two older ones.
- Generated structured data at build and verified it before deploy — organisation, author, breadcrumbs, FAQ.
That is the whole method. There is no eighth step where something expensive happens.
What we are not claiming yet
Worth saying plainly, so nobody has to work it out from between the lines.
- Traffic has not arrived. Position and coverage moved; clicks have not, and at eight days they should not have.
- The volumes are small. These are narrow queries by design. Winning them is evidence that the method works, not a revenue event.
- Query-level data is a sample. Google redacts most queries on small properties, so what we can see is a slice rather than the ledger.
- It is one week. A position four days after publishing can move in either direction, and dated queries expire by design — “google ads 17 august” will be worth nothing on 18 August.
- It says nothing about hard queries yet. Ask us about “seo cluj” in three months.
What it does show is narrower and, we think, more useful: a small site with an unremarkable recent history can reach the first organic result in several countries within days, if it picks a question that has just become worth asking and answers it properly. That is available to anyone reading this.
How this fits with what we published yesterday
Yesterday we wrote about Google’s new AI report in Search Console, and argued that impressions and position are the two things capable of moving first in an SEO programme, while clicks and enquiries arrive months later. We published our own numbers to show it: average position went from 43.6 to 12.3 in the first intensive week, with clicks exactly where they had been.
This article is the same claim from the other end. Yesterday was the theory with a chart behind it. Today is a single article, four days old, sitting second in three countries with no clicks worth reporting. Both describe the same sequence, and the sequence is the useful part: coverage, then position, then clicks, then customers.
It also lines up with something we wrote earlier in the week on why AI describes your brand correctly and still recommends someone else. Being the page that answers a specific question is what makes an assistant able to use you at all. The ranking is simply the visible half of that.
What we are testing next
We are going to keep running this in the open, because a method whose failures you never see is a sales pitch. Over the coming weeks:
- Whether dated queries expire as fast as we think. We will watch “google ads 17 august” through and past 17 August and publish the curve.
- Whether the narrow wins pull the broad ones up. The test case is “seo cluj”: 43.3 a month ago, 23.0 last week. If the theory holds it keeps improving without us ever writing a page aimed straight at it.
- Whether writing natively beats translating, measured per language rather than argued about.
- Whether the entity markup we shipped on 16 July changes anything in how assistants describe us.
- What one article a day does over ninety days, including the weeks where it does nothing at all.
If you want to follow along, the numbers turn up in these articles as they arrive. We publish the flat months too, and there will be flat months.
Questions nobody has actually asked us
You know the influencer move — “so many of you have been asking me about…” when everybody involved knows that nobody asked anything. We are eight days into publishing this. Nobody has written in about our Search Console positions, and it would be slightly worrying if they had.
What we do have is fifteen years of clients asking us things about SEO, GEO and digital work in general. So the questions below are the ones we think a reader would reasonably want answered after reading this, plus the ones that come up in real conversations about this kind of work. Honest labelling, and rather more useful than a fictional inbox.
How long did it take to rank in five countries?
Four days. The article went up on 29 July 2026, and by 1 August it held average position 2.0 in Hungary, 2.0 in Denmark, 2.0 in Portugal, 4.0 in Germany and 7.0 in the United States.
We would love to tell you this is normal. It is what happens on a question almost nobody has written a good page for yet, which is a different and more repeatable thing.
Does an aged domain make you rank faster?
It helps, and it is a genuine asset — ours has been registered since 2011. What it is not is a shortcut. The site had no real SEO work on it for about six years, and during that stretch it ranked for our brand name and not much else.
Where age earns its keep is later: once there is a run of consistent publishing behind it, an old domain tends to compound faster than a new one.
You rank second in Denmark but 23rd for “seo cluj”. What gives?
Two completely different competitions. “chatgpt ads in europe” became askable a few weeks ago and has few good answers. “seo cluj” is contested in a city with roughly 1,400 IT companies and about 20,000 people working in tech.
And it is moving: 43.3 on average across 28 days, 23.0 across the last seven. That is the first week.
Are broad keywords not worth targeting, then?
They are. They are just not worth starting with. Narrow, dated questions are winnable in days and they teach Google what your site is about; broad terms are won later, on the authority those pages build.
Anyone promising page one for a two-word city keyword in month one is selling something other than SEO.
How do you find questions this specific?
We pull the Search Console API every week and look for questions where we already collect impressions but hold no position worth having, then cross that against whatever changed in the industry in the last seven days.
The pattern that keeps working is year plus region plus platform: “chatgpt ads in europe”, “google ads 17 august”. Specific, current, and largely unwritten.
Is a good position without clicks worth anything?
Early on, yes — as evidence rather than as revenue. Coverage moves in days, position in weeks, clicks in months, enquiries later still. Judging a week-old programme by the click column tells you only that it is a week old.
We went through that order in detail in our piece on the Search Console AI report.
How long have you been doing this?
Cittago has been a digital studio in Cluj-Napoca since 2011, so fifteen years, and deliberately boutique the whole time. Our busiest stretch was 2020, with eight freelancers across projects and two large Cluj agencies as partners.
What is new is not the experience. It is that we are finally pointing it at our own domain.
Will you keep publishing the results, including the boring ones?
Yes, and especially the boring ones. We are running a series of tests on our own site over the coming weeks — query selection, publishing cadence, internal linking, entity markup — and we will publish what each one did, numbers attached.
A method whose failures you never see is a sales pitch, not a method.
Where we would start
If you take one thing from this, take the query selection. Open Search Console, sort by position, and find the questions where you already have impressions and sit somewhere between 11 and 30. Those are the pages one rewrite away from the first page.
Then find the question in your industry that became askable in the last month and has no good answer yet. Write that one properly. It is worth more than a month spent chasing a two-word keyword you will not win this year — and it is what earns you the right to chase it next year.
That is the method, and it is the same one we run for clients through our SEO and AI search work. If you would like us to run it on your site — the questions your customers actually type, the ones you are close to winning, and the ones to leave alone for now — book half an hour. We will tell you which are reachable, and we will tell you if the answer is none.
Related reading: what Google’s AI report in Search Console actually measures, the article this whole piece is about, and the difference between SEO and GEO.
Last updated: 2 August 2026. Positions are averages from the Search Console API and will move. We update this page as the tests above produce results — including the ones that produce nothing.
Which questions can you win?
Half an hour on your Search Console. We come back with the queries you are close to winning, the ones to leave alone this year, and the order to do them in.
