← All Field Notes
AEO July 2026

Your Traffic Is Down and Your Content Is Working

By Jack DeManche

A client pulled up their analytics last month, pointed at a 30% drop in organic traffic, and asked what went wrong. Nothing went wrong. That's the part nobody's ready to hear.

Their content was being read more than ever. It just wasn't being clicked. The traffic didn't leak to a competitor or fall off a ranking. It got intercepted โ€” answered in place by an AI system that read their page, summarized the useful part, and never sent the visitor along.

This is the mechanic worth understanding, because it changes what a website is for. When someone asks ChatGPT, Perplexity, or Google's AI Overviews a question, the model doesn't hand back ten blue links. It composes an answer and, increasingly, names its sources. If your page is one of those sources, you won. The user got what they needed, your brand got named as the authority, and your server logged nothing. The value transferred; the pageview didn't.

The job of the page changed

For twenty years the goal was to earn the click โ€” headlines engineered for curiosity, pages built to pull a visitor in and route them toward a conversion. That content is optimized for a funnel that's narrowing. The content that wins now is optimized to be quoted. Clean structure. Self-contained sections that make sense without the paragraph above them. Specific, checkable claims instead of hedged generalities. A direct answer in the first sentence, before the elaboration. These aren't style preferences โ€” they're the features that make a passage easy for a model to lift and safe for it to cite.

Here's the tell. Take one of your pages and read only the first sentence under each heading. If those sentences, pulled out and lined up, still answer the questions a customer would actually ask, the page is built for extraction. If they're throat-clearing โ€” "In today's fast-paced landscape" โ€” the model has nothing to grab, and it'll quote whoever wrote more plainly than you did.

Owned beats rented, again

There's a deeper shift underneath the tactics. For a decade the growth playbook was to rent attention โ€” post on the platform, ride the algorithm, chase the feed. Answer engines don't reward rented attention. They reward source material, and source material has to live somewhere they can read it: structured, durable, on a domain you control. The owned website, written off as a brochure a few years ago, turns out to be the one asset that feeds the answer layer. You can't be cited for a post that scrolled past. You can be cited for a page that sits still and answers one question well.

So the drop in the dashboard isn't a failure to fix. It's a number that stopped measuring the right thing. The question isn't "how much traffic did we get" โ€” it's "how often are we the answer." That's a harder thing to see, which is exactly why most teams are still staring at the one that's easy.

The soft spot: we don't have good tooling yet for measuring citation share. Tracking how often you show up inside AI answers is manual, partial, and platform-by-platform โ€” and the platforms could change how they surface sources tomorrow and reset the game. I'm confident the click is fading as the primary signal. I'm much less sure what clean replacement we get, or when.

This post was written collaboratively with AI. The ideas, arguments, and editorial judgment are mine. The drafting process involved AI assistance for structure, phrasing, and pressure-testing the argument.