AI Content Creation – 2026 Update

In this article:

• AI content creation can save you time, but it should never replace human content
• Google’s Helpful Content update is now penalising unedited AI content
• Incorrect info and AI hallucinations are now only part of the broader issue with AI
• Other issues include quality, originality, monotony, ‘slopification’, and obvious bot content
• AI copy editing is now essential, but 100% human writing is an even better solution

Since the last update to this post, a lot of things have changed when it comes to AI content.

When AI-produced content first began to flood the internet, reactions varied. First, there was the panic mode that it would replace human content. Then, there was the idea that maybe these tools could offer some value as a tool. Many people saw it as a way to create a structure or rough article drafts, help with content planning and ideas, and save time and effort. However, we have now entered a new era of AI.

This updated post is based on my own experience as a human writer in the content industry since 2006. Obviously, I can’t speak for everyone. I do feel that a warning is needed, because Google is going to make it harder to keep producing these tools. More importantly, we are at a cusp of technology. The jokes I used to make about robots taking our jobs are not so funny anymore. I know a writer who was literally let go because her company created a Claude script to do her job.

So with all of that in mind, I figured it was time I did an updated post on the use of these tools, what they mean for content, and (most importantly), how they affect my services as a human writer. This may not be what everyone wants to hear, of course. If you’re looking for something quick and cheap, you may want to continue using your tools. If you care about what Google and your audience think, you might want to keep reading.

When AI content first began to flood the internet, reactions varied. First, there was the panic mode that it would replace human content. Then, there was the idea that maybe these tools could offer some value as a tool. Many people saw it as a way to create a structure or rough article drafts, help with content planning and ideas, and save time and effort.

However, we have now entered a new era of AI.

Still using AI content? Here’s what to know.

Relying on AI alone was never a good move, especially when it came to generic tone, factual errors, and the lack of nuance. This made it hard to beat human content. The potential for this as a tool slowly became something different. Instead of using it as a tool and bringing in humans, it began to be used on a widespread basis, normalising AI-produced content.

This brought us to the era of slop. This is the term that describes low-quality, mass-produced content that is made using assorted AI tools. The term has become so widely used that it was named the Merriam-Webster Word of the Year for 2025. It’s officially in the Cambridge Dictionary as a noun. Dancing cats, fake news, completely crazy videos, and other nonsense are flooding the internet.

It isn’t just this type of annoying, obvious nonsense. Even Pinterest was affected. When I started noticing that home decor ideas were getting weird, I realised just how bad things were getting. I can’t open Facebook these days without seeing at least one fake animal rescuing baby video, a few AI-voiced-over recipe reels, and a service poster that looks exactly the same as the three above it, only in different colours. If I’m really lucky, I may also get some fake embroidery images, too.

Then, if I go onto a site to look for a product (any product), I see content flooded with obvious AI flags. I’ve even seen real estate listings made with ChatGPT.

What does all of this mean when you just want to put content onto your site and you don’t have a budget or the ability to do it yourself? Should you ditch your ChatGPT completely? Is Claude better? Is it ok to use AI in some cases?

Let’s take a look to get the lowdown on the risks, reality, bigger picture, and what you, as a user should know when using these tools.

Google isn’t happy

Google’s Helpful Content guidelines set the stage for the current shift towards quality content written by humans. Things like E.E.A.T. (Experience, Expertise, Authoritativeness, and Trustworthiness) became a thing. Of course, machines didn’t quite make it into the guidelines.

Even in 2025, Google was starting to get a bit iffy about non-humans producing content. As long as content was still helpful and edited, fact-checked, and useful enough that your audience would be helped, it was fine.

Now, thanks to the slopification of pretty much everything, Google’s beginning to see that maybe humans prefer content that’s written by other humans. What the latest changes are doing is penalising sites that are relying heavily on ChatGPT, Claude, and other tools.

They are coming down just as hard (or even harder) on sites doing this en masse. So if you’re assuming that you can do your entire site using a tool, it may seem faster, easier and cheaper. But Google isn’t going to see it that way.

Key Takeaway: Non-human content is going to hurt your rankings, sooner or later. Don’t wait to see what the Google Helpful Content update is going to do to sites that are not using original content.

AI watermarking is here

This is something you really want to note. AI watermarking is being used by models across ChatGPT and Claude. According to Wikipedia, watermarking works by “producing a probability distribution of the next word, and sampling over that distribution at random. Text watermarks function by modifying the selection process over that distribution.”

In plain language, it means that random bits of text are noted and assessed on whether they look human or not. This is mainly in place due to policies on transparency. You can see Claude’s overview and what it means for AI text to get an idea of how this works. OpenAI is doing the same with text created with ChatGPT.

What does this mean, exactly? Basically, it means that text can be flagged as AI-produced.

You know those labels on social media images that state things are made with AI? It’s like that, only there is no clear sign. It’s invisible. It means that your content may get flagged when using AI detectors, even if it’s been edited a bit.

The one thing to note about watermarking is that it doesn’t always guarantee being flagged. Shorter content is generally harder to detect, compared to longer content like blogs, as there are more words. The point of this is that your content will very likely be flagged as AI if you have used AI to write it, with no human editing, obvious signs, no fact-checking, and just 100% produced by whichever tool you use.

On the downside, content can be human-written and get a false positive in rare cases (meaning it gets flagged for some reason).

Key Takeaway: It’s getting harder to hide machine content now. That means that your content will be flagged as “written by [insert chosen tool]”. Obviously, this is not something you want. Especially when you’re in an industry that requires a high level of trust and authority.

Tone & voice are still bad

A major hurdle with AI copy is still the ability to convey tone and brand voice. This is because no machine will ever replace the human ability to inject nuance into written content. AI tools create text by collecting related information and linking it together. They do not understand context or flow. They may be able to create text that reads fine if the copy is simple. But they do not bring across the subtle tones that make your content suited to your brand voice. Instead, they use generic descriptive tones across all types of content – regardless of your niche.

Example: Overly descriptive terms like “enchanting”, “captivating allure”, “soothing beauty”, “bask”, “embrace”, and other terms not used naturally.

Teaching bots to capture brand voice is difficult, too. This is because every brand has a unique voice. You know how great it is when you can throw a quirky line or phrase into your copy that captures what you want to say? That’s something that bots don’t know how to do too well. AI cannot inject genuine emotion into text. Only humans know how to do that.

Key Takeaway: Tone and brand voice are essential to connect to your audience. Only humans can convey your unique tone that is suited to your niche and style. Somehow, years after arriving on the scene, these tools still haven’t gotten this right.

Privacy risks are worse

Yes, I know. That sounds a little dramatic. In the early days of AI, many were worried about the impact of rapidly changing technology. Italy, France and other countries in the EU are taking a very serious look at the privacy implications of AI. After employees at Samsung leaked sensitive data when using ChatGPT for work, corporations are getting alarmed, too. This is something every single business needs to know when using AI in content creation.

I originally wrote this article in 2023. I updated it halfway into 2025. Every year, there have been issues.

Bots are getting into trouble for using intellectual property, and as laws around AI are still a little murky, it’s harder to prevent them from stealing information. Bots can also copy huge volumes of pages in a very short time. It’s even worse now. Bots use everything you put into chats for training.

Whether you’re asking for advice on a random medical problem or planning out a business idea, all of that is being used to train. Just have a look at this list of concerns on IBM to see how your data could be used.

Key Takeaway: Privacy concerns remain a hot topic. This is because AI can steal and copy large volumes of information for training and decision-making. There is little control over how this data is collected, used, and protected.

Factual issues are still there

I mentioned AI hallucinations earlier. These sound like something out of a sci-fi movie. But then again, so does AI in general. When I pictured the inevitable robot invasion, this really wasn’t what I was expecting. Anyway, AI hallucinations are a thing, and fact-checking is now essential.

A while back, AI produced fake Guardian articles that looked exactly like the real thing, opening up the risk of misinformation, conspiracies and confusion.

Wikipedia has a whole entry on hallucinations. My favourite example is ChatGPT being asked for evidence that dinosaurs built a civilisation, then insisting that fossil remains of dinosaur tools had been found, stating that “Some species of dinosaurs even developed primitive forms of art, such as engravings on stones”.

OpenAI itself has admitted that hallucinations are a thing with ChatGPT, which is its own product. Duke University Libraries outlines the problem well, especially when it comes to why this is still happening in 2026 (and why we should be worried).

Key Takeaway: As far as content goes, this means that there is a huge need for human fact-checking. The last thing you want is errors in your copy that are missed because no human has gone over facts.

What about Claude?

Is Claude better? It seems more professional, without the obvious ChatGPT flags, right? Well, as of October 2026, Anthropic has banned users from “cruel behaviour” towards its models. The good news is that this won’t apply to casual users getting frustrated with the tool (for now). This is still a bit worrying.

We don’t need to get too technical or go into issues like rogue agents and breakouts (or its serious crime fake reporting to police). We definitely don’t need to talk about a Claude researcher quitting due to being “genuinely frightened” about humanity’s future.

These are not things the average user needs to think about, unless they want to contemplate the broader effects of AI in general. In this article, we’re talking about words.

From a content point, Claude is similar to any other tool, including Gemini. This means that it can also throw out content that has facts wrong. The tone can be better than other tools, but it’s still stilted and generic.

In some cases, using prompts to make it sound more professional can end up making it sound more robotic. The professional tone can seem more authoritative, which leads to a false sense of trust. Factual errors can still happen, and if you’re not checking content and links, this can lead to problems.

Key Takeaway: Claude is still AI and has the same challenges as any other tool. Using it without human help gives you the same risk of quality, tone issues, and generic voice.

So what do you do about AI content?

This is a whole new era. One we may not have been prepared for when it blasted into our lives back in 2023. We’ve gone from the initial panic to the idea that maybe this could be useful, and now back to the panic. If the creators of these tools are to be believed, we may be entering a phase that we could never have predicted.

The bigger questions and ethics surrounding AI are a little overwhelming for those who are focused on everyday things like putting blogs up or doing product pages. It doesn’t seem like a big deal to open ChatGPT and ask it to make some pages or maybe a poster for your upcoming sale. That’s what tools are for, right? Yes, that is indeed the point of tools.

A note here is that not every tool has to be completely useless. Asking Google why orcas like to sink boats, whether sea otters really are the ocean’s psychopaths (they are), how to do a stitch you just can’t get right, or even how to make a healthy dessert using the three ingredients you have available is not wrong.

AI overviews can even be useful if you can’t be bothered to close them (as long as you always click through to the links and read the sources/watch the videos). GEO is a whole other ballgame.

With that said, it is important to understand how the tool works for things like content creation rather than the random questions you are asking, and what it is doing on a larger scale. These are conversations we need to start having. The box has been opened, and now it’s up to us as users to see how we approach non-human content.

Ultimately, the decision on whether to use AI tools for content comes down to your budget and your goals. If you need to churn out generic copy quickly, using these tools will do the job. If you need well-written content, relying on AI completely is a big risk.

Here’s what to do instead:

• Hire a writer to do your content the old-fashioned way, written by a human
• Focus on Google’s E.E.A.T. content, rather than mass-produced content
• Take time to learn the issues with AI in general, and why slop is a problem
• Instead of using any AI tool completely, get help from a human to edit your copy
• Use a writer with experience in editing and SEO content (and understanding of AI)
• Create rough drafts and general ideas if you need to, but remember it is not human

Ultimately, a human writer is an investment that will pay off in the long run. An experienced writer will always be better than a tool, because humans know how to connect with other humans. If you still want to use tools to write your content, make sure you have a real writer to edit, fact-check, and align the content with your audience.

To avoid the potential problems of using AI for content creation (and what to do instead), give me a shout and let’s see how we can turn that boring, generic robotic text into something just right for your brand.

This AI content article was written in April 2024 and updated in June 2025. Current rewritten version: October 2026.