Still works — just not in the niches people default to. The Google AI answer problem is real but it mostly hits generic informational queries. Anything with specific, structured data that users need to act on tends to hold up much better.
Been running a niche analytics site in the greyhound racing space (trapstats.co.uk) and the SEO pattern is very different from a typical blog. Race data, form guides, track-specific statistics — this stuff doesn't get summarised by AI overviews because the freshness and specificity requirements are too high. The content is also transactional by nature, so dwell time and return visits are strong.
The niches that seem to survive 2026 Google in my experience: anything requiring real-time or frequently updated structured data, comparison tools with proprietary datasets, and deeply community-specific content where trust signals (forums, communities) matter more than domain authority.
The old "write 3000 word guides and rank" model is rougher now. The sites doing well are ones where the actual data or tooling IS the content.