How to Understand Your AI Scan Results

You’ve started incorporating GPTZero’s advanced AI detection into your work, scanning and uploading documents to check for ChatGPT and other AI chatbots. Now you’re wondering: I’ve got my scan results. What else is there to do?

We go over how best to understand and use your AI detection results, including common misconceptions about how to interpret AI scan reports.

How GPTZero works

GPTZero Quick AI Scan comes with a categorization of “AI”, “human,” or “mixed” (both AI and human). Each result includes two parts: an AI probability score and a “confidence” statement. With GPTZero Advanced scan, you get more granular, sentence level highlighted detail on which sections are contributing most to your AI/human score.

Understanding Probabilities

You spent hours writing something up and it shows up as “93% human?” Where is the “6% AI” coming from?

Remember, the percent you’re seeing is NOT saying that “6% of your document contains AI.” GPTZero’s percentage refers to our model’s probability that your document was written by either AI or human.

Probability is a calculation that predicts how likely something is to happen; in this case, how likely a document was written by a human or AI.

In other words, “6% AI” means that our software believes only 6 times out of 100 cases we’d see something similar written by AI, and 93 times out of 100 times, your document was written by a human.

Unlike our plagiarism detector, which can find an exact copy of something that is referenced or duplicated on the internet or databases, current AI detection is always probabilistic and predictive. This is because AI is constantly evolving and adapting, and so is human language.

There is a difference between AI and human writing, and we seek to make sure we preserve knowledge of that difference. However, probabilistic results, where sometimes the confidence score can be low, are also why we don’t recommend using an AI detection result as the only proof for academic punishment or discipline.

Read more on what to do if you detect AI in your classroom.

Understanding Confidence

Every score also comes with a statement about our confidence in our prediction. Confidence measures the rate at which we would incorrectly identify the text document (error rate). These can range from:

• Highly confident: our error rate is less than 2%
• Moderately confident: our error rate would be around 10%
• Low confidence: our error rate would be 14% or higher

Our confidence levels adapt and are set automatically based on maintaining the highest accuracy rate (95%+) while also keeping the lowest false positive ratio (<1%). In other words, if we aren’t confident something is AI, we generally air on the side of “human” with a low confidence score, to avoid the case where a human is falsely accused of AI. 

Understanding Sentence Impact & Highlighting in Advanced Scan

Advanced scan offers more detail into just how we make our AI probability determination. We scan your text to find which sentences and phrases are disproportionately affecting your overall AI or human score. Sentences with high impact are very likely to be adding to our confidence in your AI/human probability score, while sentences with medium to low impact are proportionately affecting your score.

Some sentences in an otherwise AI or human document might not show up as highlighted. They may still have an effect on your score, but they aren’t more likely than other parts of your document. You may find that adjusting an entire document changes your score more than just adjusting the sentences. This is because our detector holistically analyzes the document, and its prediction can depend on a pattern that affects the bulk of the text.

Understanding AI Vocabulary

AI Vocabulary is a separate scanning option that shows you the most common words that AI tends to overuse. Right now, AI Vocabulary scan doesn’t affect your Advanced or Quick scan AI probability directly, but is a useful tool to show where AI terms and phrases can crop up, even in otherwise human passages.

Our goal in providing you this tool is to make sure in your human writing you can avoid sounding like a bot.

Learn more about the most overused AI words and phrases.

FAQs

I got an AI score from GPTZero on a doc. What should I do next?

• If you’re a teacher: here are some tips to start dialogue with students about academic integrity and school policy.
• If you’re a writer or student: there’s nothing wrong with using AI tools to help you get started. But if the goal of a project requires original thinking and effort, you want to ensure AI isn’t undermining or taking credit for your effort. 
  • Step 1: Download an Authorship Verification tool like GPTZero’s Origin Chrome browser extension for Google Docs.
  • Step 2: Revise and edit your document with your own words. You can generate a human writing report and include video playback of your writing to prove your human effort.

What happens if I change my text? Will it still evaluate as AI?

Sometimes yes, sometimes no. Depending on the depth of changes, there are still parameters in AI-generated text we detect that persist even if you change words around. But in general, writing and developing your own human voice prevents you from sounding like (and being detected as) an AI chatbot.

My writing got labelled as AI but it was definitely human (a false positive). What should I do?

If you’ve downloaded the GPTZero Origin browser and have an authorship verification report showing human effort (and still are misflagged as AI by our service), please bring this to our team’s attention. We are always working to improve our tool, and as thanks, we will give you a Certified Human badge and report. If you are facing disciplinary issues as a result of a false positive (including from other tools), please reach out to our team.

How accurate is GPTZero?

GPTZero is 95% to 99.5% accurate with a 1% false positive rate. With mixed samples, we have 96% accuracy. And with AI paraphrasing detection, we have 95% accuracy.

See all our accuracy benchmarks here.