The rise of AI content generators in the digital landscape has been an interesting phenomenon to watch. While it makes it easier for anyone to create content, it has also contributed to the increase of really bland and generic content.
Having dabbled in AI content generation myself, I find it fairly easy after a while to find out if something was written by AI. Very briefly, they use similar wording, phrasing and certain sentence structures. If you are good at detecting patterns, chances are you would be able to detect this too.
One of the tasks I do at Dolce Marketing is to look through the AI generated content and try to rewrite it to make it more helpful in Google’s eyes. This involves rewriting large chunks of the article and later on checking if it was plagiarized and whether it sounds like it was written by AI.
To my surprise, I had large re-written parts of the article be classified as ‘written by AI’. Even though it was really clearly written by me. I then input the same paragraphs in other AI Content Detectors and found that the results were very different.
So this got me thinking, which AI Content Detector is the most accurate?
In this article, I will be testing the top AI Content Detectors found on the first page of Google and see how accurate they are.
Testing Methodology
This test will be very simple. I will use one of my university assignments that I wrote long time ago (pre-AI work years before AI generation). I will then input this into the top AI Content Detectors on the first page of Google and document what each detector says.
I will then ask AI to write a second paragraph based on what I initially wrote. So ideally, the AI detector should provide a 50% human 50% AI detection score. The goal of this test is to see how accurate these AI content detectors are when doing a quick test.
The written text I will use:
One of the key findings among the three interviewees was the need for convenience. As an international student with a busy timetable, there is a need for food that can be provided or prepared quickly so that there will be minimum disruption in their schedule. The process of cooking and washing dishes is too time consuming and troublesome as compared to ordering out. Such examples include buying pre-cooked food from supermarkets, fast food restaurants and even from food delivery services.
And the second half will be done by AI:
The second key finding among the interviewees was the importance of cost-effectiveness. As international students operating on limited budgets, there is a significant focus on finding affordable meal options that provide good value for money. The expense of dining out regularly or ordering delivery can strain their finances, leading them to seek budget-friendly alternatives. Such examples include buying groceries in bulk during sales, utilizing student meal deals at campus cafeterias, and sharing cooking expenses with roommates through communal meal preparation.
Disclaimer
I will be using the free version of the AI Detectors since everyone will have access to those. For most local businesses, the free version should be good enough as you would just be doing basic checks on plagiarism and AI detection.
For example if you hired a content writer, you want to do your due diligence and make sure that there aren’t any issues with plagiarism and AI generic content as that could hurt your SEO.
This will not be a detailed review on the different features that each AI Content Checker has. I do not have a subscription to all the AI Checkers and I have not used them extensively. This will simply be an ‘eye ball’ test.
I also am not affiliates with any of these AI Content detectors and do not have any affiliate links in this article.
How do AI Content Detectors work?
Before we go through the test, we should first find out how AI Detectors work. AI Content Detectors use sophisticated algorithms and techniques to distinguish between human and machine-written text.
Natural Language Processing
NLP algorithms analyze text by examining syntax, grammar, and natural language flow. This system evaluates sentence structures and identifies patterns typical of AI-generated content, looking for repetitive structures or unnatural phrasing.
The detection process relies on two primary metrics:
- Perplexity: Measuring the unpredictability of content. Because human writing typically shows higher perplexity scores.
- Burstiness: Evaluates variation in sentence length and structure, where human writing demonstrates greater diversity.
Machine Learning Implementation
The detection systems are trained on extensive datasets containing both human and AI-generated text. They use predictive analysis to recognize subtle differences between the two types of content.
Feature Analysis
Detectors also examine multiple textual elements:
- Lexical Features: Word frequency and vocabulary choices.
- Syntactic Patterns: Sentence structure and arrangement.
- Semantic Analysis: Context and meaning evaluation.
1) Copyleaks
Founded in 2015, Copyleaks first started as a plagiarism detector and now includes AI Content Detector. Copyleaks uses sophisticated algorithms to analyze text across more than 100 languages and can detect various forms of content manipulation, including paraphrasing and source code plagiarism.
Copyleaks claims to have 99% accuracy in AI detection across 30 languages. A scientific study conducted by Cornell Tech-owned arXiv declared Copyleaks as the most accurate AI detector among eight tested tools.
90 users have rated Copyleaks at 4.4 out of 5 stars and academic users have reported successful results when checking for thesis verification. Some users however reported false positives where it inaccurately flagged pre-AI student work as AI written. No AI tool is perfect after all.
How did it Fare?
Copyleaks rated it as ‘AI Content Detected’. It doesn’t give a percentage on how much is AI written/Human written. It also does not highlight sections based on whether it is AI written/Human written. While this might be fine for most, I do appreciate it if the free tool provides more feedback.
I tried creating a free account to see if I would get more detail but it basically gave the same result – 100% AI Content Detection. Suffice to say this is quite inaccurate. However I stress again that this is a very basic test and not a complete review of the full product.
From a local business perspective, I most likely won’t be using Copyleaks as a quick check.
2) Quillbot
Quillbot is next on our list, serving 56 million users worldwide and boasting 5.7 AI content checks per second. An independent test showed that it had 80% accuracy, tying for second highest among the AI detectors. It has been noted that it struggled with mixed content, with test results showing only 31% accuracy for mixed content.
The free version of Quillbot comes with a 1200 word limit per check and the premium version increases that limit to 25,000 words. As you would see later, it gives a percentage on how much is human written/AI written. And it also highlights the sentences that are likely AI written.
How did it Fare?
This is a screenshot from our basic test. It came back with a 43% AI generated and 57% human written score. I would say that this is quite an accurate reflection. Of course it did not detect the first sentence of the second paragraph.
I do appreciate that the free version provides the percentage score and highlighting of the text that needs to be improved. As a local business user, I would recommend trying out Quillbot as a quick AI content tester.
Scribbr
I must quickly add that Scribbr is another AI content detector on the first page of Google. But it is using Quillbot’s AI detection system. This screenshot from our test gave back the same result basically.
3) Writer
Writer.com AI content detector is the third on our list. It comes with a 5000 word limit per check with a paid plan increasing the limit up unwil 500,000 words monthly. A comprehensive test by academichelp.net gave it an accuracy score of 33.33 out of 50. Other independent studies showed that it was able to identify human written content correctly 95-96% of the time.
The same study however showed that it had incorrectly classified AI written content as human written. It rated ChatGPT text as 81% human written and Gemini text as 96% human written.
How did it Fare?
Based on our test, it came back with a 83% human-generated content rating. Which is quite inaccurate actually. This would line up with the results from the other independent tests done.
It also does not highlight which parts are human written and which parts are AI written. Making it slightly inconvenient as you may not know where to begin making changes.
As a local business, I probably won’t be using Writer.com AI detector. It may not be the most inaccurate, but there are more accurate free AI detectors that have better features as well.
4) ZeroGPT
ZeroGPT is one of the first AI detection tools that I used. It utilizes a proprietary DeepAnalyse™ technology to detect AI written content. It claims to have over 98% accuracy rate in detecting AI content and achieving a 100% accuracy test in a recent independent test by Word Spinner. They are also currently working on reducing the error rate to less than 1%.
The free version of ZeroGPT features a 15,000 character test with multilingual support in several languages. It’s DeepAnalyse™ technology examines the text’s linguistic patterns, syntactic structures, text randomness and variation in sentences.
How did it Fare?
So how does it fare? In this test that we did, the results came back with a 0% AI written score. That is of course strangely inaccurate. Especially when considering it’s technology and track record. Of course again, this is a simple test and results might be different when tested further.
As a local business, it might be worth a try using ZeroGPT. I have had accurate results in the past when using ZeroGPT, so this particular test we did might be an outlier.
5) GPTZero
GPTZero was also one of the first AI detector tools that I used. GPTZero is one of the leading AI content detection tools out there. It uses advanced algorithms and deep learning models to analysis the text. It boasts a 99% accuracy rate for AI vs human detection, 96.5% accuracy for mixed AI and human content, and 1% or lower false positive rate.
If you are wondering, ZeroGPT is actually a copycat version of GPTZero that launched after GPTZero’s initial success. They are however different in terms of algorithms and processes used and are not entirely the same. Both are really popular and impressive tools with impressive results. You can’t go wrong with either one.
How did it Fare?
In our test, we received a 37% AI generated rating. Although i’m not sure why the whole text is highlighted. It might be just a visual glitch as it is the free scan.
If you create a free account, you have access to 5 free advanced scans that highlights which parts are human written and which parts and AI written. This is of course slightly inaccurate although at least you know which parts are considered human/AI written.
As a local business, I think that it is worth trying GPTZero out. It is fairly accurate and the UI is quite helpful in determining which parts need to be improved.
6) ContentDetector.ai
ContentDetector.ai is another AI content analysis tool that is powered by Content at Scale’s AI detection technology. It has served over a million bloggers and academic writers since it’s launch. ContentDetector.ai does not have word limit restrictions and is able to detect content from ChatGPT and other AI models.
It claims to have 85% accuracy when detecting AI content. However, independent testers did not manage to achieve the same results with the tool inaccurately identifying mixed content as entirely human written. It regularly detected a higher probability of the content being written by AI than being human written.
How did it Fare?
Despite the concerns with inaccuracy, our simple test results shows that about 56.25% of the content is AI written. This may be slightly inaccurate but still falls within a reasonable rate. The results are also consistent with claims that it is misclassifying it as artificial content instead of human written.
Despite the slight inaccuracies, I would say a local business can consider using ContentDetector.ai. It would be also be great if it highlighted the parts where it considered it to be artificial. However not having a word/character limit is a major plus point for AI content detectors.
7) Grammarly
Everyone should know our next tool Grammarly. It started out as a writing tool that helped users check and improve their writing with grammar checks, spelling checks and more. Now it has included features to improve writing with AI and checking for AI written text.
In terms of accuracy, it detects 90% of basic grammar and spelling mistakes. It identifies 75% of the complex grammar mistakes while also recommending corrections. It also has a plagiarism checker that is 95% accurate. At the time of writing, there hasn’t been a detailed test on how accurate it’s AI detection tool is.
How did it Fare?
Based on our test, it says that 33% of our text is AI generated. This is of course quite inaccurate and it does not show specifically which parts are human and which parts are artificial. However, this could be because of the free version. If you have Grammarly premium, you would have access to the AI checker and plagiarism tool, which might provide more detail.
As a local business, I will not recommend using Grammarly’s free AI detection tool as a quick and free test. However, if you have a Grammarly premium account, it might be worth using it’s AI checker tool together with it’s writing tool.
8) Ahrefs
The last AI detection tool on our list is Ahrefs. Ahrefs is one of the leading SEO tools in the market. But it also provides free tools such as the AI content detector, keyword finder and the backlink checker. Ahrefs free AI Content Detector also rewrites your text to make it sound less AI-generated.
In terms of accuracy, Ahrefs AI content detector failed to effectively detect AI-generated content in controlled tests. It was unable to accurately identify 100% AI-generated content from ChatGPT. It also provided inconsistent results when compared to other AI detection tools.
How did it Fare?
Ahrefs AI Detection tool rated our text as 40% AI-generated. This is slightly inaccurate, but not as wildly inaccurate as other independent tests have claimed. For the fun of it, I decided to put the new ‘humanized’ version from Ahrefs into another tool such as GPTZero.
GPTZero came back with a 100% probability that it was AI-generated. The other tools had a range of 50% – 90% probability that it was AI written. You might be thinking next, what if we asked Ahrefs what it thought of it’s own ‘humanized’ version?
It rated it’s own ‘humanized’ version as 80% AI-generated. Of course I did not try the tool further as I think we got a fairly good assessment on it’s accuracy. Suffice to say i’m not really impressed with Ahref’s AI detector tool and would not really recommend it for a local business.
Conclusion
While this was a really simple test on the capabilities of the AI-generated content detectors, hopefully it would be enough to know which ones are worth using. It is important to remember that no tool is 100% accurate and the best practice is to check the written text using a few tools to get a general picture.
Would you like a more detailed test in the future? Let us know what you think.
In the mean time, read about why you need a Local Business Blog in 2025.
