Marco Kohnsgrowth13 min read
Is SEO Worth It? What Four Sites I Run Say
Is SEO worth it for a bootstrapped founder? 90 days of Search Console data from four sites I run: 16,048 clicks, 1.66 million impressions, and where it paid off.
SEO is worth it when three things are true at once: people type your problem into Google, the page you can build answers it better than the ten already ranking, and you can wait a year for the payoff. When one of those is missing, it is the most expensive channel a small team can pick, because the cost is paid in months of your own time and the bill arrives before any result does.
I run four sites and I read their Search Console every week. In the 90 days to 27 September 2026 they earned 16,048 clicks from Google between them. One of them earned 13. The gap between those two numbers is the whole answer to the question, so this piece is built on the numbers rather than on the 550% return-on-investment figure that the agency articles ranking for this query pass around without a source.
What do four sites I run say about whether SEO pays?
Three product sites earned between 1,439 and 12,772 clicks each in 90 days, and this personal site earned 13. Same owner, same tooling, same weekly attention, and a thousandfold spread in outcome. The variable was never effort. It was what each site could offer a searcher.
| Site | What it is | Clicks | Impressions | CTR | Avg. position |
|---|---|---|---|---|---|
| mypassion.ai | Career quiz, tools, blog | 12,772 | 1,022,526 | 1.25% | 7.2 |
| erp-pilot.com | ERP pricing and comparison pages | 1,824 | 577,946 | 0.32% | 18.5 |
| lovable-prompts.com | Prompt library and guides | 1,439 | 59,005 | 2.44% | 11.9 |
| marcokohns.com | This site: essays | 13 | 641 | 2.03% | 28.6 |
The pattern underneath the table is intent. MyPassion.AI's traffic arrives at pages that do a job the moment they load: a career quiz, a pay-raise calculator, a review of a competing test. ERP Pilot's arrives at pricing pages for software that costs six figures to buy, where a searcher wants a number nobody publishes. Lovable Prompts serves a prompt the reader pastes straight into a tool. None of that is essay traffic, and this site is an essay site.
| Category | |
|---|---|
| mypassion.ai | 12772 |
| erp-pilot.com | 1824 |
| lovable-prompts.com | 1439 |
| marcokohns.com | 13 |
The essay site is the outlier by three orders of magnitude, and it gets the same weekly attention as the other three. Source: Google Search Console, 29 Jun to 27 Sep 2026, per property
So the first honest answer is that "SEO" is not one bet. It is a different bet on every site, and the size of the bet is set by whether the site has anything a searcher can use in the next sixty seconds.
Why do impressions mean almost nothing on their own?
Because Google shows your page far more often than anyone clicks it, and the ratio between the two is where the money hides. ERP Pilot was shown 577,946 times in 90 days and clicked 1,824 times, a 0.32% click-through rate at an average position of 18.5. Half a million impressions sounds like a business. Page two of Google is not.
The sharpest version of this sits inside one site. MyPassion.AI's pay-raise calculator was shown 426,687 times and clicked 974 times, 0.23%, at an average position of 7. On the same domain, a RIASEC worksheet page for educators was shown 1,360 times and clicked 360 times, 26.5%. One page ranks for a broad phrase where Google itself answers most of the query, and the other ranks for a narrow phrase where the searcher needs the file.
| Category | Percent |
|---|---|
| Pay-raise calculator (426,687 impr.) | 0.23% |
| RIASEC worksheet (1,360 impr.) | 26.5% |
A hundred times the click-through rate on three hundred times fewer impressions. The narrow page earns more per impression than anything else on the site. Source: Google Search Console for mypassion.ai, 29 Jun to 27 Sep 2026
The third-party data says the same thing at web scale. SparkToro and Datos found that for every 1,000 US Google searches in 2024, 360 clicks reached a site Google does not own, and 58.5% of searches ended with no click at all. Ahrefs compared 300,000 informational keywords in March 2024 and March 2025 and measured a 34.5% lower click-through rate for the top-ranking page when an AI Overview sat above it. Ranking has not stopped mattering. Ranking for a query Google answers itself has stopped paying.
The operating rule I took from my own numbers: judge a page by clicks and by what the clicks do afterwards, never by impressions. Impressions are Google's decision. Clicks are the searcher's.
How long did it take before anything happened?
About a year from the first indexed page to traffic worth planning around, and the curve was flat for the first six months. MyPassion.AI recorded its first Google impression on 17 September 2025. Here is what the four 90-day windows since then look like, read from the same Search Console property.
| Category | |
|---|---|
| Oct to Dec 2025 | 116 |
| Jan to Mar 2026 | 1518 |
| Apr to Jun 2026 | 4491 |
| Jul to Sep 2026 | 12772 |
116 clicks in the first quarter of publishing. Anyone judging the channel at month three would have shut it down. Source: Google Search Console, four consecutive 90-day windows ending 27 Sep 2026, computed from the 360-, 270-, 180- and 90-day totals
The first window is the one to sit with. Three months of publishing, 116 clicks. Every guide ranking for this keyword says results take three to six months, and the phrase makes it sound like a delivery date. It is closer to the moment the line first leaves zero. The compounding part, where the fourth window is nearly three times the third, only exists because nobody turned the channel off during the first.
That is the whole case for SEO on a product site, and it is also the whole case against it for anyone who cannot fund a year of output before the channel returns anything. Paid acquisition tells you inside a week whether the offer converts. Search tells you inside a year, and in the meantime you are guessing.
Is a personal site at DR 22 a different bet entirely?
Yes, and pretending otherwise is how personal blogs end up with 40 posts and 13 clicks. This site has a domain rating of 22 and earned 641 impressions in the same 90 days that MyPassion.AI earned a million. It is not a failure of the writing. It is what a site of essays, without a tool, a dataset or a transactional page, looks like in Google.
The SEO work that did pay here was hygiene rather than content. When I first got Search Console access to this domain on 9 September 2026, I found 26 URLs from the old WordPress site still earning impressions and every one of them returning a 404. Together they carried 759 impressions and 12 clicks over 480 days, about 15% of every click the domain had ever recorded, going to dead pages. One of them was the old German homepage, ranking at position 1.9 for my own name. All 26 now redirect to a live page. That afternoon of redirects did more for this site's search performance than any single essay on it.
| What was broken | URLs | Impressions | Fix |
|---|---|---|---|
| German ChatGPT course pages dropped in the migration | 13 | 597 | Redirect to the AI training page |
| Old German homepage, ranking for my name | 1 | 88 | Redirect to the homepage |
| WordPress attachment pages | 8 | 55 | Redirect to the parent post |
| Flat category archives from the old tree | 2 | 19 | Redirect to the renested archive |
Where a low-authority site can still win is a narrow cluster of keywords with low difficulty and a specific intent, which is why the reflection-practice essays here target terms in the hundreds of searches a month rather than the tens of thousands. Everything else on this site earns its readers through the newsletter and LinkedIn, and I plan it that way. Treating an essay blog as an SEO bet is how you end up writing for a search engine that was never going to send anyone.
What did AI change about the answer?
It removed production cost as a moat, which means volume of content stopped being a strategy and started being a risk. I am working through a large-scale Google manual action at the moment, and the sequence and the causes are written up in the piece on Google penalties. The short version: Google rewrote its spam policy in March 2024 so that content produced at scale to manipulate rankings is a violation whether a human, a model or both produced it. Volume plus thin value is the profile it enforces against.
The distribution of outcomes was already brutal before generative models arrived. Ahrefs found in its December 2023 study of roughly 14 billion pages that 96.55% of them get zero traffic from Google, and a further 1.94% get between one and ten visits a month. Cheap production did not change that distribution. It made the losing side of it cheaper to join.
What I saw in the research I co-authored points the same way. The Journal of Business Research paper on generative AI for growth hacking, published in 2025 with Prof. René Bohnsack and colleagues, interviewed 20 startups from pre-seed to Series E on how they use these models in their growth strategies. The consistent finding was that the models compress execution, drafting, testing variants, running more experiments with the same team, while the drivers that separated companies were customer understanding in a specific niche, resilience and agility. Nothing in it suggests that writing more pages faster is where the advantage sits, and I wrote up what has changed since the study separately.
For SEO the consequence is direct. The pages that earned clicks in my table are the ones a language model cannot produce from the open web: a quiz that returns a result, a calculator, a pricing figure assembled from sources, a downloadable worksheet. The pages any model could write in ten seconds are the pages Google shows to nobody.
When is SEO worth it, and when is it not?
Run the channel through Brian Balfour's Product Channel Fit test before you run a keyword tool. His line is that products are built to fit channels and channels do not mold to products, and search is the channel where that is most true. Here is how I score it for my own portfolio.
| Condition | Worth it when | Not worth it when |
|---|---|---|
| Intent | People search for the problem or the price, in words you can rank for | Nobody searches; the need has to be created |
| Asset | You can build a tool, dataset or page that beats the top ten on substance | Your best page would be a rewrite of theirs |
| Time horizon | You can fund 12 months of output before judging the channel | You need a signal inside a quarter |
| Authority | Target difficulty sits within reach of your domain rating | Every result in the top ten is a DR 70+ publisher |
| What a click does | The page converts or captures the reader on arrival | The click reads and leaves, with no next step |
| Team | One person can ship and measure weekly without cash spend | It competes for the same hours as the product |
Applied to my own four sites: MyPassion.AI scores six of six, and the clicks show it. ERP Pilot scores five, with the click-through rate still to fix, because ranking on page two for pricing queries is a position problem rather than an intent problem. Lovable Prompts scores five on a small market. This site scores two, so search here is hygiene plus a handful of winnable clusters, and the writing exists for readers who arrive some other way.
The bootstrapped version of the test adds one more line. I evaluate every channel on four costs: my hours, cash, whether it compounds or has to be bought again, and whether the audience it reaches is the one that buys. Search is nearly free in cash, heavy in hours, compounds harder than anything except word of mouth, and reaches people at the moment they have the problem. That profile is why I keep paying for it on the product sites and why I stopped pretending to pay for it here.
What would I do with ten hours a week and no budget?
Spend the first month on the two pages a searcher can use immediately, not on a content calendar. A comparison or pricing page and a tool, calculator or template. Those are the page types that earned 90% of the clicks in my table, and they are also the pages that give a click something to do when it arrives.
- Pull the demand before writing a word. A keyword with under 200 searches a month is not a traffic bet, and a top ten made only of large publishers is not winnable without links. Both are keyword choices, not excuses.
- Build the asset before the article. The calculator, the dataset, the worksheet. The article is the wrapper.
- Fix what already ranks. Dead URLs, missing redirects, titles that do not match the query. On this site that was the highest-return hour of the year.
- Measure clicks and what they do, from month one. Impressions are noise. A page with 1,360 impressions and 360 clicks beats a page with 400,000 impressions and 974 clicks, on every metric that pays rent.
- Set the review at day 90, 180 and 365, and read them in that order. Day 90 tells you whether Google indexed you. Day 365 tells you whether the channel works. Nothing in between is a verdict.
I teach this sequence in the executive seminars at CATÓLICA-LISBON, and the objection I hear most is that it sounds slow. It is slow. It is also the only channel in my portfolio where the fourth quarter did nearly three times the work of the third without anyone spending a euro.
Three questions I get asked about this
Is SEO dead now that Google answers questions itself?
No, but the informational middle of it is dying. Ahrefs measured a 34.5% drop in click-through for the top result on informational queries with an AI Overview, and SparkToro found 58.5% of US searches end without a click. Queries where the searcher needs a tool, a price, a file or a decision still send clicks, because the answer box cannot deliver those. Aim there.
How much traffic should a new site expect in its first six months?
Plan for almost none. MyPassion.AI earned 116 clicks in its first 90 days of indexing and 1,518 in the next 90. If your plan needs search traffic inside six months, your plan needs a different channel for six months.
Is it worth doing SEO for a personal blog?
As hygiene, yes: redirects, titles, one indexable page per topic, so the readers who search your name or your one strong topic find you. As a traffic strategy, rarely. This site earns 13 clicks a quarter from Google and its readers come by email and LinkedIn, and I would rather say so than sell you the 40-post plan.
The question in the title has a short answer after all. SEO is worth it when you have something a searcher can use and a year to wait. When you do not, it is worth exactly what this site's 13 clicks suggest, and the honest move is to build the asset first or spend the hours somewhere that pays sooner.