"The web is eating itself and your metrics look fine." The scariest SEO headline of the week deserves a plain-language translation.
Duane Forrester published a piece in Search Engine Journal this week with a headline we have not been able to shake: the web is eating itself, and your metrics look fine. His argument names three documented mechanisms, source bias, retrieval collapse, and model collapse, that together explain some of AI search's strangest behavior (searchenginejournal.com). We recommend reading the original; his mechanisms deserve his own framing. What we want to add is a translation of the headline's second half, because that is the part aimed at you.
"Your metrics look fine" is the trap
Every dashboard a business owner watches was designed for the old web. Sessions, pageviews, bounce rate, rankings. All of them measure one event: a human arriving at your site.
But an increasing share of your real audience now meets you without arriving. They ask an assistant, the assistant reads the web, including possibly you, and answers. If you were the source, you did valuable brand work and your analytics recorded nothing. If your competitor was the source, you lost a customer and your analytics recorded nothing. Same flat line, opposite outcomes.
That is what "your metrics look fine" means. Not that things are fine, but that your instruments cannot register the place where things are changing. A thermometer does not detect a flood.
What the feedback loop means for small sites
Forrester's deeper worry, per his piece, is circular: AI systems increasingly learn from a web increasingly filled with AI output, and retrieval narrows toward already-cited sources (searchenginejournal.com). We will not pretend to know how that resolves. Honestly, nobody knows.
But one implication seems safe to act on either way: original, firsthand, verifiable content becomes more valuable as recycled content floods in, because it is the only thing the loop cannot generate for itself. The business owner who publishes real prices, real photos, real answers about their actual products, marked up so machines can read them cleanly, is feeding the machines the one nutrient that is getting scarcer.
Measure what the machines see, not just what the humans do
The practical move is to add a second instrument next to your analytics: a record of what search and AI engines can actually read on your site. Can the crawlers get in? Does your content survive without JavaScript? Are your facts consistent and quotable? Is your structured data valid and honest? None of that shows up in a traffic graph, and all of it decides whether you exist inside the answers.
That is the instrument we built. The free scan reads your site the way Google and the AI answer engines actually read it, and tells you in plain language what they see today. No email required to get your diagnosis. Your dashboard says everything is fine. It is worth five minutes to check whether the machines agree.
Source: Duane Forrester in Search Engine Journal, "The Web Is Eating Itself And Your Metrics Look Fine," July 9, 2026 (searchenginejournal.com). The three named mechanisms are from his piece; the translation and recommendations are ours.