Yes, you can get 50 credits by installing the free AI detection Chrome Extension to test Originality.AI’s detection capabilities. 1 credit can scan 100 words.

Yes, all scans are stored for later retrieval or sharing. Complete removal of your account and data is available upon request. 

Originality.AI’s AI detection is currently only trained and tested against English language. Plagiarism checking works across multiple languages. 

Yes, Originality.AI can detect ChatGPT content. 

We completed a correlation study of 20,000 web pages and identified a small correlation between Originality score and Google search result ranking. Potentially indicating that AI content performs worse in Google than Original content. 

Google has said that the appropriate use of AI to make content more useful is not against their guidelines. 

However, they have cotninued to be clear that the use of AI in an effort to game search results is against their guidelines. “Using automation—including AI—to generate content with the primary purpose of manipulating ranking in search results is a violation of our spam policies.” https://developers.google.com/search/blog/2023/02/google-search-and-ai-content

Or worded another way from John Mueller…

https://mastodon.social/@johnmu/110128231319270162

At Originality.AI we are not against AI content but believe that the decision to accept the Google risk of publishing AI generated content should be the decision of the website owner and not the writer. 

Yes, per our Terms and Conditions credits expire in 2 years from the date of purchase.

Yes, we have a robust API and have integrated it into the content operations platforms for some of the largest publishers / writing agencies and marketplaces in the world. 

Full documentation is in the backend. If you have a unique use case or are wondering about capabilities to handle your volume please reach out and we will be happy to discuss your needs.

Our internally built artificial intelligence uses supervised learning with multiple models including a modified BERT model to predict if content is AI or Original. Our AI has been provided with millions of records both AI and Original content then trained to tell the difference between the two. After each training session a large test data set is used to evaluate if the new model is an improvement or not. 

AI detection scores are a prediction on the chance that the text submitted was AI generated or Original. They are not a measure on the amount of AI vs amount of Original content in a given text. 

We recommend not applying a similar “hard” rule when it comes to working with AI scores and writers. If a writer consistently has high Originality scores but then on one article has a higher AI score, this could very likely be a false positive and further investigation should occur. 

‍Here is our guidance on how to interpret the AI scores for writers.

Yes, Originality.AI’s AI is the only available AI that has been trained to detect if content has been paraphrased with a tool like Quillbot. If a piece of text, regardless of whether it started as AI or Human, is run through a paraphrasing tool like Quillbot Originality.AI will identify the content as AI generated 95% of the time. 

Read more about Paraphrase Plagiarism Checker

See a quick live test: https://youtu.be/OK5JnBtbbRo

The image below shows the detection accuracy for our Paraphrase Checker:

False positives in AI detection do occur and we are sorry if Originality.AI identified your content as AI. It is different for every tool and a specific reason on what the AI identified as causing it to predict the con

Here are 7 tips for minimizing false positives…

1. A detection score of 60% Original and 40% AI is not a false positive. It correctly predicted with 60% confidence that your content was Original. The meaning of the score is not that 60% was Original and 40% was AI it means it is 60% confident the content is Original.
2. Editing AI-written content is not a false positive - it is a true positive
3. Having AI edit your work is not a false positive - it is a true positive
4. When any amount of AI touches the content it can cause the entire article to be flagged as AI
5. “Cyborg” writing where a lot of AI tools are used to create an outline, suggest edits, optimize the content can increase the chance of a higher AI score. (For this reason we created a free content optimizer tool similar to SurferSEO or MarketMuse but 100% free that does not use AI to reduce the chance of a false positive occurring.)
6. Strange formatting can reduce the accuracy of the detector tools causing an increase in false positives or false negatives. 
7. The shorter the text the less accurate the detection score. We recommend at least 100 words being checked. 

We apologize if our tool has incorrectly identified your writing as AI!

• See our in depth guide on avoiding and managing AI detector false positives 
• Students - Make sure your teacher is educated on the current AI detection accuracy. Provide any details you have on the creation process. Using a Google Document and providing a screen recording. An AI detection score alone should not be used for academic integrity discipline. 
• Writers - Create your content in a Google Document and share the free Originality.AI chrome extension with your clients so that if there is any False Positive you can prove you created the document. Share how we recommend people use the AI detection score. AI detection is not the same as plagiarism checking and should have a more nuanced approach to what is and is not acceptable.    
• Agencies - The first thing any content agency needs to be doing is to define if it is acceptable for writers to use AI and if it is not then develop a process for checking. If you are choosing to use AI in producing content then make sure to communicate that transparently with clients. If you have clients that have accused your agency of having content produced by AI rather than human than refer to your company's policy/procedure for checking work and show the scan history of the writers work. 

A false positive is when human-written content is identified as AI-written text by an AI detector. False positives do occur and can cause a lot of pain. Across hundreds of thousands of tests we currently see false positives occurring about 2% of the time that human work is submitted. 

2% false positives, despite being the lowest in the industry based on our accuracy test, is too high.  

The AI researchers and machine learning engineers at Originality are working constantly to both improve detection accuracy and reduce false positives. 

We are heavily focused on equipping writers with tools to help them prove their contents Originality.

It was for the purpose of helping writers prove that their work was original that we developed a free Chrome Extension that allows anyone to visualise the creation process of a Google Document

Free AI Detection Chrome Extension

Originality.AI is the most accurate AI content detector producing the fewest false positives while also the only tool that accurately identifies a piece of content (human or AI) that has been paraphrased. 

See the complete study analyzing how Originality.AI matches up in detection capabilities to other AI detectors here - See Case Study.

The image below shows the results of a test of 1200 text samples, 600 AI generated and 600 human generated tested across multiple AI detectors:

‍

The table below shows the features Originality.AI has including…

• AI Detection
• Plagiarism Checking
• Paraphrase Detection (only tool to offer this)

Originality.AI is more than an AI detector. We are an innovative organization building out a comprehensive suite of content QA tools that will help you gain control over the quality of your content. Readability scores will be added in the near future. 

On the last OpenAI GPT-4 model Originality.AI was tested to be 99.37% accurate with 1.56% false positives. AI detection is different for every model. Below are the detection and false positive rates when testing Originality.AI:

• GPT-3.5 Detection Accuracy: 99.9% accurate, 2.1% false positivesx
• ChatGPT GPT4 Detection Accuracy: 83.29% accurate, 2.7% false positives
• GPT-4 Detection Accuracy: 99.5% accurate, 2.2% false positives
• Paraphrased (quillbot) Detection Accruacy: 94.7% accurate, 2.5% false positives

See our complete AI Detection Accuracy study to see how Originality.AI compares to other AI detectors. 

The image below is what is called a “confusion matrix” that is used to assess the accuracy of a AI prediction machine. Each prediction our AI makes on a test data set is grouped into one of four buckets…

• True Positive - AI detector correctly identified content as AI
• False Negative - AI detector incorrectly identified AI content as Human
• False Positive - AI detector incorrectly identified human content as AI
• True Negative - AI detector correctly identified human content as AI

The image below shows the confusion matrix for Originality tested on GPT-4 generated content. 

‍

AI-generated text is words that have been written by Artificial Intelligence (AI) specifically using Large Language Models (LLMs). The most common example is OpenAI’s ChatGPT which is an AI using OpenAI’s NLP Models (GPT-3, GPT-3.5 and GPT-4) to understand a users prompt and provide a written response.

For the purposes of AI text detection we aim to flag any content that has been meaningfully changed by AI. We try and take an approach consistent when defining what is Plagiarism “Presenting work or ideas from another source as your own.” when training our AI detection model. 

Slight editing with Grammarly or similar is NOT considered AI generated. 

For reference here is what we aim to identify as AI or Original…

• AI Generated and Not Edited =  AI Generated Text
• AI Generated and Human Edited = AI Generated Text
• AI Outline, Human Written and heavily AI Edited = AI Generated Text 
• AI Research and Human Written = Original Human-Generated
• Human Written and Edited with Grammarly = Original Human-Generated
• Human Written and Edited = Original Human-Generated