At the moment, the whole tech world seems to be talking about ChatGPT. All of a sudden, it is possible to create even longer, coherent texts with an AI that make sense and read well. An input like “Write a 500-word blog article about 10 travel tips for Bangkok” delivers a text that a travel blogger might immediately copy onto his website. Even before that, tools like Jasper delivered pretty good content at the click of a button, even if the quality wasn’t quite as good as ChatGPT, which is based on a technological advancement of the underlying base technology that both programmes use.
But the technology for recognising AI-written content is getting better and better, and the supposed money-saving travel blogger could be in for a rude awakening if their site suddenly drops in Google rankings.
In this article, we want to take a closer look at the topic of AI content detection so that you understand how you should use ChatGPT & Co. and how you should not.
AI Content - definitely a disruptive technology
In a (perhaps not too distant) world where any content can be created at the touch of a button, many organisations will have to adapt. Universities, for example, will have to consider how much sense it makes to grade assignments and perhaps adapt their educational concepts to today’s technological possibilities. And today at least, copywriters and content writers who find themselves threatened by ChatGPT and similar programmes will either have to change how they work or become better at justifying their salaries by producing better quality work than AI.
We don’t necessarily want to assess here the extent to which the above travel guide about Bangkok offers truly original information – that’s up to everyone having been there to try. Instead, we want to look at what happens when a user publishes AI-generated content (almost) unedited on their own website.
Plagiarism - the first problem
First of all, we checked the text we had just generated for plagiarism with Copyscape – and we actually found something. An internal study by The Happy Beavers has shown that plagiarism systematically occurs in AI-generated texts. Especially when venturing into a more specific topic, the generated article may resemble a single source on the internet in large parts.
This is not always the case, but it is an initial obstacle that should be eliminated through further development of the technology. After all, good AI technology already exists for rewriting an existing text.
Can AI-generated content be recognised?
Short answer: yes. The technology is still relatively new, but from the user’s point of view it works like a plagiarism scanner. You take any text, upload it to the scanner and receive a result, usually some kind of percentage value on the probability that the text was written by AI or a human.
“All AI content generators on the market are essentially the same basic technology, called GPT-3,” explains Luka Karsten Breitig, CEO of The Happy Beavers. He regularly evaluates the work of content writers and has studied the new AI technology extensively. “In short, AI always uses a certain style and this style can be recognised by systems geared to it.” Sentence length, sentence structure and word choice play a crucial role.
But if an AI has the entire treasure trove of data on the internet as its basis, it is only a matter of time before an AI can also imitate writing styles or even invent a new one.
However, there are some companies that have specialised in precisely this question. They develop scanners that can recognise AI-generated text. In our tests, this worked quite reliably, both with AI-generated text and with text written by humans.
Here is an overview of the major players on the market:
Originality.ai has already been quoted a lot in the professional media. The programme scans plagiarism and whether the content was created with AI and can even scan entire websites. According to its own statement, the accuracy is 94%. The scan of 100 words costs $0.01 – an acceptable price for the service offered.
2. Content At Scale
Unlike the previous tool, Content at Scale, not only scans your content for plagiarism but can help you generate blog posts based on your list of keywords that successfully pass through AI content detection. The main benefit of the programme is humanly written AI content which can be produced in over 100 languages, Content at Scale suggests.
3. Kazan SEO
Kazan SEO is a search engine optimization tool itself that also offers AI content detection as an additional feature. Besides offering to generate optimized content and organize your keywords, it conducts a linguistic analysis when reviewing a content piece to determine whether the text is written by a human or a machine. In addition, Kazan SEO is a completely free tool so make sure to try it out yourself.
4. Writer.com's AI Content Detector
Writer.com’s AI Content Detector is a particularly useful tool for writers and publishers as offers not only scanning machine-generating content but also content editing and creation. One more convenient argument is integration with tools such as Chrome, Google Docs, Word, Outlook, and Figma which can make your life way easier. You can try out the tool for free for 14 days and then choose Team or Enterprise versions for a price of $18 per user in a month.
5. CopyLeaks AI Content Detector
Last but not least is Copyleaks AI Content Detector uses its algorithms to compare your uploaded content across various sources and shows you similarities found. With a claimed accuracy of 92%, the tool suggests whether your text was written by AI or a human being. Note, that the AI content detection feature is free while anti-plagiarism requires a fee.
What does Google do with AI-generated content?
So if the above-mentioned companies are able to recognise AI-generated content with a certain degree of certainty, then we can assume that Google also already understands this technology. Google itself is working on its own competitor product to OpenAI’s ChatGPT.
The big question now is: What does it do to my own SEO ranking if I obviously copy AI-generated articles onto my website?
Google itself defines automatically generated content in its spam policies as follows:
Spammy automatically generated (or “auto-generated”) content is content that’s been generated programmatically without producing anything original or adding sufficient value; instead, it’s been generated for the primary purpose of manipulating search rankings and not helping users
This is based on the assumption that AI-based content is inferior in quality to what a human can put on paper. Previous updates to Google’s search algorithm have also repeatedly emphasised the quality of the content on the website as a major influencing factor.
However, Google has never directly said that it is against AI-generated content per se. Danny Sullivan explained in a recent tweet that it is not the tool that is decisive, but the result: according to this, Google wants to evaluate tool-agnostically whether the content is “helpful, reliable and people-first”, regardless of the creation method.
Sunar Pichai, CEO of Google’s parent company Alphabet, also made Google’s AI-friendly position clear:
“It (AI) can make humans more productive than we ever have imagined.”
What does this mean for companies and players in the market?
So we have learned that it is currently relatively easy to recognise AI-generated content. We also know that Google does not penalise pages per se in the ranking if they use AI-generated content.
In our opinion, the following conclusions can be drawn from this:
1. The war of the tools is on
The next few years should be exciting in this area. “We will most likely see a battle where AI content creation software and AI detection software try to outdo each other permanently,” says Luka Karsten Breitig. “Constant new software updates on both sides will balance each other out for a while until one side wins.” OpenAI, the company behind ChatGPT, obviously wants to be in on both sides of this – it has released a tool that can detect AI-generated content.
2. Bad writers should be concerned, good ones not really
ChatGPT can already replace some copywriters, but these are the ones who offer their work as freelancers cheaply and unskilled on relevant online platforms. For these writers, it will probably make sense to do their work with ChatGPT for a while, until it gets noticed, and meanwhile either get further qualified or change jobs.
Good writers, however, can reliably produce articles that are better than what AI currently produces. And they will most likely stay ahead of the game for quite a while – much like a translator can use machine translators such as DeepL and Google Translate, but still be a quality guarantor that the result is better than what the machine says it is.
3. We could witness content fatigue
If content creation becomes cheaper and degenerates into a mainstream product by means of AI, then we could experience fatigue with evergreen content, which at the same time increases the demand for original work that stands out from the crowd in terms of quality. So if numerous players put AI-generated content on their websites en masse, this could also lead to an increase in the demand for quality content and thus human content creators being in particularly high demand.
4. Google’s position remains ambivalent
While Google’s official representatives have stressed that they are focused on the outcome, not the method of production. Nevertheless, this position can also change at any time. Even today, it is by no means certain that AI content will not already be judged to be inferior in quality precisely because it is not “human first”. Google’s “Helpful Content Update” ensures that a well-written article can bring more customers through a high ranking than 100 that are written by AI and therefore do not stand out from the crowd.
5. We have won a tool and it would be absurd not to use it
Just a few years ago, clients would tell translators not to use machine translators while working under any circumstances. For tech-savvy and non-dogmatic professionals, this feels like telling a designer to create a company logo without Photoshop and with just a piece of paper and pen. Just as machine translation is a tool that, when used correctly, can help produce better work, ChatGPT is a tool that can help writers write better texts. If they use it properly and not just to copy-paste the result.
The topic of AI content creation is absolutely new and much is still unclear. While AI-generated content certainly makes life easier in many areas, this definitely does not mean that the profession of content creator is now made obsolete by technology.