The moment we talk about AI gone rogue, several top-rated sci-fi movies come to our mind – Ex-Machina, The Matrix, Tau, M3GAN, 2001: A Space Odyssey, Blade Runner and many more. The fear of AI doing the wrong things or not delivering what humans want has been a core topic since the beginning of the AI era.
Today, AI is revolutionizing several processes in all spheres of life – from smart homes to the healthcare industry and marketing to politics. AI benefits are numerous, but it is still an evolving field that needs more time to come into full effect and deliver what is truly expected of it.
AI tools for content writing – Suddenly on the rise due to the hype
One of the fields where AI is being hailed as the next best thing is content writing. AI-generated content picked up traction a few years ago with the arrival of AI content writing tools. The market was suddenly flooded with AI content writing tools like Jasper, Rytr, Closers Copy, Writesonic, Article Forger, Scalenut, etc.
Like the list of sci-fi movies about AI, the list of writing software is also endless. We are not reviewing any AI writing tool in this article. The idea here is to analyze if AI content generators are successful or not and what are the improvement areas that can make them highly usable.
With the launch of ChatGPT by OpenAI, the hype around AI content writing tools keeps content writers and SEO experts busy. Many of them have already labeled the AI content generators as game changers, predicting that human content writers might not be needed anymore. This has given rise to many questions, but there’s still a large section of content purists who have spent years in the content writing field while learning the intricacies of writing and editing and how to make it both human and SEO-friendly so that it serves every purpose laid out in a digital marketing strategy.
Are AI content writers good? Let’s talk about ChatGPT first
As a freelance content writer, I explored some important sides of AI content generators, free and paid and also Chat GPT to analyze if it actually meets the goal or not. I approached the testing with some very basic methods instead of trying to write poems or code strings.
I started with a basic test on ChatGPT and tried writing a short article on how to run an ecommerce business successfully.
Content generated by ChatGPT
Observations:
Grammarly score – Low - 87
Readability score – Very low - 34
Score on Hemingway App – Very low – 14
Plagiarism as per Grammarly Writing tool – 14%
Grammarly score for ChatGPT content
The plagiarism score as per Grammarly might not look too much, but if we consider a few key areas, then it’s a serious matter:
There is nothing called acceptable plagiarism, so no percentage of it on any of the tools should make you feel relaxed about your content being unique.
Grammarly plagiarism score for ChatGPT generated content
Grammarly suggestions to improve the content
Readability score on Hemingway App
This was my first question to ChatGPT on this topic. The second time I used it by tweaking my question to ‘How to be successful with an ecommerce business,’ the result was even more disappointing. The ChatGPT bot did not paraphrase or present the information differently. I used Copyscape compare tool to make a uniqueness comparison. The number of word matching was high, and there were parts where long strings of text, even up to 64 words, were there.
Copyscape comparison tool for two different versions of same topic idea
What do the ChatGPT content results mean?
The results mean that even slightly more usage of this AI tool for content writing will lead to 100% plagiarism in plagiarism checks (not just comparison tools). Once I use content from it, post it online, get it indexed on search engines like Google and Bing and then again generate content from it around the same topic and run it on plagiarism tools, the result will be 100% plagiarized. That is because all the plagiarism checkers, like Copyscape, Grammarly, Turnitin, 1Text, etc., scan the pages online to determine the plagiarism percentage.
Once a copy is live on search engines, the content generated by ChatGPT on the same topic will start matching it. So, as more and more people start using AI content writing tools, the percentage of plagiarism in their content will start getting more. Even if content writers try their best to make it unique and relevant by adding more words, they will not be able to achieve the 100% unique content tag.
We must remember that Google’s plagiarism detector model works better than any other tools available to content writers. So how does Google detect AI content? It does so with its AI content detector tools. We do not know what tool or method Google uses, but we know for sure that they are better than any other tools available online, whether it is Copyscape or Grammarly.
Why is ChatGPT content not reliable?
The reason for the failure of the ChatGPT tool for SEO content is simple. It has limited information, and the tools used to train it to act as a small reservoir of knowledge. In comparison, the information on the internet is a large sea, but this open source AI tool is not allowed to consume it without permission. This can be said based on the pause that Elon Musk put on OpenAI after he learned that it was harnessing Twitter database for training purposes.
AI content detector tools
Besides, there are simple AI content detection tools already available that can detect if the content has been generated with an AI tool or not. An AI content detector app developed by a student to check ChatGPT generated content is now popular. It’s called GPT Zero.
GPTZero results
Hugging Face is another popular AI content detection tool. Its results are highly accurate (not 100%, though) but once you mix up content with human writing and add more words from your side to give Ai content some depth, the fake percentage on the tool starts coming down. This means the tool can be manipulated with some simple writing tricks.
Hugging Face tool for AI content detection – most accurate of all but not entirely reliable
Result from another tool for the same content – suggesting 79% as fake content
Results from another tool indicates only 8% fake content
Since AI content writing is too risky for your online presence and overall digital marketing strategy, it is best to stick to human writing. AI generated writing has gained immense popularity after the launch of OpenAI’s ChatGPT, but here is what the tool itself has to say:
What ChatGPT says about AI content writing reliability
Why human writers are better than AI content writing tools
Human writers vs AI content writer tools do not make a relevant discussion at this point because the clear winner is human writers. The ability of human writers to convey emotions and infuse creativity make them better than AI writing tools in every sense. Their ability to deeply understand the context and tone enables them to avoid misinterpretation of writing instructions and keep the ideas intact while penning their thoughts.
Since human writers use their niche-specific knowledge and unique experiences instead of following some patterns or working as per pre-formatted templates used for paraphrasing while using a limited base of words gives them a clear advantage over AI content writing software. The art of storytelling, reasoning, analyzing, evaluating and sharing experiences is unique to humans, something for which they are naturally programmed.
Analysis of content produced by AI writing tools
Before getting to the analysis of tools that are termed as specialized in writing content, let’s analyze some of the best-known programs/tools to see how well their AI algorithms work to ensure the correctness of the content.
I typed ‘Tom Hollan starred in the move Spider-Man’ in Grammarly to see if the tool is well-trained to detect mistakes in name spelling or not. Note – I used ‘Hollan’ instead of ‘Holland.’ Surprisingly, it does not detect anything.
Grammarly unable to detect the error
Google Docs identifies the mistake and suggests a correction
Google Docs detects the error
MS Word editor identifies the wrong name spelling
When I change the actor and movie name to something else but again use famous ones, the results are again surprising. None of them can detect the wrong spelling, and they think ‘Ban’ (correct is Bana) is correct. But the Google search engine identifies the problem and therefore gives this suggestion:
Grammarly fails again
This time, Google Docs fails
But Google search engine identifies the mistake and suggests correction
Things can get worse to the extent that Google, which talks about strong policies against harmful content, makes suggestions that can have extreme consequences. Here is an example.
Horrifying – the suggestion by Google Docs for the word ‘poising’
Here's an example of how Bing search engine can complicate things with it's AI-powered search results. You search for California realtors and the results in 'related people' show someone unrelated and unusual. See the highlighted person. What could be the reason for such wrong result? Too much confusing, duplicated content around the keyword? And Bing's inability to comprehend that content? Or something else?
California realtor (highlighted person) as per Google search engine. AI search algorithms in action.
A typo in MS Word ‘shaing’ receives 3 different suggestions ‘shading, shaming, shaking.’
Suggestions on MS Word
In Grammarly, the same typo ’shaing’ receives 3 suggestions ‘sharing, shipping and shaping.’
Suggestions on Grammarly
One’s suggestions are so different from the other!
With all these pointers, it can be easily said that no AI content writing or editing tool is 100% reliable. All the AI content software, including plagiarism checkers, grammar checkers, readability checkers, and writing and editing tools, whether free or premium, cannot match human writing quality.
The tools are highly prone to errors and can be tricked into providing factually wrong and misrepresented output. This is why content writers must know that manual writing, editing, and proofreading are irreplaceable. For SEO-based content writing, AI content tools should be used only as assisting tools.
Some other AI content writers that I have tested
As a top-rated freelancer on some platforms and working independently, I keep receiving testing requests for different kinds of tools – SEO tools, writing tools (including AI ones), grammar tools and much more. Recently, there have been numerous requests for AI content writing tools. Instead of going into the detailed reviews, I would present content samples from a couple of these tools.
Requests I receive for testing AI tools (client name, company name and some personal details hidden)
In the first one, I tried generating content for the introduction of a simple topic, ‘Why is SEO important.’
Here are the results:
Results round 1
Results round 2
Results round 3
Does this look like quality SEO content writing? Is it an introduction that can rank your content high on Google, leave aside the top 3 ranking on the first page? NO.
Professional content writers would know the issues – spacing, grammar, bland tone, overly basic information, highlighting the negative side, looking critical in some parts and so on. These are all things that should NOT be in a good blog.
Here are the results from another AI content generator where I tried the topic ‘Student health importance .’Output review – same as the previous one.
Results from another AI content tool
Here is the 3rd AI SEO content tool I tried with the topic ‘Email marketing mistakes.’
This turned out to be the most disappointing tool. Reason – it totally ignored the important word ‘Mistakes’ in ‘Email marketing mistakes’ and instead churned out content on ‘Email Marketing ideas’ or why ‘email marketing is important .’That beats the entire idea of the article. The tool cannot identify the important words on which the entire article will be based.
Results round 1 - Email marketing 'ideas' instead of 'mistakes'
Results from another popular AI content tool
Overall results based on the review of these and many other tools – AI content tools are not successful and should not be used for writing purposes.
A quick recap and FAQ
What is AI-generated content produced by AI writing tools?
AI content writer tools such as Jasper, Ryter and Copy.ai and ChatGPT by OpenAI use pre-existing information from other online sources (which they have permission to) and the information they have already been trained with. Based on it, the tools format and restructure the content and provide it as output to the users.
Is AI-generated content good for SEO, and can Google detect AI content?
What Google says about AI/fake content use
As per my observation and testing of many AI content writer tools, I would not rate it as ‘good .’It lacks depth and is highly robotic. Any reader or information seeker who knows what good content is or holds good language skills can easily identify AI-generated content. Google’s advanced systems can also easily identify such content. Google has clear policies against AI content, so it does not value it in search engine rankings.
How do I know if my content is AI-generated?
There are many tools like ChatGPT Zero and others that can identify AI content. Since the launch of ChatGPT new version, there has been a flurry of AI content detector tools released. I have tested many of them, but none can be called good tools for identifying fake content.
As mentioned earlier, it is easy for a good language skills holder or a good reader to identify such content. If you are a website owner, a digital marketer, a content writer or an SEO expert, you will surely have no problem spotting fake content generated by AI content writing software, including ChatGPT.
What is the best AI content writing tool?
Right now, AI content is in an evolving stage. Software companies love to market everything as disruptive and create a market with their aggressive strategies. Anything that addresses a pain point takes the fancy of users. This is why there is a lot of hype around content writing software. But in reality, none of them offer anything new or deliver what they promise. They are just improved versions of paraphrasers that have been in the market for a decade.
Who owns AI generated content?
It’s complex. Some writers cite ChatGPT as the owner of the content. But are AI writing tools and content software the rightful owners? Nothing can be said with great clarity, as too many clauses, sub-clauses, and judgments around this subject need to be studied. However, this Forbes article by Joe McKendrick provides useful information on who owns the AI content.
Does Google penalize AI content?
Yes, it does. As per Google, AI content is against the webmaster guidelines, and all AI content will be treated as spam. Google has always valued high-quality content that adds value to users' searches. It values organic content generation methods, which is what AI content lacks. So Google is clear that AI content does not fit into its search engine ranking policies.
