Scaled Content Abuse: How to Scale AI Content Without Losing Rankings (with Koray Tuğberk Gübür)

James Dooley: Scaled content abuse. There's a lot of people now scaling out a lot of content, a lot of blog posts, a lot of articles. They believe they need to kind of get that topical authority. And there's no one better to be talking about this than Koray himself, the inventor of topical authority. Thanks so much. So before we get started, can you just quickly explain what topical authority is?

Koray Tugberk Gubur: Technically, when you cover a topic for longer duration of time with lower cost for the search engine, search engine realises that they can actually rank you for all these group of queries and save money for themselves. And rather than checking 10,000 other sites, they just trust one site for one topic and they keep tolerating actually your mistakes or they define the entities or consensus based on you. So internally in our office, we've got a formula on the whiteboard and it says topical authority equals topical coverage multiplied by historical data divided by cost of information retrieval. And we're going to be talking about here now the scaled content abuse. And these people are now creating 5,000 articles, 500 articles, 800 articles. And they're doing the query templates and they think they're writing good quality content. There's no historical data being done with it. Can you give some examples on this and first of all, firstly, can you explain what the scaled content abuse actually is?

James Dooley: Of course. So according to Google, if a page doesn't have any human effort, they don't define it as quality, even if the content is perfect. They want to see the human effort very clearly there. And there are a couple of very quick ways of seeing whether a page is coming from a human or just a machine, which is about publication frequency. If you publish 5,000 pages a day, it's certainly, it's not coming from the humans. It's only coming from the machines. And in Google's actually ranking system, there is always a hierarchy of algorithms. If one of the algorithms don't or doesn't approve your content, they won't be checking the other steps anyway. It will be just waste of money again for them. And if they see that actually there is no human effort and you publish 5,000 pages in a day, since actually human effort can't be found, you won't be even considered as a candidate to rank in that area. The interesting thing is recently we see that even if you have really huge PageRank or, let's say, tens of thousands of referring domains in a quality and natural way, and even if you have a really huge brand search demand, millions of customers, the moment that you actually trigger this section of the algorithm, they just start dropping you. It looks like a very strong decision tree, especially since December 2025 core update. I mean, there is a concept that we use, search engine tolerance, and we always actually see and think that the tolerance for these type of brands will be huge. We would always consider them, they can do a little bit spam and they can get away with it. But it looks like if you are hitting this classifier mainly, it becomes a major problem. When I created the first AI sites in 2021 through the expired domains, one of them was BlackTailNYC, for instance. It was an expired domain from a bar in New York City. We turned it into the cocktail site and we came to the 8,000 clicks a day with human outsource. Then we came to the 65,000 clicks a day with just the AI. We got a manual penalty. We removed the AI content. They removed the manual penalty. We continue from 7,000 again. This was 2021. And maybe you remember in the beginning we were talking about momentum. We were saying that if you publish more, you will rank better. Today, rather than publishing more, publishing with human possible range is more important. If you publish so fast, so high momentum, they just classify you as basically lack of human effort or lack of originality and it harms domains. And we can maybe give the domain name, they can people check BrowserStack.com. It has 23,000 referring domain. It has 200,000 brand search demand. It has thousands, tens of thousands of customers, but technically they lost 70% of traffic in the eight months because they publish close to 2,000 AI-generated content in one day. And also there are the web... Majority of the site is also outdated or very stale too. You can create the sitemap and you can actually observe this by yourself.

Koray Tugberk Gubur: So with regards to scaled content abuse, you've mentioned there quite a few times about the frequency and the article velocity of what's being done, right? Do you, do you think it's that more than the human element... You, you also mentioned that the content should have some sort of human element to it? That's two different things. Or is it both just as important as each other?

James Dooley: It needs the human element. Like, what happens if I was to do the article and it did have a human element and I scaled 5,000 articles? Would that still get hit even though there was a human element?

Koray Tugberk Gubur: I believe even if you create content with just the humans, as long as you publish 5,000 pages in a day, you will still get classified by AI because there is not a certain absolute way of knowing which text is coming from a human or the machine. Mathematically not possible. And Google has some services that they claim, but maybe for images and the videos or the audio, that's easier. But when it comes to text, mathematically it's possible. Even there was a theory by the scientists because scientists usually they mock, uh, the writers or novelists, and they were claiming that even there is a possibility that an African chimpanzee can write Hamlet as Shakespeare did because it's mathematically possible. Chimpanzee will press these things, maybe after 1 billion trial he or the chimpanzee will be able to actually write the Hamlet. So mathematically, even a chimpanzee can create better content than human. So it's that level of, uh, chaotic math calculations. There is not absolute way. But I believe Google will try to find a kinds of safe border for themselves basically because we have seen some sites got manual penalty because of scaled content, but they are created in 2017. Their content is actually way earlier than AI actually came for, but still they kept the sites for six months in the manual penalty state by claiming that content is coming from AI. So meaning that Google doesn't actually care about absolute labelling. They just try to sense with the statistics what you are doing there. But technically momentum there is a stronger classifier.

James Dooley: So with regards to scaled content abuse there then, and I know the answer on this would be it depends, and I don't want... I want you to almost give a range, right? How much is too much on a daily basis? So you mentioned 5,000 being too much, but, like, I don't know whether you're saying hundreds too much, 50 is too much, 20 is too much. Like, what sort of range would it be? And I know you might say it depends on the size of the existing site and the brand and stuff like that. But for anyone who's watching this, what would you say the range should be on a daily basis?

Koray Tugberk Gubur: For instance, right now I'm creating a subfolder just for the citrines. Uh, citrine basically is a wheat type basically, like let's say gum set, a special type of citrate, and there are basically, like, uh, 5,000 types of citrines, meaning I will be creating 5,000 pages eventually. And in that case, I started actually from like five a week. Then I actually increase it to the 12, then 17, then 19, then 25, then 27. Then I am right now in the 30s. Technically, I'm trying to also find the balance with the business because if I keep everything low momentum until I publish all 5,000, like two years will pass. We don't have that type of time. So, we need to go faster as well. In that case, I'm actually keep changing my content by changing my structured prompt, which I can also show you later as well. And then after changing my prompt in the unique way, in the changed way, I publish another 25 and 30. I keep increasing it gradually. And if I see that after I publish 40, I don't get this classification, I don't get the indexation, I still understand that this is actually safe range and I go for 45, 50, 55, and etc. Another important thing is not stopping next day to the zero or next week, meaning that actually I'm increasing gradually, I stay in the maximum level and I decrease gradually as well while updating the older pages too. If you keep updating and if you keep staying in this, let's say, autocontrol state, you will always find a balance. But there is a site that you can check, or aai.com, which is an AI writer. They published like 4 million pages in 60 days and they rank really well in the beginning, but eventually, of course, they drop to the zero, uh, as well. So I believe there you should be finding this testing state. If the site is still being crawled, indexed and getting new queries, and if you do this week over week, it will be better. I usually try to grow my subfolder or site at least, like, three months old by keep publishing slowly. After three months, I also sense that it is safer. Same for, let's say, Scrabble game industry. The, the Scrabble words, let's say, you need to publish like 1 million pages for unscramble X, unscramble ZY, etc. Meaning that actually if you are classified as a spammy source, you won't be able to index anything. That's why we always try to grow the site very slowly until we see that actually it is becoming permanent on the SERP. This autocontrol in a self-balancing way, I believe, is the best methodology for now.

James Dooley: So I've got a question for a new website. Right. So someone comes along and they've got massive budgets and they say, okay, I want to enter the iGaming market and I want to go really hard on slot games. Yeah. 30,000 slot games. And I want the 30,000 slot games up, right, on a new website. I've heard from some people, and I can't remember whether it's from you or whether it's not from you, that some people say before you launch the site...

Koray Tugberk Gubur: Yeah.

James Dooley: ...robots take the result and linking it. Yeah. And, and, but it also says that you should publish majority of the pages. So that means from day one there could be 25,000 pages being created on a single day. Is that not scaled content abuse?

Koray Tugberk Gubur: That is also scaled content abuse. That was, uh, in one of our podcasts as well I mentioned. But there, for slot sites too, I wouldn't put 20,000 there probably because it's, it is still too huge number. Initially it will be ranking fine, by the way. The purpose here is actually being safe after four or five months still. Because once they classify the site in this way, it will be very hard to reverse that state as well. In that case, I will try to find the highest, uh, popular terms and the categories of the slot games, and I will try to get the most popular 10% at least to my site in the best possible way while linking there. From there, you can actually keep it safer because 10% still will be acceptable. Plus, uh, also you can directly, uh, link the site while preventing Google to actually crawl the site. This answer will be in a way actually providing a level of trust to the source. Even if they can't see the content, they will be dying to crawl it and index it. Once you open it, if there are really important query terms that you are targeting for, they would index it. And since you will get impressions and the clicks, and if you satisfy the clicks, you should be safe from there. But the click part is important. If the clicks are not satisfied, they will be way much less tolerant for such a source. And from there you can create a momentum higher to complete the 30,000 as well.

James Dooley: So on there, what I'm going to reverse it to the initial question where I spoke about if you had human content but did 5,000 articles, you said because of that frequency, you'd still probably get hit. You still would be in the scaled content abuse bucket. What happens if you only was to do 10 articles a day, but it was AI content that wasn't having the human touch? It seemed okay. It was factually correct. You did all the entity attributes. The, the page served its purpose, but you was doing it a day, but it was kind of scaled content abuse in the fact that it didn't have the human touch. It's almost like a slow death. Could they get hit if they're under the threshold, though?

Koray Tugberk Gubur: Technically, it is, uh, still scaled content abuse, but technically are scaling very slowly in their terminology at least. But in my opinion, it will be very harder for search engine to actually catch you there, especially if you have clicks satisfied. And if especially if you have a long-term indexed domain, if you have also links and the brand search demand, that will be safer. Though still, I usually suggest at least modify the centerpiece annotation and the beginning phase of the document. This way you can keep yourself safer because they check initially on the there. Meaning actually initially they can see the human effort, then the AI section likely can be coming there. And mathematically, you can also run a structured prompt. You can give your own wording structures. While we are talking, I usually say, for instance, "basically," because it's my signature word as a human. And in a similar way, your writer will be having similar different signatures, and you can structure a prompt to distribute these authorship signatures as a human does. This way you can keep your content also safer too because once mathematically they can't fit your content into an AI pattern, you'll be safe anyway. At least in the long term we can guarantee that.

James Dooley: What are my signature words? Say it again. What are my signature words? Got me off track there. Now with regards to scaled content abuse then, right? So I want to play one more example to you. In local SEO there, sometimes you, let's say, someone starts, they offering SEO services and they want... They found out that SEO services in a location, they all get search volume for 5,000 areas within the US and they provide an SEO service in all them 5,000 locations. What would you do on a 5,000-page website with regards to the different locations? Would you launch like a thousand and then start doing 50 a month, or like...

Koray Tugberk Gubur: I can give an example site, for instance, not 5,000 like, but I believe this Query Deserves a Page will be an important concept because since December 2025, I see that Google is tolerating, uh, let's say, less and less of the sites. And if majority of these locations don't have a search demand for this service, I wouldn't open them as a separate page that frequent. I will try to create a location hierarchy and I will try to merge them. And majority of the locations will be just like the subheadings. Majority of the time, a site name there might, might be maybe like Andrew Pickett. You can also check that when you delete some of the location pages there and merge sub-locations in the main Melbourne page, they, they actually increase in the rankings because when you have fewer pages, you increase ranking signal per page and you also prevent possible cannibalisations too. But if you have too many pages that is repeating the same information just by changing the location name, and if you don't have any Google Business Profile for these locations to justify having these pages, probably you will see some crawled but not indexed and discovered but not indexed sections in the Search Console, which is actually a dangerous sign because it might cause also massive de-indexations and also massive ranking loss too.

James Dooley: So, and then with regards to the suburbs around the main city, so you mentioned Melbourne. If you was doing all the suburbs and then you collapse them onto the Melbourne base, the main Melbourne page, would you then put the suburbs in the micro content as H2s? Or do you just do it over areas we cover in Melbourne and then just have it... Is it like almost like bullet points or something?

Koray Tugberk Gubur: Areas covered would be okay too, in a way. But also usually I try to make them unique as well. A numeric value section, meaning that statistics might be okay. Like say the accident statistics or compensation statistics in Melbourne. You can say and you can give statistics for every region. This way you, when you go to another major city, you can have more uniqueness because numbers are way much more uniqueness provider and also originality as well. So I find some attributes that I can use numbers to also create co-occurrences for these suburbs or sub-areas.

James Dooley: And anyone watching this, if you've actually been hit with the scaled content abuse, leave a comment in the comment section. Myself or Koray will reply and have a look at the website. Koray, it's been an absolute pleasure. Thank you.