Editor's note: this case study is told in the first person by Marco Bertelli, the owner of a specialty kitchenware shop based in Bologna, Italy. The numbers and outcomes are real. We have changed only a few small details to protect competitive information.
For three years, I paid a respectable SEO agency five hundred euros a month to take care of my website's search visibility. They were professional, they answered emails promptly, and they sent me monthly reports with charts. The charts always seemed to be going up, which was, naturally, comforting.
In total, over those three years, I paid them eighteen thousand euros. Eighteen thousand euros is a real amount of money for a small kitchenware shop. It is the cost of a new espresso machine, three trade shows in Frankfurt, or about four hundred hours of my own labor at what I optimistically calculate as my hourly rate.
Last November, on a slow Tuesday afternoon, I sat down and asked myself an honest question: what had those eighteen thousand euros actually produced? And the answer, when I looked at the data carefully and without the comfort of the agency's monthly charts, was — to put it gently — disappointing.
So I fired them. And I replaced the entire engagement with a combination of AI tools and a few hours of my own time each week. This is what happened over the six months that followed.
The starting line
Before I tell you the results, I should tell you where I was starting from. In November 2025, my shop's website was getting roughly 4,200 organic visits per month from Google. Of those, perhaps 280 were converting into something measurable — a newsletter signup, a phone call, an in-person visit referencing the website, or an online order. My average order value was about ninety euros, and my conversion rate from organic traffic was, by any honest measure, mediocre.
The agency had been focused, for the previous year, on what they called "topical authority building" through blog content. They had published forty-two articles for me, all of which I had paid extra for, and most of which I had never read. When I finally went back and read them, I discovered that they were exactly the kind of bland, AI-generated-but-edited-by-a-human-just-enough-to-pass content that, the truth is, I would not want to read either. No wonder they were not converting.
This was the situation. Now let me tell you what I actually did.
The plan
I gave myself six months. I committed to spending no more than four hours per week on SEO, and a budget of no more than one hundred euros per month on tools — a fraction of the five hundred I had been paying the agency. The savings would, in theory, be invested elsewhere in the business. Or, more honestly, would simply not be spent, which is also a kind of investment.
The plan had four parts.
First, I would do a proper technical audit of my own site, fix what I could, and stop worrying about the rest. I had been told for years that my site had "technical issues," but no agency had ever explained them to me in language I could understand or actually helped me fix them.
Second, I would rewrite my product pages and category pages myself, in my own voice, with the help of AI as a drafting partner. Not as a writer — as a partner. The distinction would turn out to be crucial.
Third, I would aggressively pursue Google Business Profile optimization, customer reviews, and the kind of unglamorous local-SEO work that the agency had largely ignored because it was, in their view, beneath their expertise.
Fourth, I would publish, perhaps, two blog posts per month — but real ones. Things I actually wanted to write about. Knife sharpening techniques. The history of Italian copper cookware. The difference between a good and a great pasta pot. Niche, specific, opinionated content from someone who actually knows what he is talking about.
That was the plan. Now the results.
Month 1: the audit shock
I ran my first AI-powered audit and immediately wanted to call the agency back to ask what, precisely, they had been doing for three years. The audit identified 127 distinct issues on my site. Fourteen of them were in the "high priority" category. Some were embarrassing — pages that had been accidentally noindexed, broken links to products I had stopped selling years ago, missing schema markup on every single product page, duplicate meta descriptions on dozens of category pages.
I spent the entire first week of the project just fixing these things. By the end of week two, my site was, technically speaking, in better shape than it had been in three years. Total cost: zero euros and approximately twelve hours of my time.
By the end of month one, my organic traffic had already increased by 8%. I had not published any new content. I had not built a single new backlink. I had simply fixed the things that were broken. This, in itself, was a humbling lesson.
Month 2-3: the content rewrite
This was the hard part. I made a list of my twenty most important pages — the product categories that drove most of my revenue, the bestselling individual products, my About page, my shipping policy, my contact page. Then, one at a time, I rewrote them.
The AI tool I used was straightforward. I would paste my existing page, write three or four paragraphs of notes about what I actually wanted to say (in Italian, in plain language, like I was talking to a customer), and ask the AI to draft a new version. The first draft was always too generic. The second draft, after I told the AI specifically what to fix, was usually better. By the third or fourth iteration, I would have something I could edit by hand into something I was actually proud of.
The key — and this is the lesson I most want to share — was that I never, not once, published the AI's draft as written. Every page that went live had been touched, sometimes substantially rewritten, by me. The AI was a collaborator, not a ghostwriter. The output sounded like me because, in the end, it was me.
This took, perhaps, ninety minutes per page. Twenty pages times ninety minutes is thirty hours, spread across two months. Manageable.
By the end of month three, organic traffic was up 23% from the November baseline. More importantly, conversions were up 41%. The new pages were doing what the old pages had not: they were actually convincing visitors to buy things.
Month 4: Google Business Profile and reviews
I had been ignoring my Google Business Profile for years. It existed, it had my hours, it had a few photos from when the shop opened. That was it. No posts, no Q&A, no recent updates, and I had perhaps responded to one out of every ten reviews.
I spent month four fixing this. I added forty new photos — not stock images, real ones, taken on my phone, of products in the shop, customers shopping, the espresso machine on the counter, the pasta we sometimes cook for tasting events. I started writing weekly posts about new arrivals and events. I responded, at last, to every review, including the few negative ones.
I also started, for the first time, asking customers for reviews systematically. Not aggressively, just at the right moment — after they had said something nice in person, or after a successful online order. The conversion rate on these requests was much higher than I expected. By the end of month four I had added thirty-one new reviews, all genuine, all five-star but for two four-star reviews from people who were nonetheless reasonably positive.
The effect on local search visibility was immediate. By the end of month four, my appearances in the local Map Pack for queries like "kitchenware Bologna" had nearly doubled. Foot traffic to the shop, which I had been tracking informally, was visibly up. Several customers mentioned they had found me on Google after asking ChatGPT for "good kitchenware shops in Bologna" — which was, frankly, the first time I had heard a customer say that, and it would not be the last.
Month 5-6: the blog experiment
The last two months were dedicated to the blog. I wrote four posts in total, one every two weeks. Each one took me, honestly, somewhere between four and six hours, including research, drafting, editing, and adding photos. Each was on a topic I actually cared about and felt qualified to write about: how to choose a pasta pot, the difference between carbon steel and cast iron pans, why I do not sell non-stick cookware in my shop, and a long piece on the history of Bolognese kitchen tradition.
I used AI to help with structure and to suggest related questions readers might have. I did not use AI to write a single sentence that ended up published. The voice was mine, the opinions were mine, the small mistakes and personal touches were mine.
Three of those four posts are now in the top five Google results for their respective queries. One of them gets cited regularly by ChatGPT and Perplexity when people ask about cookware in Italian — I have tested this myself, repeatedly, with delight. None of the forty-two posts the agency wrote for me are cited by anything.
The six-month numbers
Here are the actual results, side by side.
November 2025 (starting point):
- Organic traffic: 4,200/month
- Organic conversions: 280/month
- Local Map Pack appearances: ~850/month
- Reviews: 47 (last new review 6 months earlier)
- AI search citations: zero that I could find
May 2026 (six months in):
- Organic traffic: 6,890/month (+64%)
- Organic conversions: 521/month (+86%)
- Local Map Pack appearances: ~2,100/month (+147%)
- Reviews: 78 (+31, all from the last 4 months)
- AI search citations: 14 distinct queries, multiple platforms
Cost comparison:
- Old agency cost (6 months): 3,000 euros + 1,200 euros in extra content = 4,200 euros
- New approach cost (6 months): 480 euros in AI tools + ~100 hours of my time
If I value my time at fifty euros per hour, the total cost was 480 + 5,000 = 5,480 euros — actually slightly more than the agency. But the results were, by every measure, dramatically better. And the work I did was an investment in skills I now have permanently.
What I would do differently
If I were starting over, three things would change.
First, I would have started with the technical audit on day one of my first agency engagement, three years ago. The fact that my site had basic technical issues for that long is, alla fine, my fault as much as theirs. I should have insisted on understanding what was wrong, in plain language.
Second, I would have written my own content from the beginning. The voice of a real owner who knows his products is, without doubt, the most valuable asset a small business has online. Outsourcing that voice to anyone — agency or AI — is giving up the one thing that distinguishes you from the global ocean of generic content.
Third, I would have spent more time on Google Business Profile and reviews from the beginning. For a local business, these things are perhaps fifty percent of the entire SEO equation, and my old agency essentially ignored them.
The honest verdict
So, did replacing my SEO agency with AI work? Yes. Without doubt. The results speak for themselves and the numbers are not flattered.
But — and this is important — it worked because I treated AI as a tool, not as a replacement for thinking. Every meaningful gain came from a combination of AI capability and human judgment. The audits found problems, but I had to decide which to fix first. The drafts gave me starting points, but I had to make them sound like me. The platforms surfaced opportunities, but I had to act on them.
If I had simply bought an AI tool, run it on autopilot, and walked away, I am quite sure the results would have been worse than the agency, not better. The AI gives you leverage. It does not give you results.
If you are a small business owner thinking about doing what I did, my honest advice is this: start with an audit. See what is actually broken on your site. Fix the technical things first. Then, slowly, take ownership of your content and your local presence. Use AI to accelerate the work, not to replace the work.
You can run the same kind of audit I started with, for free, at /seo-standings. It is the tool I wish I had found three years ago, before I spent eighteen thousand euros on an agency that, in the end, produced less than I produced for myself in six months.
The truth is, the AI revolution in SEO is not that AI does the work for you. It is that AI finally makes it possible for a curious, determined small business owner to do the work properly themselves. That, in my view, is the real story. And it is a much better one than any agency report I ever received.