Search, and now AI search

The idea

I think search is the one channel where people tell you what they want, in their own words. You rank by being relevant, trusted and fast. Win the long tail first, and write headings that say what the page is.

AI answer engines read the same signals, only more literally. They want plain headings, a direct answer under each one, and the same facts about you on every page. The problem I see most often is a site written like a brochure when people came for an answer.

What a crawler can see

Kiln's site has fourteen pages, and I broke four of them on purpose. Nothing links to one page. One page shows its text only after a script runs. One image has no description, and one page's headline says nothing about the page. Watch the crawler walk the links from the home page, then switch on a fix and let it crawl again.

Why the long tail wins

Here are twenty terms Kiln could chase. The big ones, up and to the right, are hard to rank for and full of people only browsing. Down and to the left sit the narrow terms. Fewer people search them, but they're easier to win, and the people typing them are close to buying.

Where I used it

• A group audit of a small merchandise company for artists. A teammate ran the SEO audit, and my recommendations built on it. The first step was to fix the foundation: real product listings, alt text, ordered headings and a faster site. Group work. The audit was a teammate's; the recommendations were mine.

What I'd do now

In June I specified a pipeline to check a site's search health every day, then parked it to finish the course. The next pass turns it into a small open audit tool: point it at any public page and it reports the headings, the alt text, and whether each section answers a question in its first sentence. It isn't built here. This is the rewrite it would push for on Kiln's worst page: