How is Google able to differentiate bw AI and Human content?
- Thread starter VAT
- Start date
- Jul 11, 2019
- 374
- 447
Google has shown no ability to distinguish between AI and human content. Especially well done AI content. If you think about it on a matter of scale, processing that much data would be prohibitively expensive right now. Not saying that aren't working towards it or will be able to in the future, but right now they don't seem to have the ability.
- Jul 11, 2019
- 374
- 447
That's just a sign of poor content, not AI content.
- Sep 7, 2017
- 7,187
- 6,435
If it's not any of those very obvious spun content then I believe they check the number of posts you are able to publish in a short while, there's no way one or even a team can publish thousands daily. Asides that, there's almost no way for them to accurately differentiate AI from human written content. After all, some human writers can produce absolute garbage content.
To chime in, that hugging ai detector tool detected some of my handwritten texts were ai apparently, and the ai texts with sensit.ai or Jarvis we're human figure that out. Writer.com ai detector has the same issues but is detected differently, never the same conclusion that makes sense with both detectors I know of.
Basically, I think if Google is detecting AI , most white hat sites will be in trouble given what I have seen just writing my text hah.
Basically, I think if Google is detecting AI , most white hat sites will be in trouble given what I have seen just writing my text hah.
You are comparing what a human can detect and what an algorithm that has to read all of the possible articles on the internet can.
Not exactly a similar comparison.
I’ve tested every software that’s out (so far) that I could find and either they suck at actually detecting the AI (saying that it’s real) or suck at detecting real content (content that I have personally written from scratch with my own experience they classify as fake).
Can Google produce better detection? Absolutely, but is it worth their time and money if the content is good quality and has good user metrics?
In regards to the original post my guess is Google will start to look at the amount of articles written/generated over a certain period of time and then do manual reviews if they deem it "too much".
Either that or they will do something in regards to sandboxing or not indexing content on newer sites or sites that havent been active in a certain period of time.
That's just my guess as it seems easier and cheaper than scanning every article and trying to determine if it is AI.
malayan
Power Member
- Aug 26, 2020
- 608
- 553
I'll take up this challenge:
#1
Once upon a time, there was a white and grey cat named Buttercup who always seemed to be in the wrong place at the wrong time. It didn't help that she was quite clumsy too, with an endearing habit of toppling off the kitchen counter during her explorations.
One morning, after taking a tumble from the countertops, Buttercup noticed something strange on the floor near her. It was an oddly inviting puddle of milk! She didn't know where it had come from – although later it turned out mom had spilled it when making breakfast – she dived right into investigating what it might be.
Being a curious kitty, Buttercup stuck her tongue out for a taste and discovered she liked it! With newfound enthusiasm, she began lapping up the milky treat with zest. Within minutes it had all gone – much to her disappointment. But just as Buttercup thought her delight had been swallowed up (literally), another puddle of milk appeared near her...
Mom came back into the kitchen to find Buttercup happily licking up what was left of the second spill, leaving nothing but an empty bowl in its wake. Judging by her guilty expression and her fur covered in splashes of creamy white liquid, Mom knew exactly why these little accidents kept happening to their feline friend... and yet again, poor Buttercup hadn't escaped unpunished!
#2
Once upon a time, there was a daringly adventurous housecat named Bart. He enjoyed exploring the world outside his home, yet he always found himself coming back to the kitchen in search of something delicious. On particularly rainy days, Bart relished going into the kitchen and finding himself a puddle of spilled milk left on the floor near the fridge.
One day, after another storm knocked out power to the neighborhood, Bart trodded into the kitchen and was ecstatic when he saw that some of his favorite wet liquid was scattered across the tiled floor! Without hesitation, he bent down to get closer to it and let out an excited "Meow!" His licks were gentle, yet swift as he slurped up every last drop of milk from the floor.
After such an exerting adventure, Bart realized that he hadn't just quenched his thirst but also attempted to clean up for his owners on top of that! In fact, there wasn't even a single drop left where it had been. Feeling quite proud of himself for such an accomplishment, Bart raised one paw in triumph before realizing what he had done – now where could he go to find more milk?
He thought hard and then finally concluded that nothing said satisfaction like a good bowl full of cold milk. So, with one giant leap toward the refrigerator door handle, Bart stretched out his arm as far as his four little paws would allow him before hitting a dead end again. After all that effort though, at least now he could be content knowing that there would be zero chance for anyone slipping against this unexpected hazard during their midnight snack runs!
debakid
Regular Member
- May 3, 2019
- 231
- 96
How is it originality ai able to detect AI content then ?
- Jan 17, 2015
- 5,263
- 4,955
The problem with those AI detectors is that require a ton of processing. Always remember that Google has to process not only millions of domains, but zillions of subpages under each domain, making this task extremely intensive. If Google has to check for spammy links, compare, index, cache, check WPO, check certificate, check for footprints, reverse DNS, translation copies, spinning copies, understanding if they are grammatically correct, analyzing spam inside the DOM, etc, etc etc... The power of processing required now with AI detector, their servers are going to blow in 1 hundred parts.
AI is the bane for Google nowadays, and most SEO experts are observing that content is rapidly decreasing in quality because of this. The concern with AI copy is that it will have a due date. And it's very difficult to keep it up. Unless you have the method to "rewrite" your copy with new algos now and then (+ also risking your rankings). So basically, AI is a double-edged sword, that, at least, is currently reporting some benefits.
Still more experimenting must be conduct by many members in this forum. Definitely, I bet that the word AI will be, by far, the most repeated word in this 2023.
debakid
Regular Member
- May 3, 2019
- 231
- 96
I see your point, but I think one day, they will come down hard on AI content sites.
- Jan 17, 2015
- 5,263
- 4,955
Don't forget that AI also perfects itself. GPT2, GPT3, GPT4... each iteration is always 1 step ahead of detectors. So when GPT3 detectors are cheap enough processing wise, GPT4 will be up and singing. When GPT4 detectors are fast, GPT5 will be a thing.
Be wary that this is going to take some extra factors by Google to keep their SERPS as clean as they can. They will probably focus in the super-volume keywords, leaving a blue ocean for the long tails, that is precisely what is happening now. I have a list of sites with Long Tail PAA ranking with millions of visits per month. Maybe they will risk that just 1 non-long tail keyword out stands, Google picks this on time, checks the low quality of the site with pure AI articles and shadow bans it.
So perhaps a combo of flying low + long tailing will be the future for undetectable AI. Time will say.
BetterSEO
Junior Member
- Aug 25, 2022
- 116
- 134
I tried comparing my written content and open ai Chat with https://writer.com/ai-content-detector/ and it actually detected the ai content every time. It also always showed my written content as human.
But it's when I combined both that tricked it into thinking it was all human.
But it's when I combined both that tricked it into thinking it was all human.
- Jul 11, 2019
- 374
- 447
No AI detector can do this cheaply, at scale, or with any reasonable degree of accuracy. Lots of false positives and/or false negatives from all of the ones I've tried.
