I can usually tell within two sentences. Something about the rhythm. The way every paragraph opens with a transition. The way the writing covers a topic without ever actually saying anything. Scroll through the comments on any LinkedIn post about leadership or strategy and you will see it everywhere. AI-generated replies that summarize the post back to the author and add nothing, because that is exactly what happened.
I am not the only one who notices. A study published at the Association for Computational Linguistics found that people who regularly use AI writing tools can identify AI-generated text with roughly 90 percent accuracy. Out of 300 articles reviewed, experienced annotators misclassified exactly one. They picked up on specific vocabulary patterns, an unnatural level of formality, and a kind of polished emptiness that is hard to describe but easy to recognize once you have seen enough of it.
Most people are not that tuned in. The same research found that the general population performs no better than a coin flip. They read something that sounds articulate, and they assume a person wrote it. That gap between what power users see and what everyone else sees is widening every month.
But here is the thing. Whether someone can tell it is AI is not actually the problem. The problem is what happens when the person who sent it does not know whether it is accurate.
I have watched this play out multiple times in the last few months. A marketing director sent a strategic recommendation to their leadership team. It read well. It was structured, confident, specific. But two of the three data points were wrong, and the framework it recommended did not apply to their industry. They could not tell because they did not know the subject well enough to check. They trusted the output because it sounded like it came from someone who did. Leadership caught it. That person no longer works there.
This is not a hypothetical risk. The legal profession has been learning this lesson publicly and painfully. Courts across the country have flagged more than 600 cases in which lawyers submitted AI-generated filings that cited cases that do not exist. A New York attorney was fined $5,000 after his brief included six fabricated case citations from ChatGPT. He told the judge he assumed it worked like a search engine. A federal court sanctioned an immigration attorney for submitting fabricated quotations in an emergency habeas case, despite knowing the tools hallucinate. The American Bar Association responded with Formal Opinion 512, clarifying that using AI does not reduce a lawyer’s ethical obligation to verify what they file.
The common thread in every one of these cases is not that AI wrote something bad. It is that a person put their name on something they did not verify.
That is the line. Not whether you use AI. Not whether someone can detect that you used it. The line is whether you can stand behind what you sent. Whether you know the claims are accurate. Whether you understand the reasoning well enough to defend it if someone pushes back.
Most of the anxiety around AI writing focuses on detection. Can my boss tell? Can my client tell? Will I get caught? That is the wrong question entirely. The right question is: do I actually know if this is true?
The business owners I talk to are not worried about AI replacing them. They are worried about what happens when their people use AI carelessly. Sending bad information to clients. Making commitments based on hallucinated data. Publishing content that sounds authoritative but falls apart under scrutiny. Those concerns are valid. I have seen people lose their jobs over exactly this in the last two months alone, not because they used AI, but because they used it as a substitute for knowing what they were talking about.
The fix is not to ban AI. It is to set a clear standard. If your name is on it, you own it. If you cannot verify a claim, do not include it. If you do not understand the subject well enough to catch an error, get someone who does to review it before it leaves your desk. The tool is not the problem. The problem is treating AI output like it has already been checked when nobody has checked it at all.
Do not put your name on something you cannot defend. That was true before AI, and it is the only rule that matters now. It is that simple.
Tyler Kelley is Co-Founder of SLAM Agency. He advises CEOs on strategic positioning and helps organizations understand where AI creates value and where it creates risk.