I Stopped Asking AI for SEO Ideas and Started Feeding It Search Console Data Instead

In August, the website Borderfolio began appearing on the first page of Google search results within 25 days of launch. However, the pages were ranking for very specific queries, such as "spy irish domiciled schd" and "QQQ SPYL TER 0.03% depotauszug." The interesting aspect was that Google was giving insights into its understanding of the product.

Instead of the traditional SEO approach of keyword research, writing articles, and waiting, the process was transformed into an engineering feedback loop: Search Console data → telemetry → hypothesis → one change → measurement and repeat. The company Borderfolio focuses on issues faced by non-US investors holding US or Irish-domiciled ETFs, such as ETF domicile dividend withholding tax and UCITS alternatives.

This niche search surface led to the creation of content around specific queries, rather than generic keywords. The SEO strategy involved connecting Search Console data to an MCP workflow using Windsor.ai, enabling ChatGPT to access relevant metrics and analyze the data daily. The system would then recommend a single, high-value SEO change, or suggest no change if the results were minimal.

This approach quickly moved some new pages to positions 5-10 without optimization, allowing the experiment to run until more data was collected. Additionally, Search Console data revealed that country-specific withholding-tax pages generated impressions but had low average positions, indicating they were targeting broad queries rather than specific investor needs.