Fight back against the decreasing quality of AI-created content
Julia says: “The SEO industry is going back to human-written content to fight all the AI-created content in 2025.”
Is AI content not here to stay, and is it not getting better?
“I don’t think so. From what I’ve been seeing over the last few weeks, it’s actually making everything worse.
I have one relatively new client who I started working with about 2 months ago, and we’ve seen so much AI-written content on Google that nobody can even understand. Then, new content gets created that is based on that AI content, and you need to have a dictionary with you to even understand it.
It’s in the medical niche, so it is quite a sophisticated specialist topic. We’re trying to figure out how to determine what we should write when everything we see is AI-written content that nobody can read if they don’t have a medical degree.”
Can you measure this decrease in quality?
“It’s difficult to measure. It’s more of an impression that you get when looking at the content that has been created recently, especially in specialist niches. You don’t understand what it’s talking about. There’s so much terminology thrown in that it’s just too much for the average reader.
I’ve also noticed that there’s more and more misinformation out there on social media channels. You’ll get information about a teaser trailer for a ‘new’ Netflix series that’s 3 years old and there will be reports that somebody has died, and it’s not true. I obviously look at these things with a different eye than somebody who’s not in digital marketing.
I’m seeing it but I don’t know how to measure it. There are companies and tools out there that are trying to measure and fight misinformation especially. I have the feeling that the big social media platforms haven’t yet found a solution for how to identify what is true and what is false. At the moment it’s more something that I’m seeing.”
Is human-written content preferable for any industry or are you only talking about highly technical content?
“I would apply it to any industry because the issue remains the same. There is content out there that has been created using AI, and then the newer AI-generated content is just feeding off the same sources and recycling the same information.
I noticed it for medical content in particular because it’s quickly coming to the point where you can’t even understand the content anymore. In other niches, you can still understand it, but it’s repetitive, it’s not as nicely written, and it’s not enjoyable to read.
I’m also noticing content that is simply questionable. I have a personal interest in surfing and I’m in certain groups. Somebody recently posted a photo of a surf spot with a really nice analysis of it. I was thinking, wait a second, can you actually surf there? Are there waves there? You could see from the structure of the content that it was definitely AI-generated, but people were taking it for granted.
This issue applies to any industry because recycled information just gets more and more recycled, and nothing new gets added.”
Are you using AI in any step of your content production process?
“At the moment, I’m not.
Obviously, the tools that we’re using are increasingly becoming more AI-based. We’re seeing AI overviews in Google and, of course, I’m looking at that, but I’m not actively using any tool that helps with content creation.”
How does a human writer compete with the speed at which AI can produce content?
“They can’t. It comes down to quality.
As we speak, Google is rolling out more and more algorithm updates. For the past 2 years, it’s all been about what is actually helpful for people. Helpful content. It’s a buzzword in the industry. In my opinion, this is also an attempt from Google to fight the overflow of AI content that gives the same information all the time.
A human writer doesn’t need to keep up with the speed. We just need to deliver better quality, more helpful content for readers to outrank the AI. It’s not so much about the amount of content you produce, especially in SEO. It’s more about quality, and writing the right kind of content.”
How do you make a piece of content more helpful?
“It also comes down to keyword intent, what people are actually trying to find, and providing that information. What is helpful for people?
On my own personal travel blog, pieces where I write about public transport perform incredibly well, because that’s helpful for people. People want to find out where the nearest bus stop is and which bus they need to take to get to a specific place, anywhere in the world.
Here in the UK, it’s quite easy. You just open any app on your phone and it tells you exactly where the nearest bus stop is. However, there are other parts of the world where that’s just not the case. These blog posts perform so well because they fill a gap, and they are helpful. They’re telling people, ‘Here’s this website where you can find out bus times. Here’s where you can find this bit of information.’
What is helpful varies greatly between industries. If we come back to the medical niche, somebody who has a medical background will want to find research papers. That’s what’s helpful for them. Consumers who are looking to buy nutritional supplements, on the other hand, just want easy words. They want to understand it, and find out which product they need to buy or which supplement to take.”
What metrics do you look at to determine whether an article has been performing well?
“Firstly, how many clicks it’s getting from Google. Then, are people actually reading the content? How much time do they spend on the page?
I’ve found that, especially with these pieces where I talk about public transport in random countries around the world, people are actually reading that. I don’t know whether AI is reading it, but I’m sure that AI isn’t trying to take the bus.”
Are search engines currently more concerned with user experience than whether or not a piece of content has been written by AI?
“This is a big question mark at the moment. How do search engines deal with all this AI? How are they going to identify what is good and what is bad?
Google is now rolling out AI overviews in more countries, which is interesting. When they initially rolled it out in the US, it was a big upheaval in the SEO industry. Then, gradually, it seemed as though they rolled it back a bit and we started to see them coming up less often in Google search results. Now they’re suddenly rolling it out in other countries.
I am getting the impression that Google doesn’t really know if it likes AI overviews or not. They’re testing. I’ve also seen lots of examples where the overview is not very useful for people. Then there are other examples where it is good, and it becomes a bit closer to what we already had in featured snippets, where you just get a quick answer to a question.
I don’t want to theorise about what’s going to happen, but I think we will see a lot of testing and a lot of changes before Google decides what they actually want to display in terms of AI content in search results.”
Is it similar to the inconsistent quality that we get with People Also Ask on Google?
“Yes. With People Also Ask, the questions are sometimes misspelt or grammatically incorrect. When I pass it on to the content writer I work with, I say that these are the questions that come up in Google for this particular keyword. I ask them to please answer these questions in the content they write. They come back to me saying, ‘Did you misspell them? Do I need to use it like this?’ It’s strange that Google displays something that is obviously grammatically incorrect.
Another point I wanted to bring up is regulation and legal frameworks in relation to AI. Especially when we think about misinformation being out there, this can quickly become a really big political issue. I have a client who’s in threat intelligence and we’re constantly talking about these issues. There’s a lot of misinformation out there, particularly on social media, that can lead to other problems in geopolitics, so governments are trying to fight that as well. They just haven’t figured out how yet.
At some point, there may also be some legal restrictions on the use of AI in content production, and in what Google is allowed to display that has been generated with AI. We may need to get verification badges for authorship, for example.
I grew up in Germany, and Germany already has some of these things in place. Any website needs to have information about who the person behind the website is. We call this the ‘Impressum’ in Germany. I think ‘imprint’ is the English word, but nobody really uses it in the English-speaking world. There are already things in place, and some higher editorial standards, to tell you who has actually written this content.
I wonder if, at some point, we’ll start to see more of that within search in general, globally, for website content production. That’s just a theory I have. I don’t know what shape that might take, but I wouldn’t be surprised if there’s some kind of regulation.
We’re already seeing the beginnings of markup that informs you when something has been primarily produced by AI. Instagram recently added a new toggle that you should use if an image has been created with AI. There are early signs of something like that.
How does it look in reality and how feasible is it all? That’s another question, of course.”
If an SEO is struggling for time, what should they stop doing right now so they can spend more time doing what you suggest in 2025?
“I think it’s the same tip that I gave last year because I still see people doing this. Stop obsessing over title text and meta descriptions. Yes, we do need them – and, yes, we want a keyword in them – but please don’t go and rewrite title tags for 200 pages without even looking at the pages themselves, just because a tool tells you the title tag is too long.
Firstly, there is no character limitation anymore. How many characters Google displays is different every time you open anything in your browser. Secondly, if the title tag is too long or too short, there might be other issues with the page.
At least if you’re using AI to write titles and better descriptions it wouldn’t then be a waste of time, if they do it in 2-3 minutes, but you still want to review what the AI has written.
A lot of SEOs still export these lists from a tool that says, ‘These title tags have more than 70 characters. Rewrite them.’ Then they sit down and rewrite them. They might have some help from AI or formulas to get keywords in there, but I don’t think this has any value if you haven’t looked at the actual pages and checked whether the content is even suitable for that keyword.
Just injecting a keyword into a title tag and making it fit the character limit doesn’t have any impact.”
Julia-Carolin Zeng is an SEO consultant at Charlie on the Move, and you can find her over at CharlieontheMove.com.