I checked the prologue with 3 different AI text detectors: zerogpt, grammarly, and quillbot. The results were different. 33%, 4%, and 0% similar to AI text.
Critical note:
I copied the prologue text until it filled the AI text generator dashboard and the amount of text that can be checked in one session can vary, such as zerogpt, which allows up to 15,000 characters to be checked.
This means that if the chapter text cannot be checked completely, the checking session must be done twice. And this is unfortunate. The AI text detector can give wrong results if, when we copy the checked text, it 'happens' to be similar to the AI text generated, so that the results can be up to 80-100% similar to AI text.
I personally do not trust AI text detectors completely, even the AI text detector claims the results are indeed inaccurate, there can be true negatives (text detected by AI when it is actually hand-typed, but the style is similar to AI text) or conversely false positives (text detected as not AI when it is actually AI-generated text).
Therefore, instead of using an AI text detector to detect whether text is true or not.
It might be more accurate to analyze the story's moral, coherence, and correspondence.
First, the moral of the story is what moral message the author wants to convey to the reader. Generally, AI struggles to bring stories to life and have a moral. Because to create a living story, the author usually explores his or her life experiences, trauma, values, love, and hopes. This involves a creative process. No matter how good an AI model is (currently), it lacks the creativity to create something from imagination and inner experience. What AI produces is more like patterned or random text depending on the instructions given, and it is generated from billions of datasets.
Second, we can examine the coherence between sentences, paragraphs, scenes, and chapters. Do these texts have narrative integrity in forming a discourse that illustrates the story's moral premise? AI-generated text can't struggle to form connections between sentences, paragraphs, scenes, and chapters. Why? Because AI-generated text is usually random and instructed per session, unless there is a super LLM model capable of generating a complete novel containing 50,000 words in a single session, discourse coherence is possible. But free AI doesn't offer this option.
Finally, we can examine correspondence. Novels created by human thought usually have correspondence with the real world, whether the symbols used are allegories of real-world realities. Certain scenes can reflect the author's experiences. Even the novel itself can be an allegory of the dark realities of the real world.
Ultimately, a true author will be able to answer these three questions: what is the moral of the novel? How coherent is it? And how does the novel correspond to the real world? If they wrote with their own thoughts, I'm sure they would be able to answer these questions. Conversely, if it was purely AI-generated text, they wouldn't be able to answer these questions.
It's like someone copying from a reference book (cheating) and someone answering with their own thoughts. Someone who didn't copy (even if the answer is absurd) will be able to explain their answer. Meanwhile, someone who cheats (even if the answer is correct) will find it difficult to explain their answer. Even if they are forced to answer, they will say, "because the reference book says so."
