The central problem Sandeep Swadia sets out to solve is one that most AI users already sense: text produced by tools like ChatGPT or Claude is instantly recognizable as AI-generated, not because the grammar is wrong, but because it carries no individual fingerprint. Swadia's response is a five-step system designed to put the writer's voice back in control of the machine.

“AI can do all of that. Then, why does AI that we use today write in such a predictable generic fashion?”
— Sandeep Swadia1:53

Swadia poses that question rhetorically at the outset, framing the entire discussion around a gap he sees as structural: AI models are trained to produce statistically probable text, which by definition trends toward the average. The result, he argues, is writing that is technically competent but lacks what he calls a "unique touch" - the quality that makes a reader trust a specific human voice over a generic one.

The Core Diagnosis: Predictability and Hallucination

Swadia identifies two distinct failure modes in AI writing. The first is stylistic: the output is generic because the model optimizes for broad acceptability rather than individual expression. The second is factual: AI systems produce incorrect information with apparent confidence, a problem Swadia treats as a separate and more dangerous issue.

“AI hallucinates, and it is so confident while it's doing it.”
— Sandeep Swadia3:27

That observation directly motivates the first operational step in his system: before doing anything else with an AI-assisted draft, the writer should instruct the model to extract every factual claim in the text and classify each one as verified, unverified, or disputed. The logic is that AI confidence is not a reliable signal of accuracy, so the writer needs to impose an external audit rather than assume the model's self-assurance means anything.

Swadia has examined the reliability gap in AI tools before. In his earlier analysis of critical thinking in the AI era, he used the Theranos collapse as a case study in what happens when confident-sounding claims go unchallenged, asking "What needs to be true for this to be real?" The fact-checking step in this system applies the same skeptical posture to AI output specifically.

The Five-Step System

Swadia's framework is built around the idea that AI should function as a tool the writer directs, not a ghostwriter who replaces the writer's judgment. The steps move from fact hygiene through voice recovery to adversarial stress-testing.

  1. 1Ask AI to extract every factual claim from your draft and mark each as verified, unverified, or disputed.
  2. 2Feed the AI your own insight or original perspective before asking it to generate anything - the model should be working from your thinking, not producing generic filler.
  3. 3Use AI to identify where the writing sounds generic and prompt it to suggest alternatives that reflect your established voice.
  4. 4Use AI as an opponent: give it your core argument and ask it to attack it, exposing weak reasoning before a reader does.
  5. 5Run a final pass asking AI to flag any sentence that could have been written by anyone, those are the sentences to rewrite by hand.

The adversarial step is the most operationally specific tip Swadia offers. Rather than using AI to validate an argument, he recommends treating it as a stress-test: "Use AI as your opponent by giving it your insight and asking it to attack it". The goal is to surface logical gaps before publication, not after.

Voice as the Non-Negotiable Output

The thread running through all five steps is Swadia's insistence that the writer's fingerprint must survive the AI drafting process. He frames this not as a stylistic preference but as a functional requirement for communication that actually lands.

“The key question here is does your writing, or any kind of communication, have your fingerprints, your unique touch?”
— Sandeep Swadia10:32

This position connects to a concern Swadia has raised across several topics. When he examined the risks of outsourcing thinking to AI tools, he asked directly, "If you outsource your thinking, what's left of you?" The writing system described here is, in practical terms, his answer: a process that keeps the human's reasoning at the center while using AI to handle the mechanical work of drafting, fact-flagging, and stress-testing.

Swadia has also argued, in his coverage of creativity and originality, that "creativity is not some magical force" and that the edge comes from "your originality and your taste", a framing consistent with his view that AI amplifies what a writer brings to it rather than substituting for it.

The practical implication of the system is that the writer's workload does not disappear; it shifts. Less time drafting from scratch, more time auditing claims, injecting perspective, and rewriting the sentences that AI flattened into generic prose. Whether that trade-off produces better writing depends entirely on how much original thinking the writer feeds into the process to begin with.