How to Detect AI-Generated Content

Ever since ChatGPT launched, marketers have been using it to scale their content creation.

For SEO in particular, AI-generated content seems like an excellent way to speed up the content marketing process and significantly cut costs. Unlike the black hat article spinners of the past, ChatGPT (and similar tools like Jasper and Google Bard) produce intelligent, original content that reads as if it were written by a human. To the untrained eye, it’s hard to tell the difference.

So, why would you want to detect content created by artificial intelligence? If the output reads well, what’s the problem?

The issue boils down to quality. AI-generated content might read well, but it lacks depth and nuance. It can rank well in the search engines, but it’s not likely to provide adequate value for a visitor unless the topic is something extremely basic, like instructions for removing a carpet stain or directions to a business. Unfortunately, if you publish content made just for search engines, it will be considered spam by Google.

What does Google say about AI-Generated content?

You may have heard that Google considers all automatically generated content to be spam. This was their position at one time. 

In an Office Hours video from April 1, 2022, John Mueller clarified Google’s official position on automatically generated content, and stated, "If you're using machine learning tools to kind of generate your content it’s essentially the same as if you’re just shuffling words around… for us it's still automatically generated content and that means for us it's still against the webmaster guidelines, so we would consider that to be spam.

"Google’s official stance has since changed. On February 8, 2023, Google announced that AI-generated content is not against their guidelines unless it’s created to manipulate search rankings. Useful content created by automation is perfectly acceptable.

6 Ways to detect AI-generated content

Despite the well-written nature of generated content, it can be detected easily. 

AI-content-generating algorithms are basically a glorified form of predictive text, where the system knows the words most likely to come after one another. To detect AI, this process is reverse-engineered, where the system predicts the most likely word to come before a certain word.

Even though AI content is allowed, you might want to avoid publishing it to your website. If you hire writers to create content for you, and you’re not sure if they’re using an AI tool, here’s how to detect it.

1. Get GPT-2 Output Detector

One of the best AI content detectors is GPT-2 Output Detector. It’s free and you don’t need to register for an account. This particular tool is highly accurate and allows you to paste in more copy than other tools.

This tool uses a scoring system of real/fake, so the higher your content scores as real, the more likely it was written by a human. This tool doesn’t seem to have a character limit, but the more content you paste, the longer it takes to analyze.

2. GPT-Writer

Another great tool to use is GPT-Writer. Although some people have said it isn’t as accurate as other tools, it depends on the subject. For some content, this tool scores human-written samples as 100% human-generated content, while other tools score the same content at 98% human-generated.GPT-Writer limits samples to 1500 characters, but that should be enough to get the information you need. If you aren’t sure, run multiple samples.

3. AI-Detector – Content at Scale

AI-Detector – Content at Scale is another highly accurate tool and has a 2500 character limit, which is more than enough to get a decent analysis.

4. Honorable mentions

There are a few tools that don’t make the cut for reliability, but you may find them useful. The first one is AI Detector Pro (the free version). 

Your input is limited to 200 characters, which may not tell you much. However, the paid options might be more beneficial. Even so, with accurate free tools like GPT-2 Output Detector, there’s really no reason to pay for a tool unless you need the extras, like reporting.

Originality AI is a plagiarism and AI detector, so you get the best of both worlds with this tool. However, some people have said it was really easy to alter a little text to bypass the detections and get a human score.

5. Plug your suspected content into AI tools

Often, when people use AI tools to generate content, they keep the headings intact or alter them by a word or two. You can reverse engineer an article by plugging one heading at a time into some AI tools to see what content is generated. If the content was written using the same AI tool you’re using, each heading will likely generate text that can also be found in the body of the article you’re investigating.

For example, say the article you want to verify contains the heading, “Why is AI content bad?” If you type this heading into Google Bard, you’ll get several bullet points in response. If the content of those bullet points also appears in the article either word-for-word or very similarly, you can be fairly certain it was written using AI.

You can also plug in full paragraphs from the article in question. If the AI tool returns content that is also found in the article, it is more than likely AI-generated.

6. Use manual detection methods

Tools are great for detecting AI-generated content, but you can also use manual methods. Here are some things to look for:

• AI-style quirks. Sometimes AI-generation tools have an identifiable style that will be noticeable. For instance, Google Bard often provides both pros and cons to a prompt and concludes with a neutral statement. Other AI writing tools might repeat certain words within the text more frequently than a real human.
• The absence of human error. No matter how good a writer is, they will make mistakes that some AI tools won’t. For instance, some people are too generous with commas and misuse semicolons. Other times, writers end sentences with a preposition. Although it’s usually acceptable to end sentences with words like “on” when it’s part of a phrase, for example, “I lost the paper it was written on,” it’s not a format AI tools generally use.
• Too many bullet points. Bullet points are an excellent way to break up text, but if your article looks like more of an outline rather than content with a few bullet points here and there, it could be AI-generated. AI writing tools frequently generate content that contains up to 75% lists.
• A lack of contractions. AI tools are specifically programmed to produce formal content. Contractions make content more informal, and although that’s often better for readers, AI tools don’t agree. You can get an AI writing tool to use contractions and an informal tone, but you have to specifically ask for that. Content that carries a formal tone without personality is a candidate for AI-generated content.
• Obvious factual errors. Content generators are known for making mistakes with simple facts and humans don’t always correct them. Strange errors that seem too obvious to be made by a human are likely a sign of AI writing.
• No opinions or personality. Not every piece of content needs an opinion, but there should be some traces of personality. AI-generated content is generally objective and dry, devoid of personality.

Should you avoid AI content?

Although you don’t need to worry about AI-generated content hurting your search rankings simply for being AI, there are several reasons to avoid it:

• It’s too generic to get decent conversions. General content will only convert people who are ready to buy anyway. Avoid using AI to create sales pages. If an AI-generated page is already converting, your conversion rate will increase when you have a human copywriter rewrite the page.
• It’s not engaging. AI-generated content reads like a human wrote it, but because it lacks personality, it won’t capture and hold attention like professionally-written copy.
• Can a human do better? If you need content for a page and a human can do better, choose the human writer.
• It lacks personalization and creativity. AI content is bland. When the goal is to provide value to visitors, the more detailed you can get, the better. AI can’t create nuanced content that really speaks to a visitor’s pain points and dilemma in-depth. Without this, you won’t see high conversions or repeat visitors.

There are also some good reasons to use AI-generated content:

• AI is great for generating technical information. Content that isn’t supposed to sell a product or engage a reader is perfect for AI, like sports game scores, recipes, directions for how to do certain projects, and even code.
• Quick descriptions. When you just need a short description for something and you can’t think of the right words, AI can give you exactly what you need.
• Writing emails. You can get some great email templates for common situations using AI prompts.
• Social media posts. When all you want is quick content to publish to social media, AI is a good resource.
• Article outlines. If you’re stuck on ideas, you can get your articles written much faster by using AI prompts for an outline.

How you shouldn’t use AI-generated content

There are two main ways to avoid using AI content: to spin scraped content and to fill space on your website.

Scraped content is spam and spinning scraped content is plagiarism. Google has algorithms that can detect scraped content that has been altered by an AI tool. Publishing scraped content has always been considered spam, and using ChatGPT or any other tool to alter it is still considered plagiarism. In other words, don’t steal other people’s content and use AI to rearrange it to make it look original because Google will know.

Another use case to avoid is filling space. If you’re generating long articles to populate your blog or other website pages just to fill it out, you’re probably not providing value to visitors.

Back in December 2022, Google announced a Helpful Content system update that promised to detect and suppress content made for search engines and promote websites designed for humans. 

Many people noticed a drastic drop in rankings after this update, even sites with high-quality human-created content. It’s too risky to use AI tools to populate your website for the sake of filling space. You’ll waste your time, money, and you probably won’t rank.

Use AI tools with caution

Poor-quality content that doesn’t provide value isn’t going to rank whether it was created by a human or ChatGPT. Google can tell the difference between AI-generated content created to manipulate search rankings and content that exists to provide value to web visitors. If you choose to use AI tools like Jasper, ChatGPT, or Google Bard, be responsible. 

There’s nothing wrong with AI-generated content as long as it is helpful and provides value to your visitors. As long as you follow Google’s E-E-A-T Guidelines (Experience, Expertise, Authoritativeness, and Trustworthiness), you can rank AI content in the search results just like any other page.

‍