You’ve probably noticed that searching for software recommendations has become completely impossible. You go to Google, you get a list of sponsored garbage. So, you do what we all do now: you ask Perplexity or ChatGPT, expecting a clean, unbiased answer. You think you’re bypassing the marketing swamp. You’re not.
We thought AI would bypass the SEO swamp. Instead, it built a faster, dumber boat to sail right through it.
Three websites recently generated 215,128 “best software” pages. Not for humans to read. For AI to scrape. They aren’t even trying to get you to click their links. They are designing pages specifically to be ingested, cited, and regurgitated by LLM search engines. And it’s working flawlessly.
Search engines used to filter noise and surface reliable answers. Now, they are amplifying synthetic scale. When an LLM looks for authority, it counts heads. If it sees 50,000 pages saying “XYZ Software is the best CRM,” it assumes a consensus exists. But there is no consensus. There is just one bot printing 50,000 variations of the same lie.
AI isn’t laundering facts. It’s laundering manufactured consensus.
This is a creeping betrayal. The entire promise of AI search was that it would synthesize the web for you, cutting out the SEO spam. But the SEO spammers just leveled up. They aren’t writing for your eyes anymore; they’re writing for the AI’s algorithm. They know exactly how to trick the model into citing them.
What happens next? The AI you trust to bypass marketing is itself being gamed by machine-generated marketing. We are feeding LLMs their own regurgitations, creating a closed loop of exponential amplification of all their flaws. When the machine decides what is true based on what other machines said, the concept of a ‘source’ stops existing.
If you’re using AI for product research right now, you are relying on an ecosystem optimized for the AI’s approval, not for human accuracy. The AI SEO wars have begun, and the machines are winning. Next time you ask an LLM for a recommendation, remember: you aren’t getting an objective truth. You’re just getting the loudest synthetic echo in a room full of screaming bots.
FAQ
Q: What protection do LLM search engines have against training off content generated by other LLMs?
A: Almost none. Right now, LLMs are incredibly vulnerable to scale. If a bot farm generates 200,000 authoritative-looking pages, the LLM treats it as a consensus. They are trying to build filters, but the spammers are moving faster.
Q: What's the practical implication for everyday users?
A: Stop trusting AI for product research. If you don't already know what you're looking for, you're screwed. Use AI for code, use it for writing, but assume its 'best of' lists are just regurgitated marketing.
Q: What's the contrarian take?
A: This isn't a bug; it's the inevitable endpoint of the internet. The web has always been a garbage pile of manipulated signals. AI just made the manipulation faster. The only solution is human curation, which means we're going back to web directories from 1998.