Last week I left a comment on a LinkedIn post promoting a new AI Visibility Tracker tool from LucidEngine. I'd checked the tool, noticed their llms.txt file was little more than a sitemap, and suggested they look at the AI Discovery File Specifications for guidance on how to structure it properly.
The response? "Sounds a bit AI-ish com knowing my website."
I created and maintain those specifications. I literally wrote the standard she was failing to implement properly. But because my comment was well-structured and offered specific technical detail, it apparently read as automated. So it was dismissed.
This isn't a rant about one LinkedIn exchange. It's about a growing pattern that should concern anyone who writes for a living, runs a business online, or cares about the quality of information on the internet.
The Competence Penalty
There's a new informal rule forming across social media, comment sections, and professional forums: if you write too clearly, too articulately, or with too much structure, people assume you didn't write it at all.
I've been writing like this for 30 years. Before ChatGPT existed. Before GPT-3. Before most people had heard the word "neural network." My writing style hasn't changed. What's changed is how people receive it.
Structured arguments and specific detail used to signal expertise. Clean paragraphs meant you'd taken the time to edit. Now all of it signals "probably a bot." We've inverted the meaning. Competence has become suspicious.
And here's what makes it particularly absurd: the people doing the dismissing rarely evaluate the actual content. They don't check whether the advice is correct, whether the links are useful, or whether the commenter knows what they're talking about. They just run a vibe check on the prose style and move on. The substance is irrelevant. Only the packaging matters.
AI Detection Tools Don't Work. Everyone Pretending Otherwise Is Selling Something.
The informal "sounds like AI" judgement is bad enough. But the formal tools people reach for aren't much better.
A 2023 study by James Zou and a team at Stanford found that AI detection tools flagged 61.22% of essays written by non-native English speakers as AI-generated. The actual AI-generated essays? The tools correctly identified about 97% of those. So they're good at catching AI. They're terrible at not catching humans.
"GPT detectors are biased against non-native English writers."
James Zou et al., Stanford University, Stanford HAI, 2023
When I first saw that number, 61%, I thought it was a typo. It wasn't. Think about what this means for businesses in the UK. British English already confuses these tools because they're overwhelmingly trained on American English. Add any degree of formality, technical vocabulary, or structured reasoning, and the false positive rate climbs even further.
The false positive rates across commercial AI detection tools range from 15% to 45%, depending on the type of text being analysed. Academic writing gets hit hardest. So does technical documentation. So does anything that's been carefully edited. In other words, the better your writing, the more likely it is to be flagged.
That's not a bug in the detection tools. It's a fundamental flaw in the entire premise. These tools work by measuring "perplexity" (how unpredictable the text is). Well-edited, structured, formal writing is predictable by definition. It follows conventions. It uses standard vocabulary. So of course it looks like AI to an algorithm trained to equate predictability with machine authorship.
The Real Casualty Is Value
Go back to that LinkedIn exchange for a moment. My comment pointed out specific problems with how LucidEngine had implemented their llms.txt file and directed them to the authoritative specifications that would help them fix it. It was free, expert-level feedback from the person who created those standards.
None of that mattered. The response didn't address the technical points. It didn't ask for clarification. It didn't check whether the specifications I linked to were real (they are, and they're used by businesses worldwide). It went straight to: "this sounds like AI, so I can ignore it."
This is a pattern we're seeing across the web. People have built a mental shortcut: well-written equals AI, and AI equals worthless. Two assumptions, both wrong, stacked on top of each other. The first collapses under any scrutiny (humans have been writing clearly for centuries). The second is just lazy thinking (AI-generated content can be accurate, useful, and valuable).
Whether something was written with AI assistance, or entirely by hand, or dictated into a phone while walking the dog, the only thing that should matter is whether the content is correct, useful, and trustworthy. That's it. The method of production tells you nothing about the quality of the product.
People Are Changing How They Write. That's Terrifying.
Here's where this gets properly alarming. People are deliberately writing worse.
Professionals who've always used em dashes have stopped using them because AI tends to use them too. Writers are avoiding words like "delve" and "tapestry" and "nuanced" because those words appear on lists of "AI tells." People are introducing intentional typos, using simpler sentence structures, and breaking their natural voice to pass an informal authenticity test that nobody asked for and nobody agreed to.
A LinkedIn thread I wrote about last month revealed just how deep this runs. Copywriters admitted they'd changed their vocabulary. Content managers said they now avoid structured lists because "AI always makes lists." One writer said she'd stopped using semicolons entirely.
We are regressing. The entire history of writing education has been about teaching people to communicate more clearly, more precisely, more effectively. And now there's a social pressure pushing in the opposite direction: be less clear, less precise, less effective, because if you write too well, you'll be accused of cheating.
This is the opposite of progress. It's a tax on competence.
Google's Position Matters More Than LinkedIn's Opinion
While the comments section debates whether your blog post "feels AI," Google has taken a very different position. And since Google still drives the majority of web traffic to UK businesses, their stance is the one that actually counts.
"Our guidance for AI-generated content is the same as it's always been: focus on the quality of content, rather than how content is produced."
Google Search Central, Google Search Central Blog
That's not ambiguous. Google doesn't care whether AI helped write your content. They care whether it's useful, accurate, and serves the person who searched for it. Their February 2026 Discover core update doubled down on this by rewarding content quality and topic expertise, regardless of how that content was produced.
So we have a strange situation. The world's largest search engine, the one that sends most businesses their traffic, says the method of production doesn't matter. But a stranger on LinkedIn says it does. And somehow it's the stranger's opinion that's winning.
What Should Actually Matter
I run a web design and AI visibility business. I use AI tools every single day. I also write by hand every single day. Sometimes the two overlap. Sometimes they don't. And I'm not going to apologise for any of it, because the output is what matters.
When I left that comment about LucidEngine's llms.txt implementation, every technical claim was accurate. The specifications I linked to are real, public, and maintained. The advice was actionable and free. Whether I typed every word from scratch or had AI help me structure the response changes precisely nothing about the value of what I said.
We need to get past this. The question shouldn't be "did AI write this?" The question should be: is it true? Is it useful? Does it help? And if you're dismissing useful information because the tone doesn't meet some arbitrary standard of human imperfection, you're not protecting quality. You're punishing it.
The impulse to trust imperfection is understandable. In a world flooded with AI-generated noise, looking for human fingerprints makes emotional sense. But when those fingerprints become the only thing you evaluate, you've stopped assessing quality entirely. You're just profiling style.
And that, to borrow from the original LinkedIn exchange, is something worth knowing about.
If you're curious about how AI systems actually evaluate your website (and how AI discovery files help them understand your business better), the AI Visibility Checker will show you exactly where you stand. It's free and it takes about 30 seconds. Unlike the "is it AI?" question, this one actually has a useful answer.
Frequently Asked Questions
Are AI detection tools accurate?
No. Stanford research found false positive rates of 61% for non-native English writers. Commercial tools show rates between 15% and 45%. Well-edited, formal writing is disproportionately flagged because detection tools equate predictable text with machine authorship.
Does Google penalise AI-generated content?
No. Google's official position is that content quality matters, not how it was produced. Their February 2026 Discover update rewards topic expertise and useful content regardless of whether AI was involved in creating it.
Why do people assume well-written content is AI?
AI language models produce structured, formal text by default. People have started associating those traits with AI rather than with skilled human writers. Clean prose, logical structure, and specific vocabulary have become "tells" even though they predate AI by centuries.
Should I disclose when I've used AI to help write content?
For business content, no. Google doesn't require it and there's no legal obligation in the UK. Academic work is different and usually requires disclosure. Focus on whether your content is accurate and useful rather than worrying about labels.
Does British English get flagged more often by AI detectors?
Yes. Most AI detection tools are trained on American English datasets. British spelling, grammar patterns, and formal conventions register as unusual, which increases false positive rates for UK writers and businesses.
Should I change my writing style to avoid AI suspicion?
No. Deliberately writing worse to appear more "human" degrades your content quality and your brand's credibility. Write clearly, be accurate, and focus on providing real value. That's what Google rewards and what your readers actually need.
What is the "competence penalty" in AI bias?
It's the emerging pattern where writing skill works against you. People who communicate clearly and professionally face higher suspicion of AI use than those who write casually or with errors. The better your writing, the more likely it is to be dismissed.
How should I judge whether online content is trustworthy?
Check the claims, not the style. Are sources cited? Can you verify the facts? Is the advice actionable and specific? Does the author have relevant expertise? These questions tell you far more about quality than whether the prose "feels" AI-generated.
Find Out How AI Actually Sees Your Business
Instead of guessing whether content "sounds AI," find out what AI systems actually know about your business. Our free tool scores your AI visibility in 30 seconds.
Check Your AI Visibility
