I asked AI who to trust. It ignored the pages that ranked.

I ran a small test last week, the kind anyone can repeat in ten minutes. I took a handful of ordinary buyer questions, the sort people used to type into Google and now say out loud to ChatGPT and Perplexity, and I watched two separate things. Which pages ranked. And which sources the AI answer actually named.

The pages that ranked were often syndicated articles. A useful piece of research gets republished across a wire of news sites, each a real domain with real authority, and several of them land on page one. By the old scoreboard that is a clean win. The content traveled. The links accrued. The brand got a mention on sites a small company could never crack on its own.

Then I asked the answer engines the same questions. The republished articles were not in the citations. Not lower down. Not at all. Perplexity reached past every syndicated copy and every ranking page and quoted the original instead, the government agency, the university study, the trade body that first produced the number. The page that ranked and the source that got cited were two different things.

The reason is plain once you see it, and it should change how you publish. Old search rewards the page that earns attention. An answer engine is doing something else. It is trying to name where a fact came from. So it steps over the site that repeated the number and points at the one that owns it. If your content restates a statistic a research institute reported first, the machine will cite the institute, and it will do that no matter how many places your version shows up.

That is a rough lesson for the reach-first habit, and a clear instruction for what to do instead.

If you want an answer engine to quote you, be the source, not the echo. Publish something the machine cannot find upstream. Your own numbers. Your own count. The data your business is the only one sitting on top of. A parts seller knows which fitment questions actually get asked all day. A marketplace knows what prices did this quarter. A shop knows how long the job really takes. Those are facts with no authority above them, so when an answer engine reaches for the source, there is nowhere to reach but you.

Reach still counts. Getting your own data carried across authoritative sites is how it spreads and how the tie between your name and your subject gets built over time. The order is what people get backward. Originate the fact, publish it where you own it, then let distribution carry it out. Distribution amplifies whatever it points at, so if it points at someone else's number, you are paying to send buyers down the road to their door.

The web spent twenty years teaching businesses to fight for the top of a page. The answer engines are quietly rewarding something older and harder to fake. Have something true that is yours to say.

Citala checks whether AI search cites you, then builds the pages that make it happen. If you sell from a catalog and your traffic has been sliding, the first step is a free citation check.