LLM systems and video, audio and image processing “AI” is getting more and more refined. Are we soon going to be unable to tell what is human, made by humans? Is anyone going to write or act in front of cameras anymore? A short note along these lines.
For a more nuanced and detailed look at this, see:
“My LLM/AI experiement – and why I shut it down“.
1. A text I came across that looked like it’s AI-written
I will use a short example, but more and more often I see whole, long articles that look like they were made by LLM (“AI”).
I have redacted the name(s) to respect the (alleged) author’s privacy.
Cyberattacks happen daily. Will your website survive the next one? In The Imperative of Digital Security, Name_Redacted distills over 30 years of military, enterprise, and hosting experience into a clear, actionable guide for developers, IT managers, and business owners.
You’ll learn:
How hackers exploit weak websites—and how to stop them
What every database, CMS, and server admin must secure
Why phishing, malware, and brute-force attacks still work
How to build layered defenses, backups, and incident response
The exact tools and steps to harden your hosting environment
Whether you’re running a small blog, e-commerce store, or enterprise network, this book gives you the practical blueprint to protect your digital assets in an increasingly hostile internet.
Protect your users. Defend your reputation. Secure your future.
For some reason, I was convinced that this was written by AI. Can’t put my finger on why – it just looked like it was! I asked ChatGPT and DeepSeek: copy-pasted the text and asked them: “Is this ai-written?”
I will copy/paste the responses as they wrote them.
2. ChatGPT’s response
That passage could plausibly be written by either a human or AI. It shows signs of polished marketing copy, but nothing conclusively AI-specific. Here’s a breakdown:
Indicators of possible AI authorship:
- Buzzword density: Phrases like “layered defenses,” “digital assets,” “hosting environment” are common in LLM-generated tech marketing.
- Polished rhythm and structure: Sentences are tight, list-heavy, and use repetition for emphasis (e.g., “Protect your users. Defend your reputation. Secure your future.”)—a hallmark of LLMs trained on corporate writing.
- Generic tone: It lacks any distinct personal voice or storytelling, which often points to templated or AI-written prose.
Indicators of possible human authorship:
- Specificity in credentials: Referring to “Name_Redacted” and “over 30 years of military, enterprise, and hosting experience” is often done more thoughtfully by humans.
- Targeted, plausible structure: The layout mirrors real marketing blurbs for tech/security books—possibly pulled from a back cover or sales page.
Verdict:
If you told me this was from an AI prompt like “Write a book promo for a cybersecurity guide in the style of a back cover blurb,” I’d believe it.
If you told me a human copywriter wrote it for Amazon—also believable.Most likely case? Human-written, possibly edited or polished by AI (or written by AI then lightly edited by a human).
Would you like a more “humanised” rewrite or to check if it ranks well against AI detectors?
3. DeepSeek’s response
Yes, this appears to be AI-generated (or at least heavily AI-assisted). Here’s why:
1. Overly Generic, “Buzzword-Heavy” Language
- Phrases like “distills over 30 years of experience,” “clear, actionable guide,” and “practical blueprint” are common in AI-generated marketing copy.
- The tone is polished but lacks a distinct human voice—it reads like a templated sales pitch.
2. Formulaic Structure
- The opening question (“Will your website survive the next one?”) is a classic hook used in AI-generated content.
- The bullet points follow a predictable pattern (“How…, What…, Why…,”) without deeper insights.
3. Vague Claims Without Specifics
- No mention of real-world case studies, unique methodologies, or verifiable credentials for “Name_Redacted.”
- The promise of “exact tools and steps” is vague—human-written security books usually name-drop tools (e.g., “using Fail2Ban, CrowdSec, and WAF rules”).
4. Unverifiable Authority
- “Name_Redacted” doesn’t appear to be a known cybersecurity expert (no Google results, LinkedIn, or published works).
- “30 years of military, enterprise, and hosting experience” is a broad claim with no supporting evidence.
5. AI-Generated Call-to-Action
- The closing line (“Protect your users. Defend your reputation. Secure your future.”) is a triplet commonly used by AI for dramatic effect.
Is This a Real Book?
- If this is promoting an actual book, it’s likely either:
- AI-generated marketing copy (common for self-published eBooks).
- Fake/placeholder content (some scam sites auto-generate “book ads” to collect clicks).
How to Check?
- Search for “The Imperative of Digital Security Name_Redacted”—if no real book exists, it’s AI fiction.
- Look for reviews, ISBN, or a publisher. AI-generated books often lack these.
Final Verdict:
This reads like AI-generated book promotion—possibly for a low-content or nonexistent book. If you’re looking for real cybersecurity guides, stick to authors like Brian Krebs, Bruce Schneier, or Michal Zalewski.
Would you like help finding legitimate resources on web security?
Conclusion
It appears that “robots” (the LLMs I asked) are better at putting a finger on what looked to be off, i.e. what to me smelled like AI, but I couldn’t explain why.
With longer and more complex texts, this is often even less obvious, especially when they are heavily edited.
Likewise, in all this craziness and flood of AI-written stuff, I’ve witnessed first hand when something my friend and I wrote from scratch was suspected of being AI-written:
When Human Replies Look Like AI – Real YouTube Exchange
I’m still not sure what to make of this, but I am documenting it for future revision.
P.S.
Fun fact – I asked ChatGPT to read this very article. It suggested I rephrase this:
“Can’t put my finger on why – it just looked like it was!”
Into:
“Can’t put my finger on why – it just felt like it!”
Last updated:
Originally published:
