Helping you spot the AI-generated text

19 Signs of AI Content

1. Blatant AI Statements: Watch out for telltale AI utterances like “Regenerate response” or “as an AI language model, I”.
2. Flawless Spelling: AI language models don't make spelling mistakes or typos.
3. Neutral Opinions: AI content is usually neutral or noncommittal since AI models are designed to avoid causing offense.
4. Overuse of "crucial": The repetitive usage of this word is often a strong sign of AI authorship.
5. Overuse of certain verbs: Repeated use of verbs like "delve", "dive", and "discover" in calls-to-action can point at AI generation.
6. Frequent use of "unlock", "ensure": AI uses these words more often than humans do.
7. Overuse of "a dash of x": This is a colloquialism that you often only see AI use now.
8. Unhealthy dependence on demonstrative pronouns: An indicator of AI content.
9. Constant "reminders": AI uses apostrophe rhetorical devices quite a lot, starting sentences with "Remember,…”.
10. Repeated metaphor + tricolon): An AI's repetitive use of this rhetorical device is discernible.
11. Inability to construct anastrophes: Because AI pens text procedurally, one word after another, it isn't great at thinking ahead to form anastrophes.
12. Inability to commit catachresis: AI's quest for accuracy prevents it from breaking grammatical rules, as would often be necessary with catachresis.
13. Overused phrases: 'In today's world,' 'Take your x to the next level,' 'Master'
14. Fewer parentheses and em dashes: AI platforms typically don't use these punctuation marks as often as human writers.
15. Overfriendly tone: AI often starts sentences with "If you're anything like me,"

https://www.linkedin.com/pulse/identifying-ai-authorship-set-19-signs-your-content-penned-rajat-jain

 

I might be an AI.

Ok, nevermind

 

You just had to delve into it a little further.

 

Oh no! I subscribe to The Economist. They have excellent articles. I was just reading an article about swing voters in the American election this year and came across this sentence. There it is: delve. Should I be worried?

 

What about dodge, duck, dip, dive, and dodge? Are they still safe?

 

If it delves like a duck...

 

I've noticed that when I try asking certain sites for a little help describing something, the AI generated text gets super... I guess you'd call it purple prosy. Like, I'm trying to insert a little humor about a character wanting to pocket some treasure while another character is noting that they need permission from the king before they take any treasure and I check to see if it can make it a little funnier and instead it goes on this long tirade about how this character sending shivers up another character's spine and after that I stopped reading because it was just an uninteresting paragraph of word vomit. No thanks, I'd rather struggle with describing things than have my work be an undecipherable pile of words that kill off the reader's brain cells.

Furthermore, is it anyway or anyways? The grammar corrector keeps saying anyway but every time I've heard someone say it they said, "Anyways."

 

Anyway in the text, and anyways in the dialogue?

 

Don't try to rely on a handful of single words. You can generally get a pretty good feel for if something is written by AI from the overall-ness of it. The blandness and inanity, the generic-ness, the word-vomit, as @Starcatcher said. I usually get a pretty strong gut feeling. And if it isn't AI then it's a pretty bland and uninspired writer with no connection to their inner feelings or thoughts, just generic surface-level stuff.

When I said above that "I might be an AI," it wasn't an admisstion of fact, but an indictment of the method the writer was trying to use. I don't think a few single words are any kind of reliable test. I frequenlty use some of those words (which is what I was saying). The test itself is a fail.

 

A bit of both. I don't think I've ever heard anyone say anyway, that just sounds grammatically incorrect.

 

Anway is the right one. Anways is a colloquialism. Keep in mind most people aren't writers and don't know grammar very well, they just talk the way people generally talk. A lot of things most people say are technically wrong. That's the point of threads like Do you enjoy eggcorns?

 

Colloquialisms add flavor. Confusion, too, sometimes. I started to reread Stevenson's Kidnapped a few nights ago and had to keep looking up slang words.

Definitions can be imprecise, too. I've lived in rural Texas all my life and met only one person who used "ary" in conversation. I think East Texas may be the western extent of its usage, rare around here.

I still don't know exactly how to use it in a sentence. It's a version of any or anyone. "I don't think ary could help" is probably correct. I guess.

Ain't no telling. Slang is not my métier.

 

I asked ChatGPT: How can I tell if something was written by AI?

ChatGPT responded:

Detecting whether something was written by an AI can be challenging, especially with the advancements in natural language processing technology. However, here are some indicators that might suggest a piece of writing was generated by AI:

1. Consistency: AI-generated content tends to be consistent in style, tone, and grammar throughout the entire text, even across lengthy passages.
   
   
2. Lack of errors: AI-generated content often lacks the typographical errors, grammatical mistakes, or inconsistencies commonly found in human writing. However, this can vary depending on the quality of the AI model and the complexity of the task.
   
   
3. Unusual or unnatural phrasing: Sometimes, AI-generated text can contain phrases or sentences that seem slightly off or unnatural to a human reader. This could be due to the model's training data or the limitations of its understanding of context.
   
   
4. Overly complex or technical language: AI models trained on technical or specialized datasets might produce content that is overly complex or filled with jargon, which can sometimes be a clue that it was generated by AI.
   
   
5. Unusual topic transitions or connections: AI-generated content might sometimes jump between topics or make connections that seem unusual or unrelated to a human reader.
   
   
6. Contextual inconsistencies: While AI models are improving in understanding context, they can still sometimes produce content with inconsistencies or inaccuracies, particularly when generating longer passages of text.
   

 

News from big media may be a good place to hone AI spotting skills. Silly errors that no human with the slightest awareness would have overlooked seem to be more commonplace.

 

Without putting value judgements on the writing there is no way to tell if something was written using AI. Sometimes you can tell but there are plenty of times you can’t.

And as the technology improves the harder it will be to tell what is and isn’t AI writing.

Ironically i think AI technology that can be used to detect the use of AI technology will be worth billions in the next few years (or worth nothing it society decides they don’t care)

 

Sorry, I don't understand what you mean here. Isn't it value judgements that are the determining factor?

 

I don’t know that’s for people to decide.

But without those value judgements some AI writing (not all) is structurally fine. AI can write a story that makes sense it might not be “good” but it is a legitimate “story”

 

I set my sights higher when I am writing a story.

 

If you’re on this site I would hope so. But that’s not my point. Thank you.