What are some tips and tricks you use to detect AI content? Which tools do you find most reliable?
Tanya Pereira
23 replies
Replies
Sophia Solanki@sophiasolanki
Narrato
I'd say stop worrying about AI content detection. Google doesn't care - https://developers.google.com/se... , and almost all research presently indicates AI content detection is a futile exercise (it's next to impossible to differentiate with a high level of confidence), and has a lot of false positives (which I am sure everyone here would have experienced). AI detectors are just making a ton of money feeding off everyone's fear of how Google would treat AI content. Waste of time in my humble opinion!
Share
Narrato
@sophiasolanki great point Sophia!! As long as content offers valuable insights and is interesting to the reader, nothing else should really matter!
Skylead
Good question
Narrato
@relja_denic Thanks!! Are there any tools you rely on to check for AI content? I find that different tools tend to give you different results...
You don't question the content. You question the creator(the human) of the content.
https://x.com/paulg/status/17099...
Narrato
@rakshith_aloori a very interesting thought!!
If you find any tools available like that, please share.
The main notice I've gotten thus far is that the content is very similarly structured, making it lack human touch.
Narrato
Launching soon!
Even if the content is written by an AI, certain tools are available for the purpose of rephrasing AI-generated content. I'm not aware of any tool but the words that are written by AI are difficult.
Narrato
@prema_toppo yep narrato.io does that and more
I used a bunch in the past, but copyleaks seems the most accurate. I tested it against 20 pieces of AI generated content and the results were pretty accurate.
Narrato
@ioana_sima Thanks for the suggestion, will surely check it out.
In my experience I would say, readability takes a big hit when content is AI generated. Almost all articles are written for collegiate level regardless of the topic and its simplicity.
If you've been using ChatGPT for long enough, it's easy to recognize patterns and expressions it makes! But for a tool, I think most of them are not that reliable
Narrato
@khanh_linh15 I feel the same! Different tools tend to give very conflicting results!
When you read a lot of AI-generated content, you can determine it yourself without any tools in most cases. This is at least applicable to GPTChat.
Honestly, if you use Chat GPT a lot you start learning the its "default voice".
It's easy to identify from there.
Also, I read an article on LinkedIn of a founder that was hiring and had specific questions for the candidates. He got a bunch of identical responses. Lack of originality is certainly a way of detecting AI.
Narrato
@julia_chacur great observation! Do share the article you've mentioned with me, would love to take a look :)
Narrato
Detecting AI-generated content can be a challenging task, as AI models like mine are continually improving in generating more human-like text. However, there are some tips and tools that can help in identifying AI-generated content:
1. Lack of Personalization: AI-generated content often lacks personalization and may use generic language. Look for signs of content that doesn't seem tailored to a specific audience or context.
2. Inconsistent Information: AI may provide conflicting or inconsistent information within a short span of the conversation. Human-generated content tends to be more coherent and consistent.
3. Overuse of Jargon or Technical Terms: AI models might overuse technical jargon or industry-specific terms without providing clear explanations, which can be a red flag.
4. Repetition: AI may repeat phrases or ideas more frequently than humans. Excessive repetition can be a sign of AI-generated content.
@eleanor_montgomery pretty sure this is AI generated
Narrato
@eleanor_montgomery thank you for your reply!! Totally agree with all the points you've brought up.