False Positive Content
AI content detection tools use the dataset of existing content to predict the most likely words and phrases used for the related topic. The more predictable the piece of text is, the more likely it is AI writing. At least, that’s how the program scores the text. Therefore, your text can be detected as a false positive.
If you haven’t used any AI tools, but your text shows a high percentage of AI, we recommend you do the following:
If it is higher than 20%, review the highlighted sentences, consider why the detector highlighted them, and edit the necessary sentences.
If it is lower than 20%, explain to the customer that some sentences are underlined as false positives.
AI Red Flags
In order to edit the text falsely marked as generated by AI easily, you need to know the red flags of artificial intelligence, i.e., those features characteristic of Chat GPT and other similar programs. The main red flags relate to the paper's most important components, namely structure, content, grammar, and style, so we will consider them in that order.
Structure
Uneven length of paragraphs.
Absence of topic and concluding sentences.
A thesis statement is not restated in the conclusion.
Content
Poor Thesis Statement.
Confusing Content.
Grammar
Broken sentence structure.
Overused can+verb and lead to.
Style
Repetitive transitional words such as Additionally, In addition, Moreover, Furthermore, On the other hand, etc.
Repetitive phrases like it is essential/crucial/critical/central/vital etc.
Frequent repetitions.
List of Words and Phrases Often Used by AI
It’s important to note | Bustling | Take a dive into |
Delve into | In summary or In conclusion | Navigating i.e. "Navigating the landscape" "Navigating the complexities of" |
Tapestry | Remember that | Landscape, i.e., "The landscape of...." |
Testament, i.e., "a testament to..." | In the world of | Realm |
Embark | Unveil | Unleash |
Do This When Your Text Is Detected as AI
Rephrase sentence structure and format.
Use synonyms and similar words.
Use active voice.
Avoid jargon and complex terms (as long as they meet the requirements).
Share personal perspectives
Although AI is useful in many cases, it often harms academic writers more. Thus, avoid using AI and always check your texts on several checkers to prevent misunderstandings with clients, revisions, and cancellations.
TOP AI Checkers We Use and Recommend
Checkers + links | Advantages | Disadvantages |
A Reliable One |
| Shows the percentage probability for AI only for highlighted areas, but not for the text in general. |
Turnitin It's a basic checker for us, although it's not perfect |
| It's a base, although it's not perfect |
An Ok Tool | Highlights the suspicious content in yellow | Do not show a percentage of AI content. Has a limitation on free check - 5000 characters. |
The design of the results and the algorithm of detection is similar to GPT-2 Output Detector. | Registration is needed. Does not highlight the suspicious parts. | |
a reliable fee-based tool |
| The number of credits is limited. It shows a human score, not an AI probability, which may be confusing. |
BrandWell (former ContentatScale) a reliable fee-based tool |
| It does not show percentage results, but Reads like AI, Hard to Tell, etc. |
| The free version of the tool allows you to check 1000 words. It detects well texts which are either 100% AI-generated or 100% human. | |
| The free version allows to check high volumes of text. Checker offers a high accuracy of results. | |
| The free version allows only 1200 words for check. | |
| Some limitations on free checks exist, but they're not critical—15,000 symbols may be checked for free. | |
| Tricky checker with high probability of false-positive results, but many professors rely on this checker, so its results should be considered as well. |
Presentation
More detailed information and examples you will find in this presentation.