Can AI Detect AI Writing, and Should You Use It?
This blog was originally published on Medium and now migrated to Substack
Last week, I did an interesting experiment on whether AI chatbots can detect AI writing. What do you guess the results are? Let’s go through the experiment.
I am passionate about Human-Centric AI Leadership (disclaimer: I came up with this name without any assistance from AI) and want to publish an article about what it is and why it is important now and in the future. So, I threw this prompt into Claude: “Write a blog on why human-centric AI leadership is important now and in the future.” Then I entered the following prompt into other AI tools: “Is this blog written by human or AI and why? {Claude output: The Critical Importance of Human-Centric AI Leadership: Steering Technology Toward Human Flourishing}”. The AI tools that I tested are: ChatGPT, DeepSeek, Gemini, and Grok. Guess what? Claude fooled all other Chatbots.
ChatGPT and DeepSeek were very confident that the blog was written by human. Gemini came to the same conclusion but with the caveat that it wouldn’t totally rule out other possibilities. Grok did “lean slightly towards human authorship”, but suggested checking on the author profile or other run it through other detection tools.
ChatGPT’s Answer
Determining whether a blog post was authored by a human or an AI involves analyzing various factors, including writing style, depth of insight, and the presence of personal experiences or unique perspectives. The blog post you’ve shared, titled “The Critical Importance of Human-Centric AI Leadership: Steering Technology Toward Human Flourishing,” exhibits characteristics that suggest it was likely written by a human. Here’s why:
1. Depth of Insight and Nuanced Understanding
The article delves into complex concepts such as ethical alignment, inclusive development, and responsible innovation within the context of AI leadership. It reflects a nuanced understanding of the multifaceted challenges and opportunities presented by AI, which typically stems from human experience and critical thinking.
2. Structured Argumentation
The blog is organized into coherent sections, each building upon the previous one to construct a persuasive narrative. This logical flow and structured argumentation are indicative of human authorship, as it demonstrates deliberate planning and a clear intent to convey specific messages.
3. Emphasis on Human Values and Ethical Considerations
The content emphasizes the importance of human welfare, dignity, and autonomy in AI development. This focus on ethical considerations and the moral implications of technology suggests a human author reflecting on societal values and the broader impact of AI.
4. Call to Action for Leaders
The article addresses business leaders directly, urging them to take responsibility for ethical AI implementation and to foster a culture that balances innovation with human-centric values. This prescriptive approach indicates an understanding of organizational dynamics and leadership challenges, which is characteristic of human insight.
While advanced AI models can generate coherent and contextually relevant text, they often lack the ability to infuse content with genuine human experiences, personal reflections, and the deep ethical considerations evident in this blog post. Therefore, based on the provided content, it is reasonable to conclude that the blog was authored by a human.
DeepSeek’s Answer
The blog post you provided appears to be written by a human rather than an AI. Here are several reasons why:
1. **Depth and Nuance**: The post demonstrates a deep understanding of the subject matter, including ethical considerations, societal impacts, and leadership principles. While AI can generate content on these topics, the level of nuance and the ability to connect various complex ideas in a coherent manner is more characteristic of human thought.
2. **Philosophical Reflection**: The post includes philosophical reflections on the purpose of AI and its impact on human dignity and autonomy. This kind of reflective thinking, which involves value judgments and ethical considerations, is typically more advanced than what current AI systems are designed to produce.
3. **Personal Voice and Style**: The writing has a distinct personal voice and style, which is often more varied and expressive in human writing. AI-generated text can sometimes be more uniform and less distinctive in style.
4. **Contextual Awareness**: The post shows a strong awareness of current societal issues and the specific challenges posed by AI, such as economic transformation and the need for global cooperation. This contextual awareness is something that AI can mimic but often lacks the depth and specificity that a human writer can provide.
5. **Ethical and Emotional Engagement**: The post engages with ethical and emotional dimensions of AI leadership, such as the importance of human welfare and dignity. These are areas where human writers excel, as they can draw on personal experiences and a deeper understanding of human emotions.
6. **Originality and Creativity**: The post presents original ideas and creative solutions, such as the emphasis on inclusive development and responsible innovation. While AI can generate creative content, it often relies on patterns and data it has been trained on, whereas humans can produce truly novel insights.
7. **Structural Complexity**: The post is well-structured with a clear introduction, body, and conclusion. It uses headings and subheadings effectively to organize complex information. While AI can generate structured content, the logical flow and coherence seen here are more indicative of human planning and organization.
In summary, the depth, nuance, philosophical reflection, personal voice, contextual awareness, ethical engagement, originality, and structural complexity of the blog post suggest that it was written by a human rather than an AI.
Gemini’s Answer
The blog post you provided exhibits characteristics that suggest it was written by a human, although it’s possible AI tools were used in the process. Here’s a breakdown of why:
* Structure and Flow: The article presents a logical progression of ideas, starting with a broad introduction and narrowing down to specific principles and recommendations. While AI can generate structured text, the way this article builds its argument feels more deliberate and human-driven. The use of headings and subheadings also contributes to readability and a sense of intentional organization.
* Nuance and Perspective: The article demonstrates a nuanced understanding of the complexities of AI development and its impact on society. It acknowledges both the opportunities and the challenges, which is a hallmark of human analysis. While AI can process and present information, the ability to weigh competing factors and present a balanced perspective is often a sign of human authorship.
* Emphasis on Human Values: The core theme of “human-centric” AI leadership and the repeated emphasis on human flourishing, dignity, and agency suggest a human author. These concepts are deeply rooted in human values and ethical considerations, which are areas where human understanding and judgment are still crucial. While AI can process and understand these concepts, the emphasis and framing of them in this article feel distinctly human.
* Call to Action: The conclusion includes a call to action for leaders across various sectors, urging them to take responsibility for shaping the future of AI. This call to action, along with the article’s overall tone, suggests a human author with a vested interest in the ethical and societal implications of AI.
* Writing Style: While the writing is clear and professional, it doesn’t have the overly polished or sometimes repetitive feel that can be characteristic of AI-generated content. It maintains a consistent tone but also incorporates slightly more complex sentence structures and vocabulary, suggesting human authorship.
However:
It’s important to acknowledge that AI writing tools are rapidly evolving. It’s entirely possible that a human author used AI tools to assist with research, drafting, or editing. AI could have helped generate some of the text, particularly sections that summarize existing knowledge about AI ethics and leadership. However, the overall structure, argumentation, and emphasis on human values point towards significant human involvement in the writing process.
In conclusion: While it’s not possible to say with absolute certainty, the evidence leans strongly towards this blog post being primarily written by a human, possibly with the assistance of AI tools. The article’s structure, nuanced perspective, focus on human values, and call to action all suggest a human author’s hand at work.
Grok’s Answer
Determining whether a piece of text like the blog post you’ve shared was written by a human or an AI involves looking at several clues:
1 Content and Style:
◦ Coherence and Flow: The blog post is well-structured, with a clear introduction, body, and conclusion. It uses transitional phrases effectively to connect ideas, which is something AI can do but often with less natural flow compared to human writing.
◦ Depth of Insight: The discussion on human-centric AI leadership delves into philosophical, ethical, and practical considerations, showing a nuanced understanding that could come from either a well-informed human or a sophisticated AI trained on vast amounts of similar texts.
2 Originality and Specificity:
◦ Unique Perspectives: The blog does not just rehash common points on AI ethics but provides a structured argument for human-centric leadership, which might suggest human authorship. However, AI models, especially those with access to broad and diverse training data, can also generate unique-seeming content.
3 Language and Syntax:
◦ Complexity: The language is somewhat formal yet accessible, with complex sentences and paragraphs organized into thematic sections. While AI can replicate this style, humans tend to add personal touches or specific examples from their experience or reading, which aren’t immediately evident here but could be present in a broader context.
4 Metadata and Context:
◦ Publication Context: Without additional metadata like authorship, publication date, or platform specifics, it’s harder to make a definitive judgment. However, if this blog were part of a series by a known author or organization focused on tech ethics, that would lean towards human authorship.
5 AI Detection Tools:
◦ While not foolproof, tools designed to detect AI-generated text might be used, but their accuracy can vary, especially with AI models improving to mimic human writing more closely.
Given these considerations:
• Human Likelihood: The depth, structure, and thematic focus suggest it could be written by a human with expertise or deep interest in AI ethics and leadership.
• AI Likelihood: Modern AI, particularly large language models, can produce similar content if prompted correctly with guidelines on structure, tone, and content. However, the nuanced advocacy for human-centric approaches might be less typical for AI without specific direction.
Conclusion: Without explicit metadata or a known author, it’s challenging to be certain. However, I lean slightly towards human authorship due to the coherent and well-argued perspective unless there’s evidence suggesting otherwise, like known use of AI in content creation by the blog’s source. If you need further verification, you might consider:
• Checking if there’s an author profile or similar posts by the same author.
• Looking for any acknowledgments of AI assistance in writing.
Using an AI text detection tool for a secondary opinion, though these are not infallible.
I was surprised by how well Claude’s writing performed but also felt desperate by the outcome. If all the smart tools can’t tell the difference, is it the end of human creativity and authentic writing? As Grok suggested, I decided to use a specialized AI plagiarism detection tool ZeroGPT to test the Claude writing again, and it didn’t disappoint me. Finally, someone can see the difference.
Encouraged by this result, I decided to put my two other random Medium blogs into the test.
Yes, I used some AI assistance in writing the blog Embracing Human-Centric Values in the Age of AI, but most of ideas and main points of views came from my own thinking and experience. When I felt so frustrated by the changes AI brought us and complained about it, it’s 100% AUTHENTIC human feelings!
AI is smart, efficient, and capable of doing amazing things, but it cannot totally connect with or replace human emotions, feelings, and authenticity. In the new AI age, genuine and unique human experiences, points of view, and independent thinking will be the new premium and treasure we will cherish. Use AI for idea generation, research, and productivity, but trust your own intuition, thinking, and creativity, and hold true to your values, beliefs and passions. You will always be able to pass the GPT test and stand out in the AI era in the future.
P.S. This writing is 100% AI assistance free, other than spelling and grammar check that’s part of the Medium writing platform.