I have debated whether to write this post for a while now (I have shared a few thoughts on LinkedIn recently), but I decided it was worth saying out loud. This is not a complaint. This is a conversation I think the business community needs to have, because what is happening on the writer’s side of the fence deserves more attention.
The subject? AI content detectors and the false positives that are quietly damaging relationships between writers and their clients.
What Happened
Over the past two weeks, two separate clients came back to me after I submitted articles, telling me their detectors flagged the content as “probable” or “possible” AI.
To be clear, both articles were 100% written by me. Not a contractor. Not a mentee. Me. And I have been writing professionally for more than 30 years, long before AI tools were part of any conversation.
What made the situation even more frustrating was that before submitting both articles, I ran the work through two, sometimes three AI detectors myself. I pay for a subscription to do this because I take accuracy seriously. In both cases, my results came back at 0% assumed AI, across multiple tools. I felt confident submitting the work.
Then I received responses telling me the client’s detector said otherwise.
What the Detectors Actually Said
No detailed reports were included. No comments about the quality of the work, or the fact that it was delivered ahead of deadline, or that the word count exceeded the request at no additional charge. Just screenshots and summaries of what the detector “thought.”
One client’s tool flagged what it called “parallel phrasing.” I had to look that up. It also noted “slightly awkward phrasing,” which, fair enough, because a human wrote it. And then it questioned a statistic in the article, suggesting AI often inserts plausible-looking data without sourcing. Except I had hyperlinked that statistic to a source with a domain rating of 93. That is the kind of citation that SEO rewards and that a detector simply cannot see.
The second client ran the article through ChatGPT and Grok to evaluate whether it was AI-generated. The ChatGPT assessment noted “predictable structure” (it was a blog), “repetitive phrasing and filler transitions” (which described my typical writing style, though I could find no evidence of repetition when I reread the article yet again), and broad claims without strong sourcing (despite the fact that every statistic was hyperlinked to PubMed, which I also referenced by name within the article itself).
Grok flagged “perfect, formulaic structure with clean H2-style headings, short paragraphs, and a classic blog post flow.” Which is, well, that’s how you write a blog. That has been how you write a blog for years. Proper structure is a sign that the writer knows what she is doing. It is not evidence that a machine did the work.
What the Research Actually Says
A study published on PubMed Central found that AI detectors are far from reliable, with accuracy that varies considerably based on text length and how much editing the content has undergone. While some tools report over 90% accuracy on raw AI-generated output, they often struggle significantly with human-written, edited, or produced content, particularly when it features consistent stylistic patterns.
In other words, these tools can sometimes detect AI. They cannot always confirm that a human wrote something. That is a meaningful limitation that clients need to understand before acting on detector results.
What I Want Clients to Know
When you hire a writer, evaluate their credentials. In my case, I hold an undergraduate degree in English and a master’s degree in business education. I am a native English speaker. I have been writing professionally for over 30 years. I send prospective clients access to more than 200 published, bylined samples.
That track record matters. It is the context a detector cannot provide.
Before sending a message that questions a writer’s integrity based on a screenshot, I would ask you to pause and consider a few things.
- Read the work. Does it read well? Does it accomplish what you asked for? Does it sound like AI, or does it sound like a professional who knows your industry?
- Think about what you are communicating. When you send a writer a message focused entirely on what a detector said, without acknowledging the quality of the work or asking a single question, you are sending a message about trust. That matters for the working relationship.
- Understand the technology’s limits. AI detectors are tools, not arbiters of truth. They produce false positives. They cannot read hyperlinks. They cannot evaluate context, sourcing, or whether a writer’s consistent style simply looks a certain way across many pieces of work.
A Better Approach to Hiring and Working With Writers
If AI use is a concern for your business, the answer is not to run every submission through a detector and treat the results as conclusive. The answer is to hire thoughtfully.
Look for writers with documented experience. Ask for samples. Have a direct conversation about your expectations around AI. Do not hire based on the lowest price, then rely on a tool for quality control. Read what was submitted and make your own judgment.
The detectors will continue to generate false positives. Experienced human writers will keep getting flagged. And if the response is always suspicion rather than dialogue, the professional relationships that produce the best content will erode.
That is not a result that benefits anyone. Do you want high-quality human-written content for your website? Schedule a blog-scoping call today.