I run a WordPress website where I publish articles on topics like lifestyle, home improvement, gardening, and practical tips. I write and edit all of my content carefully, but lately I’ve been thinking more about AI detection as AI-generated content becomes more common.
I’ve been testing a few AI detectors before publishing, but I’m noticing that the results can vary quite a bit. The same article might receive very different scores depending on which platform I use, and that makes it difficult to know which report is actually the most reliable.
I understand that no AI detector is perfect, and I’m not expecting 100% accuracy. I simply want a tool that provides consistent results and helpful reports so I can review my content with more confidence before publishing.
For those who manage websites or publish content regularly, what has your experience been with AI detectors? Which ones have you found to be the most consistent for blog posts and long-form articles? I’d appreciate any recommendations or insights.
Personally, I do not think completely consistent results will be possible. Different models can lead to different assessments, and the results may change again as newer models become available.
For several years, I have used one simple prompt multiple times a day:
Polish this text
I place the text immediately after the prompt. As the models have improved, I have noticed that the results have become progressively closer to what I would expect from a capable human editor.
At this point, are you primarily trying to determine whether content was written by a human or generated with AI, or would it be more useful to evaluate the content itself for accuracy, clarity, and quality?
Personally, I am not opposed to posts created with the assistance of AI. What I dislike, as a reader, are long walls of text and replies that appear not to have been reviewed or fact-checked before being posted.
Welcome to modern dilemmas 101: How to detect AI content as it gets better. That is actually hard to do currently, for text at least, could always use anti-plagiarism tools found in academic circles, but those can be expensive and not 100% reliable. For images mayor vendors are starting to add mechanisms like SynthID digital watermarks. Best for now is to eep yourself informed and check new trends on that matter.
As for Wordpress side, is it an option to not get your site indexed?
There is a trick I do to test if folks are copy my post paste replies. See y’all I talk southern in real life.. folks , yonder etc. and once introduced into a response the person I’m chatting with if using AI it becomes southern.
So like poisoning (for the lack of a better word) the blog for AI so that it starts posting incoherent entries? A car blog and suddenly the AI start talking about the crunchiness of muffins eaten in the inside of a Volvo?
Data poisoning is pre AI it plays in the “no 100% rule” it’s why we have flat earthers and measles parties…
Sorry just noticed the AI explanation feature Data poisoning means deliberately inserting false, misleading, or biased information into a body of data so that people—or an AI trained on it—reach unreliable conclusions. The comment says this problem existed before modern AI: repeated misinformation can distort human beliefs, as seen in flat-Earth claims or “measles parties.” Strictly speaking, those examples are better described as misinformation or disinformation, while data poisoning usually refers to manipulating an AI system’s training data. The “no 100% rule” means neither humans nor AI detectors can identify every falsehood or AI-generated text with perfect accuracy.
Oh, not that extreme. It’s more targeted to cleanup. Not to widespread stuff. It would be a mechanism for WordPress admins to keep a tidy functional blog/site with accurate information. Detecting content alone might not suffice, and other standards have to be introduced, and functionality for content owners. That’s where I’m getting at.
I want to emphasize that OpenAI’s terms prohibit presenting AI-generated output as human-generated. Its sharing and publication policy also requires clear disclosure of AI involvement for publicly shared generations and content co-authored with the OpenAI API.
In this context, asking the community to recommend third-party services that generate content designed to pass as human-made is also not an option.
Perfect, that’s the point I wanted to make. While OpenAI is doing it’s effort on the business side of things, a standard to do this user side is what is missing. And that the problem’s solution to tackle this ain’t easy.