The problems change. The underlying questions often remain.

Making websites searchable: a patent history

In 2005 I was working on a problem where website architecture meets search: how to make the information inside a website easier for search engines to discover. The patent application that came out of it, “Method and system for improving the searchability of a web site,” was published in 2007.

Search engines are no longer the only systems reading websites. AI systems retrieve information, generate answers, and increasingly sit between people and the products, services, and businesses they discover. The technology is different, but the question is still relevant: how should information be structured so machines can find it, understand it, and use it?

The patent records a specific technical idea from a specific period. It doesn’t mean I predicted modern language models or today’s AI search systems. What interests me is how that original problem relates to the new ones appearing around machine-readable information.

From search engines to AI systems

Traditional search made discoverability a central concern for anyone building a website. Technical accessibility, information architecture, and relevance helped decide whether a page could be found and understood.

AI adds questions. Can a system retrieve the right information? Does it have enough context to interpret it correctly? Can it tell a reliable source from a weak one? And when does a website become useful to an AI system, beyond showing up in a list of links?

Some of these extend familiar problems, and others need new approaches and new evidence. I want to investigate that difference rather than assume everything built for traditional SEO will carry straight over to AI.

AI, distribution, and digital business value

AI is changing how software is built, how products are discovered, and how work gets done, and all of that matters to a digital business.

Discovery is getting more complex

Search rankings still matter, but they’re no longer the only way a business gets discovered. AI-generated answers, recommendations, specialist communities, directories, and other sources can influence which products people consider. So a ranking is only part of it. Can the business be found, understood, and chosen in the channels that matter to its customers?

I’m interested in how these discovery mechanisms work, what can be measured, and which signals actually lead to qualified demand.

Software is getting easier to build

AI can reduce the effort it takes to build software, automate workflows, and test product ideas. That opens doors for smaller teams, and it also makes it harder to rely on software alone to stand out.

Distribution, proprietary data, customer relationships, accumulated knowledge, and the ability to execute reliably may matter more relative to the cost of building a product. So beyond what AI makes possible, a founder has to ask which capabilities create an advantage competitors can’t easily reproduce.

Automation has to work in the real world

A working demonstration is not yet a dependable business system. AI-based workflows have to run within real constraints: data quality, cost, reliability, integration, oversight, and the consequences of mistakes. The value comes from changing an outcome, not from adding an AI component to a process.

My interest is in how the technology connects to the business economics around it.

What I am investigating

I want to build this work from original experiments, technical observations, and documented case studies, rather than from predictions alone.

How AI systems discover the web

What can server logs show about AI crawler activity? How does crawling relate to indexing, retrieval, citations, and referral traffic? Which of these can be measured directly, and which stay hard to establish?

What makes a digital business defensible

As software gets cheaper to create, which assets keep compounding in value? How should founders and investors judge their dependence on acquisition channels, third-party platforms, proprietary data, and customer relationships?

Where AI creates operational value

Which workflows benefit from automation in practice? How do implementation cost, human review, reliability, and maintenance change the economics? What separates a promising prototype from a system you can trust in production?

These are research questions, and I haven’t settled the answers in advance. I’ll publish findings when there is enough evidence to make them useful.

A note on evidence

I treat historical records, direct observations, measured results, and hypotheses as different things. A patent publication documents a technical record. An archived website can show what was available at a particular time. A server log can show requests to a server. An experiment can test a specific hypothesis under defined conditions. None of them proves a broader claim on its own.

When I publish research, I want the source, method, scope, and limitations to be clear enough for you to judge the conclusions yourself. Where the evidence is incomplete, I’ll say so. The goal is better decisions with the evidence available, without claiming a certainty that isn’t there.

About this work

I trained as a mechanical engineer and have worked across search, digital growth, product development, and AI. I’ve held operating and leadership roles, co-founded a technology company, and built digital products, including an AI-based valuation engine. I’m based in Stockholm and work with businesses internationally.

This page will grow as I document more of the historical material and new research becomes available.