A Trustworthy Detector or “Should Never Exist” One?
Based on 18 reviews, GPTZero has a 2.1 stars rating out of 5. Users frequently cite that the detectors misjudges the analyzed text. In other words, it recognizes human-written texts as AI-generated or vice versa. All in all, GPTZero is ranked 20th among the best AI detectors out there.
End of story, right? The service is terrible, and that’s it! Well, it could’ve been that way had we relied on the number of reviews instead of their content. Upon closer inspection of the negative GPTZero reviews, we were surprised about the minuscule number of those that provided information on the users’ exact experience with the software.
You see, most negative comments focus on how AI detectors are generally bad. And there are multiple reviews accusing the software’s creator of being a teacher’s dog, a pig, ruining everyone’s grades. And obviously, it has nothing to do with the software itself.
So, we decided to investigate the service on our own to figure out if it’s worth your time or not. And we were quite surprised with the results.
The Worst AI Detector: Allegations or Evidence?
We can all agree on one thing: baseless accusations are very wrong. That’s why you always need to have evidence to prove something. Unfortunately, living in the post-truth era, we often come across situations when allegations become synonymous with evidence.
Where are we leading, you may ask? Basing our review of GPTZero solely on customers’ comments is impossible, as they are merely allegations. Besides, only a few of them can be considered noteworthy. If the comment explicitly explains why the user’s experience with the software was unsatisfactory, it should be considered. But not when the whole review is “this software should never exist.”
Now, the reason for our investigation lies in those few comments that are worth the attention. At least those two comments provided a detailed description of their experiences with the service and explained why they found it untrustworthy. Both users described occasions when a human-written text was flagged as AI-generated.
While we cannot recommend GPTZero to our readers, the service is far from being the worst software or the most inaccurate AI detector. And it’s definitely far from being something that can ruin your grades.
First Trial: Analyzing The Allegations and Evidence
Before we proceed with our findings, we need to answer the following question: What is GPTZero? In short, it’s a platform that allows users to get their texts checked for plagiarism and AI-generated content.
There are other features – we are going to disclose them in further sections — but, all in all, everything that GPTZero provides is more or less about making sure that texts are unique and human-written. Now, with that out of the way, we can move to the next crucial question.
How good is GPTZero? On the one hand, we have two noteworthy allegations on how the software flags human-written texts as AI-generated. On the other hand, we have our own investigation. Firstly, we let the software scan one of our articles and got a 100% human-written text result. Then, we tested AI-generated text, and GPTZero stated that it was 100% AI-generated.
However, things are not as simple as they seem. For example, running this section of the article showed us that there is a 2% probability that it was AI-generated, while it was written by our team member.
Witness Testimony, or Communicating With Customer Support
While the witness was subpoenaed, they didn’t show up for the trial. But there’s no point in proclaiming a mistrial. We really wanted to communicate with GPTZero’s support team, but it’s impossible to do it via live chat. Instead of a live chat, we had an AI bot, and it was not the smartest one.
Basically, it does two things: it directs you to FAQs or offers to email customer support. Like, if everything relies on this great software, you could have an AI bot that can answer basic questions. But the FAQs provided us with answers we were salivating over.
Allegations that the creator of the software is a teacher’s dog and ruining everyone’s grades are unfair. If someone gets lower grades because their professor trusts GPTZero detector blindly, the software is not to blame.
The FAQ section clearly points out that there are cases when AI-generated content is classified as human-written and vice versa. And it’s a teacher’s duty to figure out whether the text is AI-generated or not. There are even suggestions to interrogate the student and ask them to provide the artifacts of their writing process.
Answering the Most Important Question: How Does GPTZero Work?
Given all the information that we have and the allegations about how the software frequently gives false positive results, we need to answer yet another question: how does GPTZero detect AI content?
In short, it analyzes the text on multiple levels, but the key components of the analysis are “perplexity” and “burstiness.” Don’t go googling the exact meaning of those two words: they are used differently here.
In this situation, perplexity is the predictability of the words, while burstiness is the frequency of particular words throughout the text. And, as the rise of the machines is still not happening — although according to “The Terminator” timeline, it has already occurred — AI-generated texts are more predictable, and they suffer from extreme “burstiness.”
Once, we ordered the essay “The Pervasive Tendency to Believe Liars: Insights from Elizabeth Holmes and Anna Sorokin.” Holmes and Sorokin appeared in every second sentence of this 600-word text.
To keep it short:
Does this method work? Yes. Is it perfect? No, but it allows for the identification of AI-generated texts in most cases.
GPTZero Pricing: How Much Does Preserving What’s Human Cost?
Now, it’s time to figure out the GPTZero cost. The team behind the software claims that its mission is to preserve what’s human. Given that, according to some timelines, we are trying to survive the rise of the machines, preserving what’s human shouldn’t be cheap.
While we could’ve called the question “Is GPTZero free?” irrelevant, you can use it without paying. The free plan allows you to analyze texts of up to 10,000 words per month on the basic AI scan. And you have five advanced AI scans.
The essential plan will cost you $8.33 per month and gives you access to pretty much the same features. The difference is the number of words, 150,000 per month, and document history.
Then goes the premium plan, which will cost you $12.99 per month. It has all the features of the essential plan, plus advanced scan, AI vocabulary, plagiarism checker, and writing feedback. And the per-month word limit here is 300,000.
Finally, there’s a professional plan that costs $24.99 per month, which you can use individually or as a team. You get 500,000 words per month, all the premium features, and scan sharing.
Pick the one that suits you the best.
Further Testing of GPTZero AI Detection Capabilities
After feeding the GPTZero checker with human-written and AI-generated texts and getting predictable results, we decided to make things a bit more difficult.
We started feeding it with mixed texts. Here, we learned that the detector reacts differently based on the share of both contents.
Let’s say a 1000-word long text containing 20% of AI-generated content can pass as human-written. However, text with over 2000 words containing the same amount of AI-generated material gets flagged as mixed. If you opt for the advanced scan, you get a detailed report of which part is written by human and which is generated by AI.
Now, returning to “perplexity” and “burstiness,” everything depends on the way the text is written. In other words, if you have quite a robotic style, you can be perceived as AI. If AI can craft text with a complicated style, avoiding its typical wording, it can pass as human-written text.
So, what’s the problem? The problem is that there are other detectors that can fare better for lower prices.
AI Vocabulary: Does It Work?
AI Vocabulary plays a significant role in shaping AI detection. Of course, in today’s fast-paced world, AI is constantly evolving. Thus, software like GPTZero aims to enhance its methods of detecting AI-generated content.
AI has its own vocabulary, which consists of words that it uses frequently. No matter the themes it explores in its texts, it can’t avoid its typical phrasing. It does show promising results; still, most of its texts face criticism because of their robotic vibe.
Does the previous passage look strange? If so, we succeeded in our goal. The text contains multiple word combinations from the AI Vocabulary. And, as GPTZero relies on the AI Vocabulary, it marked the text as containing 37% of AI-generated content.
And here, we cannot come to a definitive conclusion. On the one hand, relying on AI Vocabulary is pointless, as the software flags human-written text as AI-generated when it contains word combinations from the said vocabulary. On the other hand, let’s be honest: no human with well-developed emotional intelligence writes like that.
GPTZero Plagiarism Checker: Is It Worth Your Time?
We more or less understood how the GPTZero AI checker works. But hey, the premium package includes a plagiarism checker, so we decided to test it as well. And we must admit that it pales in comparison with the AI detector.
The good thing, though, is that it doesn’t give false positive results. When the text is unique, it shows as unique. But when it comes to plagiarized texts, it shows less than other plagiarism checkers.
We took some of the essays that we received when checking different paper writing services. Those essays contained significant amounts of plagiarized content. But, while other plagiarism checkers were showing between 25% and 30% of plagiarized material, GPTZero showed between 10% and 15%.
Yet, who knows, GPTZero is less than two years on the market, so maybe its plagiarism checker will get better.
Closing Statement: The Pros and Cons of GPTZero
So, is GPTZero the best AI detector? It surely utilizes interesting methods to recognize AI-generated content. However, our investigation proved that you can trick the software into giving false positive results.
Other features that the service offers are not that impressive. GPTZero plagiarism checker is subpar, to say the least. But in no way, shape, or form is GPTZero a service that shouldn’t exist, and it can’t ruin your grades. The jury should remember that before deliberation.
.webp)
Is GPTZero legit or not?
GPTZero is a legit service that provides its users with software that allows detecting AI-generated content. Throughout our investigation, we didn’t find anything that can suggest the service being not legit.
Is GPTZero trustworthy or not?
The unique method GPTZero uses to detect AI-generated content is simultaneously the tool’s strongest asset and its weak point. As it looks for a particular writing style and word combinations, the software can identify incompetent human-written text as AI-generated. Hence, we cannot call GPTZero 100% trustworthy.
Can I use GPTZero for free?
You can use GPTZero for free, yet with limited options. The per-month word limit is 10,000 words, and you only have five advanced scans per month. The essential, premium, and professional plan will cost you $8.33, $12.99, and $24.99 per month, respectively.
Does GPTZero scam its users or not?
Despite multiple complaints about the quality of the software’s AI detection, there are no complaints that GPTZero is a scam. Throughout our investigation, we also didn’t find any proof that GPTZero scams its users.
Is GPTZero safe to use or not?
GPTZero is safe to use when it comes to your personal and financial data security. The service is not involved in any fraudulent activity. However, if you want 100% accurate AI detection, you need to have a clear understanding of how the software functions.
Is GPTZero reliable or not?
Despite its unique approach to AI detection, the software can, from time to time, classify human-written text as AI-generated and vice versa. Hence, we cannot claim that GPTZero is a 100% reliable service.
*Screenshots used for review and criticism purposes fall under the 'fair use' doctrine of the United States Copyright Law.
The latest customer reviews
We've received your feedback!
You Might Also Be Interested In
