AI detector to keep your texts authentic

Pre-2022, no one would have even thought your paper needed checking for AI. So, tools like Turnitin focused on finding precise or near matches to existing content to flag plagiarism.

These days, however, educators are on the lookout not just for plagiarism but for AI-generated content, too. Turnitin now checks for both by default, for one. Your school probably already added provisions to its academic integrity policies to forbid submitting AI-generated content as your own, too.

The thing is, even if you know you wrote your paper yourself, passing an AI content check isn’t a done deal. AI detectors watch out for overly generic passages, for example. So, if your writing lacks specifics and your arguments are vague, your assignment might end up getting falsely flagged as AI-generated. The same applies if the text lacks variety in sentence structure and length or uses generic transitions.

On top of that, if you’ve been using AI tools as writing assistants, your paper might be at higher risk of getting flagged as AI-generated. For example, asking AI to polish your draft can make it appear less original.

At the end of the day, running an AI content check before submitting your paper gives you peace of mind. A detailed report will highlight the sentences likely generated by AI, giving you an opportunity to rewrite them. That, in turn, will help you avoid unpleasant conversations with your instructor, outright accusations of AI use, or even penalties for the perceived academic dishonesty.

AI content checkers have their undeniable advantages. They can quickly give you an idea of whether your content is likely AI-generated, at any time of day and for free.

But is there a 100% AI detector? That’s the thing: there isn’t. Even the most rigorously tested detectors will still have false negatives and false positives. (False negatives mean AI-generated content goes by unnoticed; false positives refer to situations where human writing gets mistaken for AI content.)

Don’t worry: most detectors keep their false positive rates under 2%, so it’s a fairly rare occurrence. That said, potential false positives and false negatives aren’t the only reason to review your draft manually (or ask a professional writer to do it).

While AI content detectors excel at detecting patterns inherent to AI writing, they don’t check the text for factual accuracy. If you were using an AI tool to create an outline, you may have unknowingly relied on hallucinated claims. An AI content scan won’t catch those fallacies, but a human expert will.

On top of that, a reviewer will be able to grasp whether the text brings forth qualitatively new ideas. That’s important because AI tools can’t generate something completely new; they only mix and match existing ideas in their databases.

So, in addition to running an AI content scan, look at your draft with a fresh pair of eyes, and ask yourself: Does it contain original ideas? Is it factually accurate? Are all sources reliable and relevant?

Even the best AI content checkers aren’t 100% reliable. They may miss a myriad of telltale signs of AI writing, like factual inaccuracies, bogus sources, or a lack of original ideas.

But those aren’t the only ways AI detectors can fail. However, before breaking them down, we’ll need to take a quick detour to discuss how an AI detector essay works.

AI detectors assess two key indicators:

• Perplexity, i.e., unpredictability in a sequence of words. The more unpredictable it is, the less likely it is to have come out of an AI tool.
• Burstiness, i.e., diversity of sentence structure and length. AI tools tend to produce content with very similar sentence structure and length.

The thing is, your own writing might be falsely flagged as AI if it fails to score high on both of those indicators. That can happen if:

• You lack vocabulary and writing skills to introduce variability and unpredictability to your writing
• You’re a non-native English speaker (AI detectors tend to falsely flag writing from non-native speakers)
• You’re following the established conventions of the paper type (e.g., lab reports have a very rigid structure)
• Your text is fairly short (fewer words and sentences mean less information to analyze)

All of this is why you shouldn’t rely solely on AI checkers to evaluate whether a text is human-written. Add manual review to the mix. Plus, don’t get too despaired if your draft doesn’t earn a 0% score: even Turnitin deems a score under 20% unreliable.

Ready to run an AI content scan on your text? Great! But before you do, keep in mind that you should choose your AI detecting tool wisely — and use it responsibly, too. Here’s what it means in practice.

Privacy

What will happen to the text you paste or upload to the AI checker? Double-check the terms and conditions to answer this question before running a scan. Your goal here is to make sure the text isn’t stored or used for purposes other than AI content checks. (Some detectors may use your input to train AI models, for example.)

Accuracy

Is this tool accurate enough, and is it transparent about its accuracy rates? Make sure it clearly addresses false positives and false negatives and can catch the use of anti-AI detector and AI rewriting tools. You can also check the accuracy yourself by running scans on several human-written, AI-generated, and mixed texts.

Responsible use

It’s not the end of the world if your genuine writing is partially flagged as AI. Educators are privy to the limitations of AI content scans and mostly use them to kickstart a conversation about AI use, not downgrade assignments. They won’t (or at least shouldn’t) blindly rely on AI content scores to draw conclusions.

So, it’s better to treat AI content check results as writing recommendations. Take a look at the flagged portions of the text. Can you make them more specific? Or rewrite them so that they read like something uniquely yours?