Detailed Analysis
Substack CEO Chris Best's interview surfaces a problem that has been quietly metastasizing across the internet for the past year: the flood of AI-generated content, or "slop," that is eroding trust in online writing. Best cites a striking statistic from Pangram, an AI text-detection company, showing that roughly 40% of long-form writing on LinkedIn is now fully AI-generated. While he's careful to note that AI-generated content and "slop" aren't perfectly synonymous — not all AI writing is soulless, and not all soulless writing is AI-made — the broader trend he identifies is unmistakable: the proportion of human-authored text in major online spaces is collapsing rapidly, and platforms are largely unprepared to deal with the consequences.
What makes this conversation notable is the specific example Best and his interviewer land on involving Claude, Anthropic's AI model. The illustrative worst-case scenario they discuss — telling an AI system to "write a thousand viral posts, make no mistakes, off you go" — captures a qualitatively different use of generative AI than someone using it as a tool to refine or accelerate work they actually believe in. Best draws a clear distinction between AI as "a paintbrush," used in service of genuine authorial intent, and AI as a cynical mechanism to mass-produce content divorced from any human conviction or accountability. This framing is significant because it doesn't treat AI itself as the enemy; instead, it locates the ethical problem in the intent and transparency behind its use — a distinction increasingly central to debates about responsible AI deployment.
The stakes Best describes go beyond content quality — they strike at the epistemic foundation of online discourse. He argues that mass-produced, intent-free AI content functions like "a denial of service attack against the public square," because it forces every reader into a state of perpetual suspicion, unsure whether anything they're consuming reflects real human thought. This matters enormously for platforms like Substack, whose entire business model and cultural positioning depend on being a trustworthy alternative to algorithmically driven, engagement-optimized content ecosystems. If AI-generated material becomes indistinguishable from human writing at scale, the value proposition of "real people writing things they believe in" erodes unless platforms actively signal and safeguard authenticity.
This conversation reflects a broader inflection point in the AI industry, where the focus is shifting from raw model capability to downstream consequences of ubiquitous access to powerful generation tools. Anthropic, Claude's creator, has positioned itself around themes of responsible AI use and human-AI collaboration rather than replacement, and this interview implicitly reinforces that narrative — the problem isn't Claude or generative AI as a technology, but how thoughtlessly or cynically it gets deployed. As detection tools like Pangram proliferate and platforms grapple with provenance, authenticity, and disclosure, the industry appears to be entering a phase where distinguishing human intent from automated output becomes as important as the underlying AI capabilities themselves — a challenge that will likely shape platform policy, detection technology, and public trust in digital content for years to come.
Read original article →