Substack is integrating AI tools directly into its platform, but CEO Chris Best's real argument isn't about convenience, it's about a fundamental misunderstanding of what makes content good or bad. The platform is building a Model Context Protocol (MCP) server that will let AI assistants like Claude and ChatGPT read, write, and act on creator accounts. This move reflects a broader industry shift toward AI integration, but Best's philosophy about content quality offers a fresh perspective on alignment in AI systems.

What Is "Slop" and Why Does It Matter for AI Alignment?

When Best talks about low-quality AI-generated content, he reframes the problem entirely. He argues that "slop" is not something AI invented; it's content made "without intention," and AI simply scales it massively. This distinction matters because it shifts focus from the technology itself to the human choices behind it. If creators use AI tools thoughtfully, the output reflects that intention. If they use them as shortcuts to pump out volume, the result is indistinguishable from any other low-effort content.

"Slop is not a thing that was made by AI. It was a thing that was made without intention. AI didn't create the problem; it just massively scaled it," said Chris Best, CEO of Substack.

Chris Best, CEO of Substack

This perspective connects directly to how AI systems are trained and aligned. Reinforcement Learning from Human Feedback (RLHF), a technique used to fine-tune large language models based on human preferences, can inadvertently reward engagement-driven behavior over quality. Best noted that his experience at Business Insider, where he was monitored and performance-tracked, created a form of RLHF that "drove a race to the bottom." The platform's metrics incentivized clickbait and sensationalism rather than depth.

Best

How Can Creators Use AI Tools Responsibly?

Best's vision for responsible AI use on Substack centers on intention and alignment with creator values. Rather than banning AI outright, he proposes that creators should use these tools as amplifiers of their existing work, not as replacements for it. His dream tool would auto-clip podcasts, post them across networks, and translate them into multiple languages, all while maintaining the creator's voice and vision.

• Maintain Intention: Use AI to enhance work you genuinely believe in, not to generate content for its own sake or to game algorithms.
• Leverage Distribution: Let AI handle repetitive tasks like formatting, translating, and cross-posting so creators can focus on original thinking and relationship-building with readers.
• Align Incentives: Build on platforms where reader subscriptions, not algorithmic engagement metrics, determine success, reducing pressure to create sensational or low-quality content.

This approach mirrors broader conversations in AI alignment research about how to design systems that reward genuine quality over proxy metrics. When platforms measure success by reader subscriptions and long-term relationships rather than clicks or impressions, the incentive structure naturally discourages slop.

Why Does Substack's Subscription Model Matter for AI Alignment?

Best emphasized that subscription-oriented metrics are "more tethered to long-term reader value and less to sheer brain-hacking attention spikes." This distinction is crucial for understanding how platform design influences AI behavior. On algorithmic feeds driven by engagement metrics, AI systems optimized for those metrics will naturally produce sensational, low-quality content. On subscription platforms, the incentive structure flips.

Best

Substack's approach also reflects a philosophy about free speech and editorial responsibility. Best argued that readers, not the platform, should decide who gets paid. This decentralization of editorial judgment means Substack doesn't curate content the way traditional media does, but it also means the platform's success depends on creators building genuine relationships with audiences. That dynamic naturally filters out low-effort content because readers won't pay for it.

The platform's growth metrics support this model. While Best wouldn't share specific numbers, he noted that Substack continues to grow steadily and that the number of creators earning over one million dollars annually is now "a lot more" than the previously cited figure of more than 50 creators. This suggests that the subscription model, combined with thoughtful AI integration, can sustain quality content creation at scale.

What Does This Mean for the Future of AI Integration?

Substack's MCP integration is part of a broader industry convergence. Beehiiv, another newsletter platform, has already opened itself to MCP, and a growing list of platforms are wiring themselves for AI integration. Best's argument is pragmatic: if creators want to use AI tools, platforms have to meet them there. Refusing integration doesn't stop creators from using AI; it just means they'll use it elsewhere.

The real question isn't whether AI should be integrated into creative platforms, but how. Best's framework suggests that the answer lies in alignment with creator values and reader incentives. When platforms reward intention over volume, and when creators have tools that amplify their genuine work rather than replace it, AI becomes a force multiplier for quality rather than a factory for slop.

As Best noted, "Make yourself legible to AI" is becoming the new SEO cliché. But beneath that catchphrase is a deeper insight: the platforms and creators that thrive will be those that use AI thoughtfully, in service of genuine relationships with audiences. That's not a technical problem; it's an alignment problem, and it's one that Substack's approach to platform design and creator incentives is actively trying to solve.

As Best