How I took a tax calculator site from near zero to 100K Google clicks using AI

How it started

About a year ago I was watching YouTube videos about getting rich over a weekend using AI. I have always thought that if something is too good to be true, it is not true. Then I found a video about someone making $10K a month from a calculator website with very little work. I was very skeptical.

But I thought, why not? I spent a while using AI to research whether I should build a calculator, and what type. I found there was space for another international tax calculator. There was competition, but it was not overly crowded. So I bought a domain, and I was pretty chuffed with the name: countrytaxcalc.com.

Does Google allow AI-written pages?

I did some research before I started, and what I found was that I could scale a site with AI, but only if I stayed within Google’s guidelines. Google’s page on generative AI content (last updated October 1, 2026) does not ban it. It says that using AI tools to generate many pages without adding value for users may violate its spam policy on scaled content abuse, and that it is critical to manually fact-check and review AI-generated content before publishing. Read Google’s page. My early mistake was the second half of that.

The accuracy problem

For the first few months I used AI to create loads of pages and assumed everything it produced was 100% accurate. Then I started spot checking the tax data and found it was hallucinating a lot of information.

So I took a step back and audited the whole site, found loads of mistakes and fixed them. Then I built a workflow where everything I publish gets checked against official government pages, or the closest I can find. I also keep a record of when each page was last updated, a sort of “tax calendar”, so I do not miss a country changing its rules. Most of the time I spend on the site now is this kind of maintenance.

What changed the growth

Another turning point was connecting Claude Code to Google Search Console through a free, open-source MCP server called mcp-server-gsc. From then on I created content based on near misses and real data, not random pages.

I started with about 40 countries plus all the US states. Over the next few months I built it up to 111 countries, plus every US state and Canadian province. Clicks from Google went from a few a day to hundreds a day.

The numbers

The chart above is the site’s own Search Console data, clicks only. Traffic was close to zero for months. It took off between April and June 2026 and has held at roughly 560 to 700 clicks a day on monthly averages since then; October is a little lower so far. It is not still climbing.

Figures from a Search Console export and the Pages report, dated as shown. Not every page is indexed: 612 were not indexed in October 2026.

The money

I waited to add ads until I qualified for Raptive. As I remember it, I needed about 25,000 monthly page views, so check Raptive’s current requirements if you are planning around it. The site has been monetized for a few months. In September and October 2026 it earned about $1,000 a month.

What I’d tell someone starting today

It has not been all plain sailing. It has taken hundreds if not thousands of hours, evenings and weekends in front of my computer. Here are my four takeaways.

1. Accuracy is the moat. Anyone can generate 2,000 pages with AI. Very few people will check every figure against an official source.
2. Build the checking process before you scale. I wish I had done it in month one, not month four.
3. Target real searches. I did not build pages just to have pages; each one answers something people actually search for.
4. Be patient. Traffic was close to zero for months before it took off. I almost certainly would have quit if I had judged it in the first quarter.

The skills I use to build and check a site are free: download the Bevizio SEO skills.

What AI assistants say about the site

I wanted to build an AI visibility machine so founders can see this, because I see it becoming more and more relevant. So I pointed my own product, Bevizio, at this site. Bevizio is mine, so weigh this section accordingly.

Between September 20, 2026 and October 18, 2026, Bevizio asked ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews 20 questions a tax calculator site should answer, repeatedly. Of 2,377 answers it could read, 1,045 (44%) named countrytaxcalc.com. In 773 of those (74%) the name was in the answer text. In the rest the site appeared only as a cited link.

It depends heavily on the question. The question “Which website provides updated tax calculations for over 100 countries?” named the site in 102 of 122 answers. That question almost describes the site, so read it as the best case. “How can I calculate my take-home pay after taxes if I work in multiple US states?” named it in 1 of 118. When it was not named, the assistants most often named SmartAsset, OECD, Tax Foundation, Taxes for Expats and PwC.

What this can and cannot show: one site, one set of 20 generated questions (not what real people type), a 28-day window, and a count of mentions, not of visitors. It does not show that ranking in Google causes an assistant to name you. How an answer is collected and judged is on how we measure. To see the same for your own product, run the free check.

What’s next

I am working on Bevizio, which shows whether AI tools like ChatGPT recommend a product (about Bevizio), and on the free SEO skills I used to build and check all of this. I share what I learn as I go. Follow along: @danielkbuilds. If you are building something similar, I would like to hear how it is going.

How this page was made: the story is adapted from my X article of October 15, 2026. Bevizio checked and dated the figures, and the page was edited with AI assistance. This page gives no tax advice. Written October 18, 2026; figures are dated where they appear.