Jon Gillham

Originality - ChatGPT AI content detector & a modern plagiarism checker

Combining AI Detection and Plagiarism Checking to determine the Originality of a piece of content. A modern plagiarism checker built for serious web publishers and the most accurate AI Detector capable of detecting GPT-3 and ChatGPT-generated content.

Add a comment

Replies

Best
Shahana Akhter
This app is a time demand when so many AI apps are in the market, and most writers use these tools. So, to detect AI writing, this app will play a vital role. Moreover, we need to spend extra money to check for plagiarism. Try to make the app perfect and less costly.
Jon Gillham
@shahana_akhter we hear you on the price. We don't plan to be the cheapest. There are free GPT-2 detectors but our plan is to be the most accurate and useful detector for serious web publishers.
Erica Martin
Do you have any plans to launch a Microsoft Word extension?
Jon Gillham
@erica_martin1 not yet... you are the first user to ask for this. What is your use case as a writer or editor or student?
Erica Martin
@jon_gillham I'm a writer. I write most of my stuff in Word and then upload or copy and paste it into the website I'm submitting it to. So having a Word add-in would be helpful.
Jon Gillham
@erica_martin1 great thank you for the idea... we have added it and your comment to be considered as a future feature.
Spencer Haws
Congrats on the launch! I'm glad that I could be an early beta tester and user of Originality. I've hired well over 20 writers in the past year, and going forward I'll be using Originality to make sure I'm getting original and quality content before hiring. I'm sure this is just the beginning of an awesome AI content detection tool!
Jon Gillham
@nichepursuits Thanks Spencer... appreciate your input along the way and support on this project!
Filip Ailer
As a professional Copywriter, it is my duty to deliver plagiarism-free content to my clients. The quickest way to check my articles for plagiarism is by using OriginalityAI. It not only tells me what percentage of my articles is considered plagiarized but also finds similar articles on the Internet. While some percentage of plagiarism is acceptable (due to keywords and specific descriptions), it should ideally lie under 10%. OriginalityAI has made my job much easier in this regard. I also like how it allows you to detect any trace of AI in your work. My clients often pay extra attention to AI presence in their articles. This is because Google rejects AI-written pieces for SEO purposes. One way to prove my articles are AI-free is by using OriginalityAI. I am grateful for the free credits OriginalityAI provided me with. Little did I know I would soon find the best proofreading tool on the Internet. I've since decided to invest in more credits, as it makes my job much easier. My only question is about the nature of AI detection. How exactly does it work? It sometimes rates the AI content in my articles as high as 20 %, even though I use no such tools at all. Does some percentage of content always resemble an AI? Regardless, I'll certainly be using OriginalityAI for years to come. Thank you for developing such a wonderful product!
Jon Gillham
@filip_ailer Thanks for the comment and glad you have been successfully using Originality.AI. No AI detection tool will ever be 100% accurate and our tool will have some false positives. It is best to think about the AI score across a large sampling of content not just a binary decision on 1 article. What I can suggest is that... 1. Look at a larger sample than 1 article to make a decision on a writer 2. A writer that gets typically very low scores has 1 higher score... not necessarily AI For example if you have a writer who submits 10 articles and these are the AI scores... Writer 1: 99, 75, 99, 99, 0, 15, 99 65, 74, 99 - Very good chance they are using AI with some minor edits to some articles Writer 2: 0, 0, 15, 0, 1, 40, 75, 0, 0, 20 - Very good chance they are not using AI but had a couple articles where they needed to write very formulaic content which can trigger a false positive Writer 3: 45, 55, 68, 99, 0, 43, 25, 15, 20, 65 - Hard to say anything conclusively but based on what we have seen this is likely heavily edited AI or AI used for some parts of the content
Nathan Petitpas
Jon, with uncertain times and countless google updates we're able to consistently* catch AI written content that could negatively impact our sites that generate traffic via SEO. Although like many others, I also wonder about thresholds and accuracy. I've contacted support, ran the tool on 100s of articles, read the comments here and all blog posts from Originality.AI but still don't have a great gauge on what % should "fail AI" checks. Since this seems like a recurring issue or question, myself and many others would benefit from a public statement or protocol. Something like: "10,000 words from the same source or 5 articles longer than 2k words with a combined AI% of 20%+ is likely 99% AI generated content after considering the current inaccuracies and state of the tool." Would you all be able to provide a suggested protocol like this?? -- After many tests, our current threshold is no more than 15% certainty it's AI or we send it back to the writer. But when I test my writing and close colleague's writing, it's never more than 5% AI. So this raises concern if the writers are re-writing AI content to make it easier on them, how accurate the tool really is, and how different writing styles may be interpreted. For anyone else here, what're your current thresholds for other writers and what % does your personal writing typically flag for?
Jon Gillham
@nathan_petitpas Hi Nathan. Appreciate the questions and fully agree this would be helpful. Everyone ultimately needs to make their own risk-based decision. For example, some people are trying to figure out... how much do I need to edit AI so that it is undetectable while others are trying to determine what is the AI suspected threshold to KNOW it wasn't AI. The reality is that predicting AI is not the same as determining plagiarism. I am working on some communication/guide on how to help with different objectives. What I can suggest is that... 1. Look at a larger sample than 1 article to make a decision on a writer 2. A writer that gets typically very low scores has 1 higher score... not necessarily AI For example if you have a writer who submits 10 articles and these are the AI scores... Writer 1: 99, 75, 99, 99, 0, 15, 99 65, 74, 99 - Very good chance they are using AI with some minor edits to some articles Writer 2: 0, 0, 15, 0, 1, 40, 75, 0, 0, 20 - Very good chance they are not using AI but had a couple articles where they needed to write very formulaic content which can trigger a false positive Writer 3: 45, 55, 68, 99, 0, 43, 25, 15, 20, 65 - Hard to say anything conclusively but based on what we have seen this is likely heavily edited AI or AI used for some parts of the content Is this the type of information you are looking for to be able to make an actionable plan on how to handle Originality.AI scores?
Nathan Petitpas
@jon_gillham That makes sense and I see the rationale behind different uses and "thresholds'. Yet the data I have on this is limited to a few 100 articles and even a few 1000 articles may not be enough for us to have a concrete understanding of what flags/ why it flags/ set fair thresholds since it seems the interworking are somewhat proprietary. Plus, I'm not a data analyst. Yes, this is exactly the type of information that would help those in the SEO realm feel more confident in the scores and how we handle the content. We don't want to falsely accuse writers either so something from the horses mouth -- even if you make no guarantees on the threshold/ protocol -- would allow us to be more firm on our thresholds or stances. I'll be keeping an eye out on the blog in hopes you all release something like this but let me know if I can subscribe to a newsletter or anything to be notified of new publications. Thanks!
Jon Gillham
@nathan_petitpas thanks! super helpful insights.
Anjum Sayed
This is excellent! It's great to see tools like this emerging. What model do you use to analyze the text and how much data were you able to train it on?
Jon Gillham
@anjum_sayed Here is a detailed post to answer how it works... https://originality.ai/how-does-...
Vijay Singh Khatri
@jon_gillham I would love to check it, but what if there should be a trial period for it?
Vijay Singh Khatri
@jon_gillham Also, I have signed up for an affiliate program too, It's fantastic for me because every marketer is looking for these tools.
Jon Gillham
@vijay_singh_khatri signup during the launch day to receive 500 free credits. No credit card required.
Jon Gillham
@vijay_singh_khatri agree, good luck
Ayesha Awan
Thank you for making such an amazing product! I am so excited to use it and see the results!
Abid Unnisa
I have tried some apps claiming to detect the AI content detector. But they dont work. Eager to check this one out
Jon Gillham
@abid_unnisa some other apps are built on testing GPT-2 which makes them not as successful at detecting ChatGPT as Originality.AI is.
Irfan
Kudos, great launch at the perfect timing. With AI taking over the internet, it is necessary to verify genuine work. I found it very good. I don't know about the results but it seems pretty awesome. I did find some bugs as well. I will send an email to the support team regarding the bugs. Good luck.
Jon Gillham
@irfaniffy Hi Irfan, appreciate you being willing to share some bugs! Not a perfect tool yet but working hard towards it!