How can we detect AI?

Great to interview for BBC Berkshire and Oxfordshire today about how to detect AI online. Here are some tips.

🤖 Can you spot AI-generated content on social media?

Most people look for strange fingers, distorted backgrounds, or robotic writing.

But I think we're focusing on the wrong clues.

The real giveaway isn't usually a technical flaw.

It's the absence of lived experience.

Each platform has developed its own AI "accent":

💼 LinkedIn
Perfectly structured thought leadership.
Three lessons. Five takeaways. Seven secrets.
Lots of wisdom. Few actual stories.

🐦 X
Technically flawless threads.
Little humour. No eccentricity.
Content that sounds like Twitter rather than people talking on Twitter.

📸 Instagram
Beautiful images.
Inspirational captions.
Very little detail about where, when, how, or why.

🎥 TikTok
Polished delivery.
Consistent pacing.
Not much spontaneity.

📺 YouTube
Accurate summaries.
Efficient scripts.
Limited original reporting or personal insight.

What's often missing?

📍 Names
📍 Places
📍 Dates
📍 Mistakes
📍 Contradictions
📍 Unexpected details

Human communication is messy.

We forget things.
We change our minds.
We tell stories badly.
We include irrelevant details because they mattered to us.

AI content often optimises for clarity, engagement, and certainty.

Humans optimise for making sense of experience.

Ironically, as AI gets better, the strongest signal may become excessive smoothness.

The future of authenticity may not be proving that content is human.

It may be demonstrating that you've actually been there.

What are the signs that make you suspect a post was written by AI?