I learned more about SEO by trying to help sparking.chat grow than by reading generic SEO advice. Later, I applied the same workflow at HY.AM Studios.

How it started

My first real encounter with SEO was with Fotokatalog, a product I built earlier. Back then I hired an SEO agency to handle it. They did the audits, wrote the recommendations, ran the reports. It helped at the time, but looking back, everything they did is something I can now do on my own. That experience planted the seed, but I did not have the workflow yet.

The workflow clicked when I helped my wife build her first product, sparking.chat. Small enough to experiment on, real enough that bad decisions showed up in a week. The SEO work there never felt like content marketing in the abstract. It felt like research, product thinking, positioning, and systems design all mixed together. When I later brought the same approach to HY.AM Studios for service pages and search visibility, the core lesson held: SEO gets much more useful the moment you stop treating it as a publishing problem and start treating it as an opportunity-mapping problem.

Most SEO advice still boils down to "publish more, write longer pages, find more keywords, keep shipping articles." There is some truth in that. But it dodges the harder question: what actually deserves to exist? Where do you already have signal? Where are you almost visible but not yet convincing enough? Which pages need improvement versus replacement? Which queries fit your positioning, and which ones are vanity?

Once I reframed the work around those questions, the entire process changed. Instead of asking "what should we write next," I started looking for where opportunity was already structurally visible.

The tools, and what each one actually did

For sparking.chat, the workflow came from layering three data sources, each doing something the others could not.

Search Console was the most honest input. It shows where real demand already exists, where pages are getting impressions, where the site is close enough to relevance that improvement might actually move the needle. Some of the best-performing pages on sparking.chat started as afterthoughts that were already pulling impressions before I ever thought to optimize them. Most interesting SEO opportunities are like that: already there, buried under weak page structure or vague positioning, waiting to be noticed before someone goes chasing greenfield keywords.

DataForSEO made the work feel less like manual tab-hopping. Site audits, keyword research, search opportunity mapping, all coming back as structured data that could actually be reasoned about systematically rather than eyeballed across browser tabs.

Claude Code was the synthesis layer, and the part that made the whole process usable at speed. I used it to work through exported data, cluster opportunities, identify patterns, and compress the distance between raw inputs and decisions: summarizing noisy exports, comparing opportunity buckets, structuring page plans. The SEO industry has mostly used AI to scale content production, but the bottleneck was never production. It was knowing what to produce. Claude Code helped with that second problem.

What changed at HY.AM Studios

The sparking.chat workflow translated surprisingly well into agency work. At HY.AM Studios, the same approach shaped how I thought about service pages, page targeting, and where the company had a realistic shot at building search visibility. I also added Semrush to the stack here for competitive framing: what the keyword landscape looks like, who already owns a space, how crowded a category feels, where adjacent opportunities exist. Search Console tells you where you have traction; Semrush tells you what you are up against.

Some pages did not need rewriting from zero — they needed sharper alignment with the query they were already attracting, and once the page architecture and messaging improved, early traction that looked marginal in a dashboard became genuinely meaningful. Other opportunities looked attractive in tools but were not a fit for how HY.AM Studios actually positions itself. Knowing which to pursue and which to ignore was the real work, and it had more in common with brand strategy than with content production.

Underperformance, in most cases, was a clarity problem. Not thin pages — indecisive pages. A clearer title, sharper intent match, better structure, stronger differentiation. The pages that win are the ones that make sense for the query and for the business behind the page. When those two things are misaligned, no amount of keyword optimization fixes it.

Where this leaves me

The workflow is not finished. SEO stays iterative, and every improvement creates a better feedback loop for the next decision. The open questions are real: how to prioritize across multiple data sources without drowning in spreadsheets, how to balance classic search visibility with whatever AI-search visibility turns out to be, and how to make page planning more systematic without losing the editorial judgment that makes pages actually win.

The pattern I keep seeing across Fotokatalog, sparking.chat, and HY.AM Studios is the same: the systems get you to the right question faster, but someone still has to decide the answer. Whether that stays true as the tools get better is the part I have not figured out yet.