are 16 of the most effective AI and ChatGPT content detection tools

AI content detectors have also gained in popularity as alternatives to ChatGPT are appearing daily.

This article will show you 16 of these detectors, and how they scored some paragraphs from my unpublished, original writing, compared to a few paragraphs about the same subject from ChatGPT.

I will also discuss the types of tasks that AI detectors can (and cannot) perform, as well as how editors, marketers, and SEOs need to think about them.

What are the best and worst detection tools?

The generative AI and chatGPT content poses a number of issues, as I have discussed in other articles:

• AI-generated data can be inaccurate, outdated, harmful, or misleading.
• The quality of AI-generated writing can be low.
• Google does not necessarily trust AI-generated content and may view it as human-created.
• AI content can “trick” editors and businesses into thinking they are paying for content created by humans.
• can repurpose AI content without crediting.

The current AI detectors don’t solve all of these issues.

These tools are not designed to fact-check AI content or improve the quality of content. They also do not provide citations or information derived from other sources.

AI detectors are useful in a number of areas, including:

• Plagiarism : These tools all have some form of plagiarism detection, which allows them to check whether AI content is largely derived from other sources.
• Penalty Prevention If you are concerned that AI content will be devalued in search results for some reason, these tools can give you an idea of how easy it is to detect AI content. Google will have its own tools and checks.
• Auditing AI use If you have an AI policy that you want to enforce, or a way to compensate authors for AI-generated content versus original content, then these tools will give you an idea of how much AI is being used by a writer. Note that these tools can return false positives and negatives.
• Understanding Search Results: Chrome extensions are available that can help you determine whether your competitors or other websites have AI content.

AI software: How it works

Each tool has a different approach and each is unique. ChatGPT detection tools mainly grade content by how predictable phrase choices within a piece are.

The likelihood of a piece being scored as AI or human is largely determined by the software’s ability to detect if the writing follows the pattern that an AI would likely follow when creating content.

Two core concepts that surround this process are:

• Burstiness : Predictable length and pace of sentence structure.
• Perplexity : Randomness in the selection of words within a sentence.

It’s unlikely that a generative AI will include in an essay on the founding of America a random and unevenly written story about the first penguin they saw. This would look like human-written text to a detection system.

Tools are being developed that will allow you to bypass the detectors. The detectors will likely be thinking of ways to detect bypassers.

Undetectable and Quillbot rewrite content to make it harder for AI detection tools to detect.

Some people have also found ways to make ChatGPT, and other AI-writing tools, produce content that is “more human”, on a human-to AI scale. They use prompts for defining perplexity and burstiness to tell ChatGPT to write more with each.

Does detection accuracy matter to you?

Before you begin using these tools, it is important to ask yourself:

What is your level of concern about the detection of AI-written content? Why?

It may not matter if the content generated by ChatGPT passes AI writing checks if you are using it to rewrite title tags or generate email copy.

If a writer generates a great copy using AI, the score may not be important.

The “detection race” will be fought by these tools with the un-detection and prompting tools I mentioned earlier.

Compare the best AI-based writing detectors

We’ll look at each one and see how it “scored” when comparing human-generated copy with AI copy or AI copy that uses this prompt in order to “beat detection.”

Note : The detection of a few paragraphs is not a comprehensive test of the detection abilities of these tools. It should give you an idea of the results you can expect if you use these tools.

You can see the samples that were entered into the tool: the “human” example written by me and the “AI” example written by ChatGPT using GPT-4. The updated version based on this topic is available here.

The table below shows how each tool scores the copy that I wrote, the copy that I copied directly from ChatGPT without prompt modification and the same copy with the “perplexity” prompt.

Here is a quick overview of some of the key features of the tools in the table.

1. Originality.AI

Originality.AI, a paid-for tool, is described as the “most accurate AI content detector and plagiarism checker for serious content publishers.”

The cost is $.01 for 100 words. The AI scanner is accompanied by a plagiarism detector.

This is the only AI scanner that can be 100% confident that the AI content is AI and the human content is human (while being accurate in every case).

Hive Moderation, (discussed in the article), came to the same conclusion. It gave a score of 0% AI for the human-generated content with 99.9% certainty that the AI-generated text was AI.

In addition to the Chrome extension, and the overall score, the Originality.AI has added a feature that highlights the specific sections in the content that you paste which it predicts won’t or will be AI.

This tool is used most often by me to verify content for AI. I use this tool most often to check content submitted by writers that we work with.

2. Writer

Writer provides a free detection that only works for up to 1,500 character and requires API access.

This tool assumed that all submissions were likely human-generated.

The free content detector allows you to enter a URL for checking.

This tool does not have a plagiarism detection feature. It gives an overall score, but does not mark any specific sections as AI-generated.

3. Copyleaks AI Content Detector

Copyleaks AI Content Detector is a powerful tool with a number of interesting features.

• You get a general verdict (human or artificial intelligence).
• You can hover over a specific text to see if it is AI or human copy.
• The tool pre-loads examples of GPT and human outputs so you can see how it works
• There is a Chrome Extension.

It’s free and, while it doesn’t scan for both, you can also use the plagiarism detector.

The software correctly identified a large portion of the content generated by humans as being created by people and a large percentage of AI-generated content, as being AI. It wasn’t always accurate and confident in certain sections.

4. OpenAI’s AI Text Classifier

AI Text Classifier, a free tool by OpenAI (the company behind GPT and ChatGPT), is the product of OpenAI.

Our test proved that the popular AI tool maker was not the best detector.

The tool classified the human content as being “very unlikely” that it was AI generated , but classified the content of OpenAI’s platform as “unclear”, if it were AI-generated. It also classified the content as being “unlikely” AI produced with or without the prompt.

The tool is completely free (you will need to have an OpenAI account). The tool isn’t very fancy and there’s no plagiarism check.

It will give you a score (like the one in the screenshot), but not a precise result. The software will not highlight certain sections of text based on their likelihood to be AI-generated.

5. Crossplag AI Content Detector

Crossplag allows you to perform a few scans for free without registering. You can get unlimited free scanning when logged in. The tool gives an overall score, but does not allow you to mark individual content sections or check for plagiarism.

Crossplaq determined that the AI and human content was 99% human. The tool is constantly stating that it works better with longer texts.

The samples we have here are above the minimum 200 characters but still on the short side.

6. GPTZero

GPTZero is available in both free and paid versions. The free version provides you with a score as well as some examples pre-loaded.

The paid version offers a plagiarism-checker, AI content highlights, and bulk uploading of multiple pieces of content. (This is designed for classroom environments, as are many of these tools.

The website does not list the price of the paid tools, but there are forms you can use to request access.

The free version of this tool detected that all three of our samples of writing were likely to be human.

7. Sapling AI Detector

Sapling AI Detector is available in both free and paid versions. It also offers a Chrome Extension and an API. The overall score, sentence highlighting, and an API are all included.

Sapling determined that the AI content was almost exclusively AI and the human content was mostly human. It identified “perplexed” and “bursty”, ChatGPT-related content, as being highly likely to contain human.

8. Content at Scale’s AI Detector

The Content at Scale AI Detector offers a free score that breaks down the different elements of content (predictability and probability) with pre-populated examples.

There is no plagiarism detection feature, and the individual content sections cannot be highlighted.

All three samples were rated as human by Content at Scale.

The content that was generated using the simple prompt had the highest percentage of human content, followed by content modified by prompt.

9. ZeroGPT

ZeroGPT offers a score, but does not detect plagiarism.

The tool returned low percentages for AI-generated probability across the board. The AI percentages roughly corresponded with the humanness of the samples (the human-generated sample receiving the lowest AI percentage followed by the burstiness/perplexity prompt and then the simpler prompt).

10. GLTR

GLTR (or the Giant Language model test room) is a free software that was developed by Hendrik Stroebelt, Sebastian Gehrmann and Alexander Rush of the MIT IBM Watson AI lab and Harvard NLP.

The tool provides some overall scores, but it mainly tells you whether each word in a text is one of the 10,100,1,000 or outside the 1,000 words most likely to be generated by AI using the context within the text. This tool does not have a plagiarism feature.

The tool returned the most “likely” words and the words that were in the top 10 of the list for the text generated using the simple prompt.

In this small test, distribution of the prompt and human copy was the same.

11. ChatGPT Detector on Hugging face

This free tool gives a score and a prediction. This free tool does not have any features to highlight or detect plagiarism.

The tool detected human content in all three samples with a probability of over 99.9%.

12. Corrector AI Content Detector

Corrector AI is a free program with a limit of 600 words and a score in percentage. The tool does not have any plagiarism detection or highlight features.

The tool assigned each sample a very low probability percentage of being artificial intelligence (AI) or “fake”.

13. Writefull GPT Detector

The Writefull GPT is a free, simple tool that provides API access and a simple score. There is no highlighting or plagiarism detection.

The results showed that all three samples were likely to be human, but the content of the ChatGPT with the simplest question with the highest AI probability (19%) was found.

14. Hive Moderation AI-Generated content detection

Hive Moderation AI Detection of Content is free. (With character limits, and requiring signing in after multiple uses.)

The software does not allow for plagiarism or highlighting, but it allows you to separate content into sections and receive segment-specific scoring.

Hive scored each sample of writing correctly, with a score of 99.9% for AI samples and a score of 0% for human content.

Only Originality.AI was able to maintain a high level of accuracy across all samples.

15. Paraphrasing Tool AI Content Detector

This free AI content detection tool includes a paraphrasing feature, a rewriter and an overall conclusion. It also has highlighting features and does not contain any plagiarism.

The Paraphrasing Tool concluded each text was probably human.

16. AI Writing Check

AI Writing Check, a free tool, offers a score and has a maximum word count. It does not have text highlighting or AI probability features.

AI Writing Check identified the three samples of writing as being human.

The key learnings from the testing of 16 AI and ChatGPT content checking systems

It’s important to note that my findings are based on very limited samples. Three short writing samples is not enough to make a firm conclusion about each tool.

This said, I discovered a few interesting patterns relating to AI writing detectors generally:

• Calibration The majority of tools classified all three writing samples as either highly likely to be written by a human or highly unlikely. When it comes to grading content, different tools are likely to be “harder or “softer”. Understanding how the tool is calibrated will help you determine their usefulness.
• Rare Features : Some of the rarest features were bulk uploads, plagiarism detectors, Chrome extensions, and highlighting certain sections that are likely to be AI rather than human.
• Paid vs. Free All the tools listed here, except Originality.AI, had at least a free version. The tools that produced the best results and had the most “rare” features were usually the ones with a paid-for version.

If you’re interested in AI detection for your projects, I recommend that you keep an eye on a few of these.

Your favorite tool may soon become obsolete due to the rapid pace of AI innovation in content generation, editing, and detection.

The article 16 best AI and ChatGPT Content Detectors compared first appeared on Search Engine land.