How ChatGPT can murder your SEO

Introduction

ChatGPT is the ultimate solution for almost any of your business-related writing needs. Or is it? Well, let’s find out.

Ever since it started gaining popularity, there has been an unending avalanche of articles, blog pieces, and studies about it. Some praised it as the next best thing for work automation since steam power, some painted it as the Trojan horse that is bound to bring down the organic traffic of anyone who dares use it. Naturally, this would not be a respectable blog if we didn’t express our opinion on it.

So, here we are. In this post, we will try to figure out whether ChatGPT is as big of a deal as people make it out to be. We will do so by running a few tests and experiments that will show just how dependable this AI can be.

Ready to get started? Then let’s begin.

What ChatGPT is NOT?

There has already been enough talking about what it is, so let’s start with the opposite.

ChatGPT is not a revolution. It is merely the most popularized and the easiest to access text-writing AI. It is quite advanced, yes, but only in the context of what is available to the public.

As Meta’s chief AI scientist Yann LeCun put it in an online discussion series by Collective[i] Forecast, ChatGPT is “not particularly innovative”. The underlying principles behind its mechanisms have been known for quite a while, and are nothing new to AI experts.

Think of it as an iPhone. There were smartphones before it, many of them actually more advanced. But it was Steve Jobs and Apple who polished the idea and put it in an accessible and appealing form that truly changed the market.

What can ChatGPT do?

To keep this brief, I will only mention the kinds of tasks that would be helpful for an eCommerce business.

1. ChatGPT can write your business emails for you.

Tell it what you want it to say in one short sentence, the tone it should use, and watch the magic happen.

In this example, I spent less than a minute typing my request and got an email template ready within seconds.

Could I have written a better email than the AI? Probably. Would it take more time? Absolutely. Would the marginal improvement in quality be worth the time investment? Hard no. The target audience of this email would be angry no matter how I phrased it.

You can also use it to communicate with clients and business partners. Very helpful if English is not your native language, or if you simply are not fond of writing in a dry corporate style.

2. As we have already described in this post, ChatGPT can handle the basic ad creation tasks.

Now, descriptions seem to be more challenging for ChatGPT. As you can see, none of them really fit within the 90-character limit I specified. Still, cutting them down to size can be easier than making them from scratch.

3. Write guest posts.

In this case, the output was far from perfect, as it cut off in the middle of the “Include negative keywords” section.

And after I asked it to go on, the AI skipped some parts of the plan entirely. Still, it is possible to make it write content by feeding it your plan one small piece at a time.

4. Write or debug code.

You can tell ChatGPT to create simple pieces of code for you and get outputs that will probably work. If you are not familiar with coding and need to solve a simple problem without bothering your developer, this functionality can come in handy. Alternatively, you can paste your own code and ask the AI to take a look at it.

5. Write product descriptions.

As you can see, I did not even need to give it any specific details to get an accurate product description. The AI did well to describe the most appealing features of the product. The text it created does not have any technical inconsistencies.

Still, keep in mind that the AI does not come up with the output on its own. Its knowledge is limited by the information it has been trained on. So, if you are dealing in a product that is not widely represented on the internet, the AI will be incapable of assisting you.

What ChatGPT cannot do?

The next thing we should do is clarify the areas it cannot help you with.

1.Work with sensitive topics and regional laws.

ChatGPT is incapable of generating content that includes adult themes, violence or drugs.

However, its limits can be a little vague, and the laws it follows appear to be American for the most part. For instance, it will refuse to write a product description for a set of erotic lingerie, but will see no problem with giving you one for a firearm, even if those are illegal to possess in your country:

And before you try it, it will also refuse to help you with “optimizing” your taxes:

2. Work with new information.

Here’s what the AI itself says: “As a language model, my limits include the information that I have been trained on, which has a cutoff date of 2021.”

This means that ChatGPT will not be able to help you in any way if your text needs to include fresh information. In dynamic, fast-changing fields like digital marketing, this limitation makes it pretty much useless.

3. Have its own opinion.

If the kind of text you need calls for a personal voice or an opinion, the AI will not be able to help you with it. Its attempts at humor and creativity leave much to be desired. Generally, if you want your text to be engaging for your readers, ChatGPT is not the best choice.

4. Help you if you are not in its green zone.

Although ChatGPT is easy to access, it is not easy to access for everyone. The list of countries it is available in is pretty limited. Ironically, those that have more to gain from using it (non-English speaking countries), are denied access.

However, this limitation can be easily bypassed by using an online SIM (or a friend living in one of the countries where ChatGPT is available) and a decent VPN.

Can ChatGPT be detected?

The simple answer is yes. There are already multiple online services available that can identify AI-written content. And if some free-to-use web-based tool can do that within seconds, it is safe to assume that the watchful eye of Google’s algorithm will have no issue with it whatsoever.

Now, let’s look at two texts. One is written by a human, and the other is the product of ChatGPT. The topics are the same.

***

How to Build A Sales Funnel For Amazon With Facebook Ads?

A sales funnel, to put it simply, is a client's journey to making a purchase. With immaculately created Facebook ads you'll get those purchases right away. If not, add additional touchpoints to the landing page.

The landing page has several functions to increase and track conversions:

Filter traffic – sloppy targeting can attract different audiences who can potentially have low buyer intent and never make it to the Amazon page. In order to not be left empty-handed, you can give them something for free, like a promo code or a free shipping coupon in exchange for their email, which you can later use to move them further down the funnel.

A coupon code – a way to introduce some incentives for purchases with discount codes. Building any Amazon page includes the need to distribute promo codes for better turnover. Along with coupon codes, you can insert affiliate links on the landing page, which are unable to be placed on a Facebook Ad. With affiliate links, you can track final sales.

The theory on setting up Facebook Ads is fairly simple and making more Amazon sales is no hard concept. However, do not forget to test your marketing techniques.

***

How to build a sales funnel for Amazon with Facebook Ads?

Building a sales funnel for Amazon using Facebook Ads can be a powerful way to drive traffic and increase conversions on your product listings. By using targeted advertising and lead generation tactics, you can attract potential customers and guide them through the purchase process.

The first step in building a sales funnel for Amazon with Facebook Ads is to create a Facebook Ads account. Once you have set up your account, you can begin creating and running ads that target potential customers. One of the most effective ways to do this is by using retargeting ads, which allow you to target people who have already visited your website or interacted with your brand.

***

Here are our stats for the first one:

GPT-2 output detector: 99.97% real.

Writer AI content detector: 99% real.

Content at Scale AI content detector: 100% real.

And here is what we have for the second text:

GPT-2 output detector: 0.08% real.

Writer AI content detector: 8% real.

Content at Scale AI content detector: 6% real.

There is no plot twist here - the second text is written by an AI.

As you can see, identifying it is not an issue.

Can you fool the detectors?

Now, you are probably wondering why I did not include the “cyborg” texts - written by AI, but with human edits and additions. The reason is that it is not as simple as it looks.

For the following experiment, let’s take this sample:

***

Building a sales funnel for Amazon using Facebook Ads can be a powerful way to drive traffic and increase conversions on your product listings. By using targeted advertising and lead generation tactics, you can attract potential customers and guide them through the purchase process.

The first step in building a sales funnel for Amazon with Facebook Ads is to create a Facebook Ads account. Once you have set up your account, you can begin creating and running ads that target potential customers. One of the most effective ways to do this is by using retargeting ads, which allow you to target people who have already visited your website or interacted with your brand.

***

We already know that the text is fully AI-written. Let’s try to find out how much of it we need to edit to make the detection software start having doubts.

For this experiment, I will only be editing the existing text, without adding anything new. ChatGPT’s text will remain italicized, while the parts I change will be in normal bold font.

Here is what the metrics for this particular piece of text look like right now:

GPT-2 output detector: 0.08% real.

Writer AI content detector: 8% real.

Content at Scale AI content detector: 6% real.

***

Building a sales funnel for Amazon using Facebook Ads can be a powerful way to drive traffic and increase conversions on your product listings. You can use targeted advertising and lead generation tactics to attract potential customers and guide them through the purchase process.

First, you will need to create a Facebook Ads account. Once you have set it up, you can begin creating and running ads that target potential customers. One of the most effective ways to do this is by using retargeting ads. These allow you to target people who have already visited your website or somehow interacted with your brand in the past.

***

At this point, here is what our tools tell us:

GPT-2 output detector: 18.05% real.

Writer AI content detector: 61% real.

Content at Scale AI content detector: 77% real.

A big jump in the last two, but the GPT-2 detector seems to be harder to fool. Let’s take it a little further.

***

Building a sales funnel for Amazon using Facebook Ads can be a powerful way to drive traffic and increase conversions on your product listings. You can use targeted advertising and lead generation tactics to attract potential customers and guide them through the purchase process.

First, you will need to create a Facebook Ads account. Once it is set up and ready, you can begin creating and running ad campaigns that target potential customers. Retargeting ads are among the most effective ways for doing that. These allow you to target people who have already visited your website or somehow interacted with your brand in the past.

***

Here is what our results are looking like now:

GPT-2 output detector: 91.92% real.

Writer AI content detector: 95% real.

Content at Scale AI content detector: 98% real.

Looking good, but not quite there yet. GPT-2 output detector is still having second guesses. Let’s keep going:

***

Facebook Ads can be a great place to build your sales funnel for Amazon. It is an amazing way to drive traffic and increase conversions on your product listings. You can use targeted advertising and lead generation tactics to attract potential customers and guide them through the purchase process.

First, you will need to create a Facebook Ads account. Once it is set up and ready, you can begin creating and running ad campaigns that target potential customers. Retargeting ads are among the most effective ways for doing that. These allow you to target people who have already visited your website or somehow interacted with your brand in the past.

***

GPT-2 output detector: 26.13% real.

Writer AI content detector: 80% real.

Content at Scale AI content detector: 92% real.

Well, this looks surprising. Apparently, continuing to edit the AI-generated text does not always translate to reduction in its visibility. But let’s try some more anyway.

***

Facebook Ads can be a great place to build your sales funnel for Amazon. It is an amazing way to drive traffic to your online store and increase conversions. You can use targeted advertising and lead generation tactics to attract potential customers, then guide them through the purchase process.

First, you will need to create a Facebook Ads account. Once it is set up and ready, you can begin creating and running ad campaigns that target potential customers. Retargeting ads are among the most effective ways for doing that. These are the ads that target people who have already visited your website or somehow interacted with your brand in the past.

***

GPT-2 output detector: 1.70% real.

Writer AI content detector: 86% real.

Content at Scale AI content detector: 50% real.

And now it’s even worse. I suggest we stop now.

For better clarity, here is a comparison of our final text with the original by Copyleaks:

As you can see, this free web-based tool had little trouble with identifying paraphrasing.

Now, let’s analyze what we have learned from this.

Editing an AI-written text to make it fool detection tools is a game of trial and error. Editing more does not always make the text look more human. In fact, past a certain point, it is entirely possible to achieve the exact opposite.

I did all of that just for two paragraphs consisting of 113 words in total. A half-decent copywriter could write the same amount of text in half the time, if not faster. And the AI output detectors would have no issues with their copy.

Keep in mind that we do not know just how reliable these tools are. Their metrics kept going up and down in response to seemingly miniscule word changes and edits.

But they won’t be the ones looking into your website’s content. It will be Google. And its algorithms are alien tech compared to the tools I have used.

But what if we do the opposite?

It is clear that, as of now, ChatGPT’s output can be detected. But writing is not the only thing it can do.

It can also edit.

What if we reverse the experiment and ask it to edit a human-written, but substandard text? Can the AI correct its flaws, while retaining enough of the text’s humanity to make it slip past the output detectors? Let’s find out.

***

The world have become a global village where people from different walks of life who reside in distant places can stay connected through Facebook and other apps. This has also delivered about a change in the global economy since much businesses now conduct their marketing activities online and make a huge ton of money on Facebook.

Let's look at the data. Statistically more than ten million people use Facebook in advertise their businesses. Why? Facebook has over 3 billion users and the majority of these users spend close to two hours daily on this platform. Hence, advertising on Facebook allows you to reach many potential customers without spending a huge number of money on marketing. Cunning marketers use this attention to their advantage by utilizing Facebook ads management.

***

As you can see, the text above has quite a lot of mistakes. Here’s what the detectors have to say about it:

GPT-2 output detector: 99.98% real.

Writer AI content detector: 100% real.

Content at Scale AI content detector: 100% real.

Now, here is what it looks like after I asked ChatGPT to edit it:

***

The world has become a global village, where people from diverse backgrounds who live in far-flung locations can stay connected through Facebook and other apps. This has had a significant impact on the global economy, as many businesses now conduct their marketing activities online and generate substantial revenue through Facebook.

Statistics show that over 10 million people use Facebook to advertise their businesses. The reason is simple - Facebook has over 3 billion users, and a large majority of these users spend close to two hours daily on the platform. Hence, advertising on Facebook allows you to reach a large number of potential customers without incurring high marketing costs. Savvy marketers take advantage of this attention by utilizing Facebook ad management.

***

Before I put it through the detectors, let’s analyze the quality of AI’s work.

It did well correcting the grammar and punctuation mistakes. It rephrased the parts that sounded way off from a stylistic standpoint, and its choice of synonyms is on point. Savvy marketers certainly looks more natural than cunning marketers.

However, it did nothing to fix the village analogy. While it can certainly be subject to debate, I believe that it does look good in a text like this.

Let’s take a look at our metrics:

GPT-2 output detector: 99.90% real.

Writer AI content detector: 96% real.

Content at Scale AI content detector: 100% real.

Looks good, but let’s take it a step further and ask ChatGPT to do something about that village thing.

***

With the advent of technology, people from all corners of the world can now stay connected through platforms like Facebook and other apps. This has resulted in a significant shift in the global economy, as more and more businesses are now conducting their marketing activities online, leveraging the reach of Facebook to reach potential customers.

According to statistics, over ten million businesses now advertise on Facebook, attracting the attention of its over 3 billion users who spend an average of two hours daily on the platform. This offers a cost-effective way for businesses to reach a vast audience, and smart marketers take advantage of this through the use of Facebook ads management.

***

This looks a little better. Let’s see what the detectors have to say.

GPT-2 output detector: 50.31% real.

Writer AI content detector: 94% real.

Content at Scale AI content detector: 43% real.

Seems like we have lost some humanity along the way. Also, the text feels a little bland now. Let’s ask the AI to add a little creativity to it.

***

The world has transformed into a global interconnected network, where individuals from various corners of the globe can stay connected through social media platforms such as Facebook. This has brought about a sea change in the global economy, as more businesses are now taking their marketing efforts online and reaping huge benefits through Facebook advertising.

Statistics reveal that over ten million businesses use Facebook to advertise their products and services. The reason being, Facebook boasts of over three billion users, and a majority of these users spend close to two hours each day on the platform. This provides a golden opportunity for savvy marketers to reach a vast audience with minimal investment in advertising. By leveraging Facebook's advertising tools, businesses can capture the attention of potential customers and turn it into tangible results.

***

GPT-2 output detector: 92.07% real.

Writer AI content detector: 97% real.

Content at Scale AI content detector: 50% real.

Both the text and its performance are looking better now. I think it is time to wrap this up and summarize what we have learned from this.

It is relatively safe to use ChatGPT for basic editing. In fact, it is possible to go beyond that and ask it to replace specific parts of the text that might not fit stylistically.

But here lies the main problem: it can only do so when asked to. For that, it needs a human overseer who can identify and point it to the parts that feel off. And even then, the result is far from perfect and detectable.

Now, let’s proceed to the most intriguing question we must ask ourselves.

Is ChatGPT bad for your SEO?

Before we start deliberating our verdict, let’s give a word to the defendant:

This answer is a perfect example of ChatGPT’s reliance on old information. Because if it had access to all the new blog pieces about itself, it would have much more to say.

Let’s go back to the question in the headline. The short, and the most honest answer is: we don’t know.

There have been proven cases of AI content usage leading to severe losses of organic traffic. However, there is still very little understanding of just how much AI content is needed to cause this.

In order to find this breakpoint, one would first need to invest serious money into building a website and its SEO, then bring it all down with generated content. As you can understand, few have enough passion for science to try it.

Besides, even if we knew this breakpoint, there is no way to tell when and how it would change. And it will change.

AIs and AI output detectors are in a permanent arms race against each other. When an AI learns to bypass detectors, new detectors emerge, specifically designed to identify this AI’s output. And then the next version of AI will be designed to bypass those detectors, and so on and so on.

Managing a website with AI-generated content in an environment like this is like navigating a boat through the worst storm you’ve ever known. The boat gets constantly thrown from one wave to another with no rest at all. And somewhere beneath those waves, there’s the Kraken - Google.

It might not notice the boat at first, but the longer it stays in the storm, the higher the chance that it will be attacked. If you want to fill your site with evergreen content, using an AI is likely to get you in trouble sooner or later. It might take months or even years, but eventually, the detector tools will find a way to see through the AI you used.

But what if your goal is to deliver fresh content that is up to date and in line with the trends? Well, the AIs can’t really help you with that either. The data they are taught on is very limited, and in many cases, obsolete. And it’s safe to assume that those limits are not going anywhere. Because the moment they are gone, the Internet will be plagued by fake news. Remember that the issue of AIs is bigger than eCommerce.

Is this really new?

The difference between human and AI-generated content is only really clear when both are working at their maximum. But realistically, it is entirely possible for a human-written text to be inferior both in quality of writing and substance to one written by AI.

Take content farms, for example. Back in 2011, the Internet was plagued with articles and blog pieces of the worst quality you could find. They were mass-produced by low-paid writers with little knowledge of the subject and even less grasp of what good writing is. The target audience of those texts were not humans, but search engines.

Although the echo of them still lives on, most of the content farms were killed off, while the rest were forced to step up their quality and stop being content farms. What caused this change, you might ask? Google’s Helpful content update.

Every three months Google updates its search algorithm. Sometimes, nothing really changes, but sometimes, websites and businesses come crashing down.

We had a client that was affected by one of them almost ten years ago. They used cheap copywriters to fill their website with content of questionable quality, assembled from bits and pieces “borrowed” somewhere else. And their traffic was great.

Until one day, it wasn’t.

When the latest Google helpful content update rolled out, their organic traffic decreased tenfold overnight. Last we heard, they never truly recovered from it.

Chances are, the same can happen to those abusing AI content now.

So, should we be worried?

It might seem like we are about to step into some kind of dystopia-flavored future where machines generate content for other machines. But fortunately, Google has made their stance clear in their August 2022 Helpful content update: humans come first.

As long as your content is created with humans in mind and adds something useful to the vast ocean of information on the Internet, Google doesn’t really care how it was created.

But can something with a sense of humor like that really write for humans?

I think not. And here is my prediction of what will really happen.

The net quality of writing on the internet will be improved, just like improved when content farms were disposed of. The AIs will replace the lowest-paid copywriters and become the new bottom standard, accessible to anyone. The other writers will be forced to drastically improve to put themselves above them.

In fact, we might see blogs deliberately shifting to genres and styles inaccessible to AIs in order to make themselves stand out among others. Ever wondered what gonzo journalism applied to digital marketing would look like? You might actually see it here, if you stay long enough.

The key takeaway from all of that is this: if ChatGPT is an upgrade in quality for you, you might be in much deeper trouble than you think.

Summary

At the end of the day, it is up to you to decide whether you should use AIs like ChatGPT for your content or not. If you like to live a dangerous life and are willing to risk your long-term website performance for the possibility to drive more traffic, go ahead.

But if you are planning for the long term, you might want to hold your horses, at least until the discourse around AIs stabilizes. We don’t know if AI-generated content will be a problem. But we do know that as long as we are the dominant species on this planet, nobody will have an issue with original, human-made content.

And remember that there absolutely are many amazing applications for AIs. There are plenty of writing tasks they can help you with that will have no impact on your SEO whatsoever. Google’s Algorithm does not care about your emails.