Is A.I. Content Causing SEO Indexation Problems?

[welski]

Most people don't understand that AI tools use different AI models. A lot of AI sites fail because they probably use tools that are derived from lower tier AI models so the patterns are obvious or the output is simply generic. For example Frase uses GPT-J which is in the low tier trained with a mere 6billion data parameters

There are plenty of large AI models trained with hundreds of billions of data parameter(Soon Trillions) that is so smart it can code on its own or even generate convincing images just base from text.

Can't wait for GPT-4 next

 

[Barry]

I just received an article back from an article writing service that I've been using for a couple of years without issue, but this latest order is full of AI-generated content.

I'm seriously pissed!

Online publishers who hire writers now have a new problem to deal with, because some freelance writers are now using these AI tools to generate more content faster, to make more money. That means we have to waste time checking over every article to ensure we're not being scammed.

I don't want auto-generated Ai content on my sites - I want properly researched content with character, written by a human - there's a huge difference.

Anyone who thinks Google doesn't care about their indexed being pumped full of auto AI-generated content is mistaken.

There are already free tools out there that can detect this stuff - just think about what Google could do with the resources they have when they decide to tackle this problem.

Even if AI content is ranking and evolves to avoid detection, Google can turn to their age-old tactic of turning the dial up on big brand authority, making the ranking landscape harder for all of us.

 

[Ryuzaki]

Word on da skreet is that People Also Ask (PAA) is back to the SERPs nearly to the same level it was before it dropped off:

Rank Ranger

Semrush

​

We could craft all kinds of theories about how this wasn't a way to target AI content abusers, or it was and they fixed something and felt safe rolling PAA back out, but the truth is we just don't know. Could have been a bug, even.

​

 

[Taq]

I just received an article back from an article writing service that I've been using for a couple of years without issue, but this latest order is full of AI-generated content.

I'm seriously pissed!

Online publishers who hire writers now have a new problem to deal with, because some freelance writers are now using these AI tools to generate more content faster, to make more money. That means we have to waste time checking over every article to ensure we're not being scammed.

I don't want auto-generated Ai content on my sites - I want properly researched content with character, written by a human - there's a huge difference.

Anyone who thinks Google doesn't care about their indexed being pumped full of auto AI-generated content is mistaken.

There are already free tools out there that can detect this stuff - just think about what Google could do with the resources they have when they decide to tackle this problem.

Even if AI content is ranking and evolves to avoid detection, Google can turn to their age-old tactic of turning the dial up on big brand authority, making the ranking landscape harder for all of us.

This is a concern of mine when hiring writers. However, I do not know any accurate tools to detect it

 

[Jake N Elwood]

I know a guy whose content is also AI and his content is far far worse than mine. How come he's ranking fine? So many others are ranking fine with their Jasper content? (I could send you their sites if you want, aglowlifestyle com, this is Emily Dyson's website, she reveals it publically in her newsletter and twitter.)

But okay maybe in time they will get deindex sure. But, how come the articles which are written by garbage human writers still ranking whereas the AI articles seem to be of a much higher quality aren't??

Also, if the article gives helpful advice to the reader what's the issue? I don't see any issues if the article is good enough for the reader. I understand if there are other articles that are more useful ranking but if there are none and my articles are providing value to those who read them, why should they get deindexed??

FYI That AI Detector, I put my articles in and it couldn't detect if it was AI or not. I got like 0.08% fake.

Hugginface is a crap detector, there are a couple of GPT3 ones out there, much better.

 

[Google PE]

Hello. Interesting discussion.

From my own experience, I can say the following.

1) Google normally indexes artificial intelligence-generated content.

[broken image]

2) Google normally ranks such content.

[broken image]

These are screenshots from a blog where 99% of the content was created using GPT-2 and GPT-3.

 

[Ryuzaki]

@Google PE, your other images were 404-ing with easycaptures.com. I recommend using Imgur for reliable 3rd party image hosting.

Anyways, of the one image that's viewable, I see it's a DR49 site, but after 6 months you're getting 56 organic traffic per month and it's 80% from India. I'm not trying to be rude (honestly, especially since we couldn't see your other two images) but I'm having a hard time calling that a success or proof that Google ranks this content. We do know that Google ranks the content for a stretch of time until they detect it and trash the site.

How many articles are on this DR49 site with 56 monthly traffic after 6 months?

 

[wikibum]

Anyways, of the one image that's viewable, I see it's a DR49 site, but after 6 months you're getting 56 organic traffic per month and it's 80% from India. I'm not trying to be rude (honestly, especially since we couldn't see your other two images) but I'm having a hard time calling that a success or proof that Google ranks this content. We do know that Google ranks the content for a stretch of time until they detect it and trash the site.

I was going to say the same thing but didn't want to be rude. 50 visitors per day is not statistically significant to make a conclusion that AI content is not a bad longterm investment, especially with traffic mostly from india.

 

[CCarter]

50 visitors per day is not statistically significant to make a conclusion that AI content is not a bad longterm investment

I dunno I think this is pretty conclusive results that allow us to state A.I. content is a complete waste of time.

If in 6 months that's the best evidence... that the EXACT type of results I expected to see from these A.I. content operations.

I don't know why so many people are pretending.

AND then the question is - do you guys even care about the visitor or solving their problem? What are we doing here guys?

 

[Google PE]

@Google PE, your other images were 404-ing with easycaptures.com. I recommend using Imgur for reliable 3rd party image hosting.

Anyways, of the one image that's viewable, I see it's a DR49 site, but after 6 months you're getting 56 organic traffic per month and it's 80% from India. I'm not trying to be rude (honestly, especially since we couldn't see your other two images) but I'm having a hard time calling that a success or proof that Google ranks this content. We do know that Google ranks the content for a stretch of time until they detect it and trash the site.

How many articles are on this DR49 site with 56 monthly traffic after 6 months?

This is actually an experimental blog set up to test whether Google will index all AI-generated articles.
There, articles are published from time to time, but no one is engaged in search engine optimization.
All published articles are indexed. This means that Google has no particular claim to the quality of content, despite the fact that it was created by AI.

Many people believe that Google does not index AI-generated content. I just gave you an example to show that this is not the case.

 

[Ryuzaki]

Many people believe that Google does not index AI-generated content. I just gave you an example to show that this is not the case.

How many articles have they indexed on the site?

I'm with you. They will index it, no question about it. But I've also seen that they'll eventually de-index it, especially if it starts to get a little bit ahead of its skis.

This was an example from earlier in the thread:

​

This seems to be the fate of most I've seen. I've actually got notes in my "Notes File" of other AI sites that I'm watching, waiting to see when Google de-indexes them.

Seems like that one guy from Reddit has the right idea. These should serve as a testing ground for finding achievable keywords for real sites. In the meantime, they can earn a little money and you could certainly churn and burn them. I suspect if you could trick a high RPM ad network into running ads on the site, or just use Amazon Associates before they ban you, you could make some more-than-decent money too. But the end result is burnt bridges with these networks.

I'm seriously pissed!

Hugginface is a crap detector

I played with it quite a bit the other day and my ultimate conclusion was that it was trash.

I could take content that I know wasn't AI because I just wrote it into the prompt myself, and it'd think it was AI. I could take three paragraphs and it thought they were real. I could remove one and it said with full certainty they were AI. I could rearrange their order and they'd jump from real to fake with the exact same text, well over like 500 tokens (way past the 50 they say to have at least). The thing was garbage, and it's for GPT-2 anyways and we're well past that now with GPT-3 and other models.

I'm trying to find the quote and can't but the people that made that detector flat out said that their own conclusion was that it's unreliable and people should "manually" read the text and use their own judgement on whether or not they're looking at AI text. Complete waste of time.

 

[wikibum]

I dunno I think this is pretty conclusive results that allow us to state A.I. content is a complete waste of time.

If in 6 months that's the best evidence... that the EXACT type of results I expected to see from these A.I. content operations.

I agree. AI content website will fail in many ways than one. Even if you get to rank it, a user reading the post WILL NOT CONVERT. Only play I see is getting some display ads running, making some money and then do it all over again. To me that is a waste of life.

We are trying to build long term assets here. This is not BHW.

What about this thread. https://www.buildersociety.com/threads/how-did-this-site-get-so-much-traffic-this-fast.6250/. This website was spamming AI content for a longtime and ranking amazingly. Dude probably made a ton of money in a few months, but like all blackhat stuff, it will eventually go to shit.

@LinkPlate already provided an example of an AI website (in this same thread) that went to shit. There are many others, but this one stands out because of the massive amount of content it had and the frequency it was posting. It's really obvious its AI content.

This is actually an experimental blog set up to test whether Google will index all AI-generated articles.

Google indexing content doesn't really mean anything. If you are not in the top 10 for your keyword - it means absolutely nothing.

 

[thisishatred]

This seems to be the fate of most I've seen. I've actually got notes in my "Notes File" of other AI sites that I'm watching, waiting to see when Google de-indexes them.

+1 for this.

Similar to Ryu, I have a number of these I track as well just to see if things repeat themselves.

Another example: https://askinglot.com/
Was huge for a while, appears to be manually deindexed now using site:https://askinglot.com/

Couple more examples:
site:answerstoall.com (this is a 301 to answer-to-all.com, let's see how fast this one gets deindexed)
site:everythingwhat.com (301d to https://contextresponse.com/ - already deindexed)

I'm down if folks wanna waste their time with this shit though.

 

[CCarter]

https://twitter.com/x/status/1554311325039337472

Doing the math, if this girl simply does 2 deliveries a day, she makes more money that people using A.I. content and journaling here in the lab. Lots of pageviews that result in less money than a Dominos employee.

How is this not a waste of time??

What baffles me about this topic of approach is... Why? What are you guys gaining, cause it's not money nor time.

What's going on? I feel like I woke up in the twilight zone and everyone is insane. I don't get it.

 

[zentrader99]

I just received an article back from an article writing service that I've been using for a couple of years without issue, but this latest order is full of AI-generated content.

Which tool did you use to get this 90% fake result?

 

[Ryuzaki]

Which tool did you use to get this 90% fake result?

That’s the HuggingFace GPT-2 Detector. It sucks and even the creators don’t recommend it be used or trusted for real application.

 

[zentrader99]

That’s the HuggingFace GPT-2 Detector. It sucks and even the creators don’t recommend it be used or trusted for real application.

I tried this tool for Copy.ai, Copysmith and Articleforge for long-form content/bullet point expander. I even proofread the content and most passed Copyscape without any issue. Even though it looked quite good for a real human reader. As a control, I used several top-ranking sites and my own content.

Almost 80-90% of the time, Huggingface detects AI-generated content successfully. Sometimes, it gives 40-50% fake ratings to real content but is more likely to be right than wrong.

Sample Output from one of the above tool.

Zoosk and eHarmony are both online dating sites, but which one is better?

Zoosk is a free online dating site that makes it easy for you to meet singles in your area. It's fun and easy to use, with user-friendly search tools and easy-to-read profiles. Zoosk also has a clean, attractive design that makes it feel like a central location for online dating.

eHarmony is an online dating service that specializes in helping people find long-term relationships based on compatibility. The site offers several different features designed to help you match with potential partners based on shared interests and values, providing more than just a photo of your date's face when they've logged in.

In general, Zoosk has more features than eHarmony, although eHarmony's features are more advanced than Zoosk's. However, Zoosk is easier to navigate because of its simple interface; eHarmony has too many options for people who aren't looking for a long-term commitment.

99.97% Fake

 

[Ryuzaki]

Somewhere on the forum I commented that I saved 4 AI sites that I had come across with the intention of seeing how they fare after a few updates. Well we had 3 updates within 2 months and here's the update:

DR47 - 6.8m to 23k organic traffic​

Deindexed completely it seems if you look on Google for it. Seems like they restored content from the original domain on a new crap theme and are experimenting to see if it can still be useful at some point.

You can see that the traffic tanked completely from almost 7 million to nothing, and the indexed pages started dropping out, too.

DR0.5 - crashing and burning themselves​

These guys know the end is nigh, so they're now filled with popups and overlays and those scripts where if you click the window 5 other windows open asking you to install spam software and all that crap. It's interesting to see their indexation counts wavering.

I guess they're trying to suck up any money they can since they didn't lift-off like others did. They also seem to be skating by because they don't get as much traffic and don't have a backlink profile to write home about:

DR48 - 2m to 473k organic traffic​

Here's another site that was doing well and still is for the time being, but is definitely not doing so hot.

Their indexation is staying up but the traffic is tanking off a cliff.

DR32 - 1.8m to 600k​

You can see this fairly big one is getting the same treatment. Indexation is starting to drop out as traffic tanks.

Here's the thing with all this. One of these managed to sneak into Mediavine (they and the exchanges dropped the ball there) while the others are having to make due with Adsense and Ezoic. The latter ones likely weren't making enough money to make much of a dent, though the Mediavine one probably did pretty well while it lasted, even though the RPMs were likely in the toilet.

These sites have a few things in common: zero effort. It's a barebones theme slapped up with the footer credit links in place, a crappy logo added, and zero images in any of the articles.

It seems to me that the Helpful Content Update started to tag these sites but didn't do much damage, and then the September Core Update decimated them. Which is very interesting. There's zero knowing what really went on behind the scenes with all these updates overlapping.

Anyways, it'll be fun to watch this war continue, because the AI content users smell blood in the water now. The window of opportunity wasn't open very long and only a handful of operators seem to have taken a real shot at it. But now lots of people will be clamoring to make it work and since it's all public and covering so much SERP real estate, nothing will stay secret long. Both sides will have to escalate the war and evolve. We know who ends up winning but there should be a lot to learn by paying attention.

 

[JOoa0ky]

@Ryuzaki fabulous data and observations there.

Yeah... I think it's safer to just take it slow and steady to win this type of race since G is hiding in the bushes with it's speed radar.

 

[Boy]

Why are we pretending these AI mass-generated sites are meant to be long-term operations, not just churn and burn? They're just mass-produced organic traffic leaks meant to make as much money in as little time then they're gone. Get 6 million visitors in a month or 4, snag $1 CPM, take that $20k, buy 1000 more domains, and run the script on all those sites.

 

[CCarter]

long-term operations

Cause this isn't BlackHat World. You don't need the Digital Strategy Crash Course nor need to create 3 customer profiles for market research for blackhat projects that are designed to tank eventually.

90% of the advice and topic we talk about HERE are for long term projects with a majority of them turned into successful 7-8 figure exits after years of investments.

People that recommend A.I. content aren't stating that last part where the projects are designed to eventually tank. So a ton of people waste time and months/years on these projects because they took advice designed for short-term gains versus long-term, and then get hit with an update and come running to us in private messages asking what's wrong their website. Or since several of us sell links we get approached for consults and end up wasting time giving people advice only to find out "oh yeah this content is A.I." ¯\_(ツ)_/¯

You got guys like @freshpeppermint that wanted to "prove me wrong" and in turn wasted 7 months of his life on this nonsense and ended up saying he was un-successful.

I'm all for making a ton of cash using darker arts then going and building something not in the shadows, aka startup funds.

It's all about the end goal. If your end goal is a long-term exit A.I. content isn't the way to go. If you are looking for quick fast cash, then there might be something to this blackhat strategy.

 

[Boy]

Cause this isn't BlackHat World. You don't need the Digital Strategy Crash Course nor need to create 3 customer profiles for market research for blackhat projects that are designed to tank eventually.

90% of the advice and topic we talk about HERE are for long term projects with a majority of them turned into successful 7-8 figure exits after years of investments.

People that recommend A.I. content aren't stating that last part where the projects are designed to eventually tank. So a ton of people waste time and months/years on these projects because they took advice designed for short-term gains versus long-term, and then get hit with an update and come running to us in private messages asking what's wrong their website. Or since several of us sell links we get approached for consults and end up wasting time giving people advice only to find out "oh yeah this content is A.I." ¯\_(ツ)_/¯

You got guys like @freshpeppermint that wanted to "prove me wrong" and in turn wasted 7 months of his life on this nonsense and ended up saying he was un-successful.

I'm all for making a ton of cash using darker arts then going and building something not in the shadows, aka startup funds.

It's all about the end goal. If your end goal is a long-term exit A.I. content isn't the way to go. If you are looking for quick fast cash, then there might be something to this blackhat strategy.

I'm not talking about what we typically do here, I'm talking about the sites that @Ryuzaki shared.

But I think I misinterpreted what the post served as. I think it was the first thing I read when I hopped on the computer.

I read it initially as 'these 4 sites got hammered' (and I thought, well there are thousands more that can be created), whereas I think the intent was 'if you use AI, you're probably going to get hammered eventually.'

My bad!

 

[Ryuzaki]

I think the intent was 'if you use AI, you're probably going to get hammered eventually.'

That and just “keeping our finger on the pulse of the industry” in general.

As @CCarter is saying, there are way more people who see this as a viable way of overcoming the content production hurdle than there should be. They don’t have the time in the trenches to understand that this is another spam game that will end with them having wasted a lot of time.

And ultimately to make it work for any small amount of time will require going above and beyond with images, formatting, on-page… and that’s to survive for 6 months. If you’re going to do all that, might as well do it with real content. Because at the end of the day Google will figure this out enough that the sites never really lift-off at all without some extra effort better spent elsewhere. They will sort it out.

Google doesn’t really care if we manipulate it as long as what we publish is helpful to users and we aren’t doing crazy link schemes. We pretty much have the green light on going balls to the wall with SEO and making gobs of money. It just ends up resembling an actual business to pull it off and a lot of people resist that for a lot of reasons, who could otherwise succeed and change the trajectory of their lives for the better. I hate to see well-intentioned people get seduced by spam because nobody told them the truth.

 

[CCarter]

People Also Ask (PAA)

About 10 days ago, Google has implemented new code that goes through GREAT lengths to hide PAA from scrapers/crawlers of their search results.

It's interesting they are hiding specifically the PAA data behind additional AJAX calls and really nothing else - not even rich data or anything.

The theory is starting to hold up that people were utilizing PAA questions and then getting A.I. bots to create content to answer those questions.

- "Reporting live from the deep trenches of the digital mines"

 

[voLdie]

I have a stable of writers at this point and every week I am having to fire a decent writer because they try to slide in some AI shit content after a few months.

Finding an editor has been challenging but I have learned you have to check every piece of content to some degree.

Writers get lazy and I am sure getting slammed with Jasper / AI ads all day.

Gay.