What Experts Share Is a Better AI Signal Than Clicks
The stories that get clicked and the stories that matter are not the same stories.
When we set out to build The State of AI · Q2 2026, we made one decision that felt almost reckless: we threw out our own traffic data.
We have plenty of it — eleven years of clicks, opens, and every link our readers have ever followed. Ranking the quarter by what got clicked would have been the easiest thing in the world. We didn't. Here's why, and what we found instead.
Clicks measure you, not the news
A click doesn't tell you what mattered. It tells you what you promoted. A story buried at the bottom of an issue can't out-click a story in the subject line, no matter how consequential it was. Optimize for clicks and you learn, with great precision, the shape of your own past decisions. It's a hall of mirrors — and every publisher who "follows the data" is, to some degree, trapped in it.
So we built the recap from a different signal: what the experts we track shared on their own, unprompted — the researchers, founders and engineers whose attention is a better proxy for "important" than any headline's click-through rate.
The gap was the story
The most interesting finding was the distance between the two signals.
The single most-shared link of the quarter wasn't a product launch or a megaround. It was Anthropic's public objection after the U.S. government recalled its most powerful model — passed around by 33 of the experts we follow. A story about a company pushing back on a regulator out-traveled every shiny demo of the quarter, in exactly the circles that build this technology.
Then there were the stories almost no one clicked, because almost no one ran them. We found nine genuinely important Q2 developments that the press largely skipped but the connoisseurs quietly traded among themselves — the "under the radar" section turned out to be some of the most-opened material in the whole recap.
Watch what practitioners share, not what trends
The lesson generalizes well past our newsletter. If you want to know where a field is actually moving, watch what the practitioners share with each other — not what trends.
Trending measures reach. Sharing-among-experts measures conviction. And the two diverge most exactly when something important is happening that the mainstream hasn't priced in yet. The gap is the signal.
That's the method behind the entire recap: assembled from the expert-share graph, deduped against our own coverage, every claim linked back to its original source, and our click data deliberately left in the drawer. You can read the result — six forces, the nine under-the-radar stories, and a poll in every section — here: The State of AI · Q2 2026.
— Alexis