Stop trying to detect AI-written content. Start spotting generic content instead.
The last couple of years have seen everyone under the sun asking whether AI-written content can be detected. In fact, I’d say it was, and perhaps still is, one of the biggest questions in marketing.
I'm just not sure it's the right one we should be asking.
Large Language Models (LLMs) have improved dramatically in a very short time. Human writers increasingly use AI as part of their workflow. Editors are rewriting AI drafts as standard. And AI is rewriting human drafts as standard.
The lines have blurred to the point where confidently declaring, "Yes, this was written by AI" has become increasingly difficult.
If we're being honest, it’s also become increasingly irrelevant. Because the question I find myself asking instead is much simpler:
“Would anyone care if this content disappeared tomorrow?”
Because what frustrates me at the moment is the amount of generic content out there… not the amount of ‘AI-generated’ content.
Generic content existed long before AI
Let's not pretend bland content arrived with ChatGPT; marketing teams have been producing safe, predictable, keyword-filled articles for decades.
AI has simply made it much quicker to create them.
We're now surrounded by content that explains the obvious, repeats what dozens of other articles have already said, and leaves readers no better informed than when they arrived.
Whether it came from a human or an AI model almost doesn't matter; the outcome is the same.
Why AI content detectors aren't always reliable
It's tempting to believe there's a tool that can definitively tell us whether content was written by AI, but unfortunately, that's not how it works.
Detection tools can be useful at picking up GenAI ‘signals’, but they shouldn't be treated as proof.
Like most tests, false positives happen. Human-written work can be flagged as AI-generated. And AI-assisted work that's been heavily edited may not be detected at all.
So for marketers and businesses, that means focusing less on where the words came from and more on whether they're worth reading.
What about the classic AI ‘tells’?
Every few months, a new list of supposed AI giveaways starts doing the rounds on places like LinkedIn.
Em dashes
Bullet-heavy formatting
Perfect grammar
The use of certain words or phrases, like, “It’s not X, it’s Y…”
Repetitive sentence structures
Overly enthusiastic conclusions
The problem is that none of these prove anything.
I know plenty of experienced writers who use em dashes (I'm one of them 🙋🏻♀️).
Others deliberately avoid contractions. Some editors simply insist on short paragraphs and frequent subheadings because they're easier to read online.
Equally, I've seen AI-generated articles that avoid every one of these so-called ‘red flags’ because they've been properly edited.
Looking for individual stylistic quirks is a bit like trying to identify a chef from the type of plate they use. Sometimes you're right… but often you're not.
That's why I think we're asking the wrong question.
Rather than trying to spot whether AI was involved, it's far more useful to ask whether the content is original, insightful and genuinely helpful.
Related:Content marketing and AI in 2026: 40+ stats for marketers
An editorial framework I keep coming back to…
After reviewing hundreds of articles over the past few years, I realised I was asking myself the same questions every time. Eventually, those questions became what I now think of as The Generic Content Test.
It's worth stressing that this isn't a scientific test or a way of proving whether AI was involved; it's simply an editorial framework I use to assess whether a piece of content is distinctive, useful, and worth publishing.
The Writeful Generic Content Test
Before you hit publish, ask yourself:
Originality: Have I added something genuinely new?
Opinion: Have I actually taken a position?
Experience: Could only someone who's done this have written it?
Specificity: Have I replaced vague claims with concrete examples?
Editing: Does every paragraph earn its place?
Differentiation: Could a competitor publish this unchanged?
Value: Would someone bookmark or share it?
AI visibility: Could an AI assistant confidently quote this as one of the best answers available?
The more "no" answers you have, the more likely it is that your content will feel generic to readers. And that may not be because AI wrote it, it’s because nobody’s improved it!
I find these questions far more useful than trying to guess whether AI was involved.
Why these questions matter:
Originality
Many articles explain. Far fewer teach.
If someone could skim your article and predict every paragraph before they reach it, you've probably added very little value.
Opinion
The best content doesn't sit on the fence. Instead, it makes an argument, challenges assumptions, and acknowledges trade-offs.
Generic content often avoids committing to anything because it's trying to appeal to everyone. Ironically, that makes it memorable to no one.
Experience
Readers can usually tell when someone has actually done the work they're describing, and that doesn't always mean telling anecdotes.
It might mean explaining:
a common client misconception
a mistake you've seen repeatedly
why conventional advice doesn't always work
a practical example from a real project
These details are difficult to invent because they come from experience, not research.
Coincidentally, these are many of the same qualities that search engines and AI-powered answer engines increasingly reward: genuine experience, original insights and evidence that a real expert has contributed to the content.
Specificity
Generic writing hides behind vague claims.
"Powerful…" "Seamless…" "Transformative..." "Game-changing..."
Specific writing tells readers exactly what changed, who benefited, and why it mattered.
In a nutshell: Specificity builds credibility.
Editing
One habit I see in generic content is that you can see the same thing being said in several different ways.
Good editing is often about ‘removing’. If a paragraph doesn't introduce a new idea, strengthen an argument or move the reader forward, it's probably there because nobody challenged it.
Differentiation
This is one of my favourite aspects to test. Replace your company name with a competitor's.
If nothing else needs changing, the article probably lacks originality.
Your experience, examples and perspective should make it unmistakably yours.
Value
This might be the simplest test of all.
When someone finishes reading, have you given them something they'll want to come back to, like:
A useful framework
A memorable insight
A practical checklist
An example they can apply
If the answer is “no”, it's worth asking whether you've really contributed anything new.
Related: AI isn't coming for your creative words
AI visibility
As more people discover information through AI assistants and answer engines, another question is becoming increasingly important.
If every point in your article already exists in hundreds of other places online, why would an AI system choose yours?
What these engines are looking for are original frameworks, real-world experiences, strong opinions, and practical examples.
These things make content more valuable; not just to your readers, but to the systems increasingly deciding what gets surfaced online.
The same principle applies to gated content. If your best insights sit behind a lead form, AI assistants and answer engines can't easily access or cite them. That's why I believe it's important to think carefully about when content should be open and when it genuinely deserves to be gated.
The irony is that good AI-assisted writing often doesn't sound like AI at all
One of the biggest misconceptions I still hear is that AI inevitably produces ‘robotic content.’
In reality, the quality depends far more on the person using it. Experienced writers don't publish their first draft. They challenge it, remove clichés, add expertise, inject examples.
They’ll question weak arguments and rewrite sections that feel too generic.
AI can speed up parts of the process, but it doesn't replace human judgement.
That's still the job of the writer.
The future belongs to editors, not just writers
As AI becomes more capable, I think one skill will become increasingly valuable:
Not proofreading… editing.
And that means:
Knowing what to remove. What to question. What needs evidence. Where a stronger opinion belongs… and where an example would make an idea more convincing.
The brands producing the best content are not necessarily the ones using the least AI; they're the ones refusing to publish generic work, and they edit well.
The question worth asking
So the next time you read an article that feels a little... flat... don't be so quick to ask whether AI wrote it. Ask whether it actually taught you something. Whether it made you think. Whether it offered a perspective you couldn't have found elsewhere.
Readers don't care who wrote your content.
They care whether it’s worth reading.
=====
FAQs: Detecting AI content
-
Not with complete certainty. AI detection tools can provide useful signals, but they shouldn't be treated as proof.
Human-written content can be incorrectly flagged as AI-generated, while heavily edited AI-assisted content may not be detected at all.
It's generally more useful to evaluate a piece of content on its originality, usefulness and quality than to focus solely on how it was created.
-
In many cases, it's not AI itself but a lack of editing. Content often feels AI-generated when it relies on vague language, repeats familiar ideas, avoids taking a clear position or lacks real-world examples.
Those same characteristics can appear in purely human-written content too.
-
The biggest difference comes after the first draft. Add your own experience, opinions, examples and practical insights, then edit critically.
Remove clichés, replace generic statements with specifics and make sure every section contributes something genuinely useful that readers won't find elsewhere.
-
The Writeful Generic Content Test is an editorial framework designed to help assess whether content is distinctive, useful and worth publishing.
It isn't a scientific method for detecting AI use. Instead, it encourages writers and editors to consider factors such as originality, opinion, experience, specificity, differentiation and reader value before publishing.
-
As search engines and AI assistants prioritise helpful, original content, articles that simply repeat existing information are less likely to stand out.
Content that demonstrates expertise, offers unique perspectives and provides genuine value has a better chance of earning visibility, citations and engagement.
=====
Now that you’ve read this, you may be interested in
