Menu
Log in
  • Home
  • January, 2026 Feature - GPTZero

GPTZero

by Kelley Garcia




I watched a video on YouTube where a teaching assistant was explaining how the professor he supported requires him to use a tool called GPTZero. Here is the video 

https://youtube.com/shorts/_S7nwMXMZ5g?si=DhzgrCeywscA0ybI

I was curious. 


The GPTZero application was created by Edward Tian, a Princeton University undergraduate student, as part of his senior thesis project. He launched the tool in January 2023 to address concerns about AI-generated content in academic settings.


Like most software, there were a few options: Free, Premium or Professional.  I tried the "Free", more about that in a moment.


I downloaded both the extension for the Google Chrome browser, the add in for Microsoft Word, and added it to Google Docs as well.  From the Extensions page overview

(1) Live AI detection checks instantly as you write

(2) Customizable feedback and writing comments in Google Docs

(3) Writing video replay and report, with a new typing pattern analysis

(4) Bibliography, citation, and a research-focused fact check


GPTZero’s Writing Replay brings your Google Docs writing process to life, showing exactly how a piece was written, edited, and pasted. Now supporting multiple editors, switch to see how different writers contributed to the final document.


Gemini claims that unlike other writing replay solutions like Draftback, Brisk Replay, Grammarly Authorship, or Turnitin Clarity, the GPTZero Writing Report includes:

• Multi-user detection: See how much each writer contributed by percentage.

• Instant replay: Fast-loading, high-resolution writing playback.

• Activity insight: Identify frequent edits, pastes, and writing bursts.

• Human vs. AI analysis: Built-in AI detection flags large pastes and unnatural text.

• Typing pattern insights: Our proprietary algorithm uncovers human typing behaviors.

• Security: Enterprise grade data security and privacy, SOC2, and FERPA compliance.


It ranks 4.2 out of 5 based on 182 ratings. 


The first test I did was on a document a friend of mine had created. He was using Gemini (Google's AI Assistant agent) and was asking about various possible scenario outcomes. The document contained both his questions and Gemini's responses. 


 

For something almost completely written by AI (Gemini), I was surprised it only returned at 55% AI probability. 


The interface is a little clunky because it was advising to select at least 250 characters but simultaneously it read the entire document. 


Also, as you can see, it has "4 free left". That is regarding "Advanced Scans". I used two in the creation of this article. To get unlimited you have to pick a plan. In addition to Premium there is also Essential ($15 / month if billed monthly and limited to 150,000 words per month) and Professional ($46 / month,  if billed monthly and limited to 500,000 words per month).  Their subscription plan page compares all the features. 


I selected the Typing analysis and I have to admit it was creepy to see my edits to the document being played back to me in a recording. Places where I typed (in Google Docs) questions, comments, or suggestions about the information or what to ask next. It plays it back as well as provides feedback; such as, "Unnatural typing patterns such as large text insertions and /or unusual changes in speed or rhythm. I can remember when this sort of thing was only referred to as Spyware.


The extension identified the original creator and myself by name as the two editors, even informing that the originator edited 61% of the document and showed I had  edited39%.


It also shows the dates and times, numbers of edits, pastes, time spent, and word count. 

The typing analysis is "recorded" and you can play it back with the ability to change the playback speed from 1 to 30x.


When you upgrade to a Premium plan you get a plagiarism scanner. And there is a Hallucination Detector which identifies and flags hallucinated citations and fake academic sources in text. 


The tool also gives Writing Feedback like Content/Clarity/Grammar and an option to ask for custom feed advice as well as citation accuracy. 


So not only is it a tool for TAs and Professors (or whomever) to use in checking documents, it can also be used to create documents. 


There is also this "Guidance for Teachers and Reviewers: The nature of AI content is changing constantly. These results should not be used to punish students. We recommend educators use our behind-the-scenes Writing Report for a holistic assessment. See our FAQ for more information.


I came across many posts in REDDIT and other social media sites of writers complaining their work was flagged as AI.



Draftback, Brisk Replay, Grammarly Authorship, or Turnitin Clarity are alternatives but I did not test any of them. I did see where many of them, like Brisk, started as Free but now subscriptions are required.  As AI gets better at mimicking humans these detector tools will need to improve.  


I wanted to see how it worked in Microsoft Word but didn't like it - it doesn't have many features and can only handle small chunks of text. The interface in Google Docs is much better. I removed it from word and I hardly ever use Google Docs so I don't plan to get a subscription but, because I added the extension now it shows up in the browser for every page I visit. I decided to test it on one of my web pages.  A little swirl on the right edge of the page.  When I click that I have three choices: Scan Full Page,  AI Detection, and Fact Check. There is also a small icon for Settings. 



I ran the Scan Full Page it returns its opinion that the page is 100% written by a human.  When I selected to see Full Results it looked like the same returned information as from the AI detection selection. 


When I clicked on the orangey highlighted one in the middle, the one with Disagree (3), the results were for something completely not related. Some legal summaries from a case about American Eagle's portal. So clearly not at all related to the topic of the post. 


The nuanced results were also completely irrelevant. 


When I ran the same report again I received different results. 


Wanted to do more but then received the disappointing message that I had used up all my freebies and would have to select a plan to go further. 


I have to say that my experimentation with GPTZero was not very inspiring. It was not accurate. It was a bit scary to think that students' work is being judged by this tool. Also, I always dislike being lured in to use a FREE tool only to then be told I have to pay to use it. The ole bait-n-switch routine. Not good. 


Nonetheless, I'm glad I tried it and have seen for myself how it functions.


If you have used it or use it regularly in your work perhaps you'd be interested in sharing your experiences with the rest of us as the February Feature?  



Powered by Wild Apricot Membership Software