Memorandum
- From
- Delaney Sawyer via Fast Company
- Date
- Filed
- Culture·6 min to read
- Re
How to Sound Human in the Age of AI Slop
ReHow to Sound Human in the Age of AI Slop
As AI-generated content floods the internet, writing experts say the best defense against sounding like a machine is to build specific expertise, share personal anecdotes, and take clear positions.
AI-generated content has become so pervasive that The Economist named «AI Slop» its 2025 word of the year, warning that «slop merchants clog up the internet with drivel.» The term captures a growing frustration: search for a health question and you will find endless articles filled with AI-written filler; scroll through Instagram and you will soon encounter fake clips stitched together with an AI voice-over. The sheer volume of this material raises a question for anyone who writes for a living or simply wants to communicate clearly: if we are surrounded by AI slop long enough, will we start to sound like it ourselves?
That concern prompted one writer to test his own work through several AI-detection services. The results were reassuring — his writing still registered as human — but they also surfaced practical lessons about how to communicate distinctively in an era when machines can mimic general patterns of human speech. The first marker of human writing, according to the detection tools, was specific domain expertise and synthesis: the ability to draw on knowledge from different fields and connect it in unexpected ways. AI can learn broad patterns, but it struggles to replicate the kind of synthesis that comes from deep, field-specific experience.
Creativity researchers have found for decades that expertise is essential for top-notch creative work, and that expertise must be specific to a particular domain. Performance expert Anders Ericsson, who studied expertise throughout his career, identified deliberate practice as another critical element — identifying the components of a skill, getting feedback, and working deliberately on weak areas. That process goes far beyond the pattern matching AI uses to learn. The implication is that the best way to avoid sounding like an AI is not to eliminate superficial tells like em dashes, but to build genuine human expertise by reading widely, trying things, failing, and accumulating a database of experiences that is uniquely yours.
Another lesson concerns specificity. Many of the strategies AI uses to sound human resemble the techniques mentalists use to create the illusion of supernatural powers. They rely on general statements that encourage people to project their own ideas and feelings onto whatever is being said. Barnum statements, for example, are vague personality descriptions that seem tailored to an individual but could apply to almost anyone. The rainbow ruse pairs a statement with its opposite, and subjective validation exploits the human tendency to map general concepts onto personal experience. An AI might tell you that you are generally private but become outgoing around people you trust — a Barnum statement — or that you are kind and considerate but feel deep anger when someone breaks your trust, a rainbow ruse.
Human writing, by contrast, offers specific, personal anecdotes. Instead of a vague paragraph about trust, a human might describe a particular time someone broke their trust, how it felt, and what they did about it. The more concrete the experience, the less work the reader has to do — and the more distinctly human the writing becomes.
A third lesson is to have an opinion. When AI is used for editing, it often tries to dial back writing, gravitating toward safer, more generalized language. Humans form points of view tied to their own experiences and expertise, and those views are a key component of identity. As mathematician G. H. Hardy once put it, «It is not worth an intelligent man's time to be in the majority. By definition, there are already enough people to do that.» Offering a distinctive point of view means taking a stand on an issue worth caring about. AI hedges; people should not. Writing to please everyone is pointless, and if you build a significant platform, you will attract criticism. The only alternative is to be boring, which is exactly what AI slop does — offering generic information and advice that offends no one and is rarely useful.
Finally, writers can think of themselves as their own learning corpus. AI consists of three components: compute, algorithms, and the learning corpus. When using AI, the learning corpus is the part you control most directly. If you ask AI to draw from the broader internet, it will learn from everyone, regardless of quality, and may even learn from its own slop, leading to what experts call model collapse. But by controlling the material it works from — feeding it your own writing, your own research, your own experiences — you can keep your voice distinct. The broader takeaway is that sounding human is not about avoiding certain punctuation marks or tricks. It is about building a life and a body of knowledge worth sharing.
