Update, October 2026: since I wrote this, a lot of searches get an AI-written answer before any links, from Google’s AI Overviews, ChatGPT search and the like. Those answers cite and link the pages they drew on, and the easiest page to quote is a useful one that says the answer plainly. That makes the dull rule at the end of this post matter more, not less. I’ve written up what I changed on this blog because of it in Optimising a Blog to Be the Source AI Assistants Cite.

Why following the crowd mostly works

#

Best practices are best practices for a reason. They’re distilled from data across thousands of sites, and if you’re new to SEO they’ll get you most of the way there without drama. Staying on the beaten path also means fewer nasty surprises — you’re unlikely to get punished by an algorithm update for doing exactly what Google’s own documentation recommends. And since the algorithms keep shifting, just keeping up with the current guidance is honest work in itself.

The catch is that everyone else is reading the same guides. When every site in a niche has the same structure, the same keyword-heavy headings and the same FAQ block bolted to the bottom, none of them stand out. You also end up permanently reactive: waiting for the next update, then scrambling to comply.

The case for trying your own thing

#

Experiments are how you find the tactics that fit your site specifically. What works for a cosmetics community isn’t what works for a recipe blog, and no generic checklist will tell you the difference. If an experiment lands, you get a stretch of time where you’re the only one doing it. And if it lands publicly, people start treating you as the one to copy, which has its own compounding value.

The honest downside is that most experiments fail quietly. You can sink weeks into one and have nothing to show for it. If search traffic is what keeps the lights on, that’s a hard sell to whoever approves your time.

Where I’ve landed

#

Do the boring stuff first, and do it properly. Then carve out a small, contained slice of effort for experiments, and actually measure them instead of judging by vibes. A small site should lean heavily on the established stuff and only experiment at the margins; a big organisation with resources to burn can afford to place real bets.

The one rule that’s survived every algorithm update is also the least exciting one: make something genuinely useful for the person searching. Everything else is fashion.