213: WHY GOOGLE & CHATGPT ARE IGNORING YOUR "DEAD" CONTENT WITH JONO ALDERSON

The Threshold Has Collapsed

Every marketing strategy for the past century operated on one assumption: get humans across a threshold, whether a physical store entrance or a digital URL, and you control the message. Once they're on your site, you own the experience. You can shape the narrative, run CRO experiments, squeeze value. That era ended quietly, and most marketers haven't noticed yet.

AI agents now intermediate this relationship. They make decisions about what information reaches users without ever sending those users to your website. The whole discipline of conversion optimization assumed you'd have access to the audience. What happens when you don't?

Your Brand Is Everywhere You Forgot About

The concept of a website as a neat bundle of pages on a domain made sense for human consumption. But when ChatGPT trains on the entire web corpus, it doesn't understand that your 2013 Christmas microsite on a separate URL is no longer relevant. It doesn't know your blog subdomain is a separate entity you barely maintain.

Your brand, to these systems, is an aggregation of every statement ever made about you. That includes Reddit threads, YouTube videos where someone drove past your billboard, Glassdoor reviews from disgruntled employees, and the product listing you forgot to update. The smallest unit of consumption is no longer a URL. The boundaries we drew around our digital presence have dissolved.

This creates an attribution nightmare. The measurement frameworks marketers relied on, tying impressions to clicks to conversions, were always somewhat fictional. Now they're completely inadequate. Old-school advertising metrics like salience studies and panel surveys might be all that's left.

The Zombie Web

Consider every dentist website on the planet. They all have the same twelve pages, the same blog posts about teeth whitening, the same generic content. None of it ever said anything new. LLMs have already extracted every piece of value from this commodity content. It's been memorized, synthesized, and compressed into the model's knowledge.

What's left? A husk that functions as a digital brochure for the rare human who types in a URL directly. But as an input to AI systems, these sites are worthless. Why would Google or OpenAI send traffic to content that solves no new problems?

The Machine Immune System

Marketing copy is designed to persuade. It says "we have the best product" when the product is average. It claims "we love our customers" while support tickets go unanswered for weeks. This worked when humans were the primary audience, easily distracted and susceptible to emotional appeals.

AI systems process this differently. When a model encounters persuasive language that conflicts with signals from reviews, forums, and other sources, it treats that marketing copy as noise. Potentially even as a hallucination to be filtered out. The immune response activates.

The incoherence penalty is real. Machines can spot the gap between your homepage claims and your three-star average on consumer review sites. They aggregate sentiment across sources you don't control and probably don't monitor. That dissonance becomes a credibility penalty applied to everything you publish.

Why llms.txt Won't Save You

Some have proposed creating separate, agent-friendly versions of websites. Clean markdown files that give AI systems exactly what they need. The problem is obvious once stated: why would a machine trust a special file that differs from your public-facing content?

If your website says one thing and your llms.txt says something cleaner, that's another incoherence signal. You're essentially asking machines to trust your curated summary over the messy reality they can observe elsewhere. They won't.

Upstream Engineering Is the New SEO

The optimization that matters now happens far from your website. It's your return policy, your shipping speed, your customer service response times, your product quality. These operational realities generate the signals that AI systems actually trust.

Title tags and meta descriptions still exist, but they're cosmetic compared to fixing why customers complain on Reddit. The engineers and operations teams who handle logistics, fulfillment, and support are now doing more for your search visibility than your SEO team. That's a difficult organizational conversation, but it's the reality of 2026. The brands that survive will be the ones that stopped optimizing their marketing and started optimizing their actual business.