Why You Need to Edit Your AI Content

A guide to proofreading and editing your AI-generated content.

August 30, 2023

By Wolfestone

AI tech like ChatGPT and AI content generators seem like the best thing to happen to content creators since the invention of the ‘Share’ button on social media. And that might be true.

After all, AI makes it easy to generate articles, emails, social posts and more in seconds and at practically zero cost.

At first glance, the content seems pretty good too. It’s much better than much of the content produced by content mills and cheap writers.

But a closer look reveals that these AI large language models (LLMs) have their share of flaws. Not only are they rife with inaccuracies, but most of the content produced is unoriginal and unengaging—especially to niche audiences.

Don’t worry; you can still use AI to save time and money. But human edits are essential to creating original, engaging content that maintains its ranking in the search results.

We’ll explain why and show you some examples below.

Why Human Edits are Essential to AI Content

AI content reads smoothly and sounds good. But it doesn’t take long to realize that there are many problems with LLMs’ output. These are some of the most serious.

1. AI Makes Mistakes (Constantly)

AI content is not trustworthy. And that’s not just our opinion. It’s the belief of Sam Altman, the CEO of OpenAI (the company that developed the technology behind nearly every AI content generator, including ChatGPT).

"I trust answers generated by ChatGPT less than anybody else on Earth," Altman said at a recent event.

Altman knows that LLMs are not designed to provide factual answers. They’re designed to satisfy a prompt by the user. While OpenAI has attempted to put some limitations on the AI’s likelihood of giving dangerously incorrect info, it still gets it wrong all the time.

Take a look at this example:

In the example, ChatGPT correctly states that the user’s information is false. But it goes on to provide a completely false explanation.

As time goes on, won’t AI and LLMs get smarter? Not necessarily. In fact, a recent study by Stanford University and UC Berkeley has found that ChatGPT’s accuracy is getting worse.

Publishing outright false information has huge implications for your business. Not only will customers lose trust in your brand, but in some cases, it could lead to legal trouble.

2. AI Has Bias

Does AI have a voice all its own? New studies have found that LLMs do, in fact, have biases that show through in their content. OpenAI and Meta’s technology were quick to take stances on issues like feminism and democracy, with GPT tech showing a liberal bias and Meta’s LLM leaning to the right-wing.

Today’s political climate is unstable enough. You don’t need AI inserting inflammatory viewpoints into your content—or even suggestive viewpoints that could be misinterpreted by your readers.

Because LLMs are trained on data sets based on human sources, researchers believe that AI content generators will never be without some form of bias.

If you want to ensure your brand’s tone, values, and mission remain aligned with your content, it’s critical to have a human edit your content for these nuances.

3. AI Invents Information

Referred to as “hallucinations,” AI is prone to fabricating information to satisfy the user’s prompt. This is rarely obvious. The AI is very good at confidently presenting information as fact.

Take a look at this ChatGPT answer:

It sounds 100% legitimate. But in fact, the information in the “study” published by the Journal of the American Academy of Dermatology has been invented.

AI can’t be trusted to fact-check itself, which is why human editors need to verify every source for accuracy.

4. AI Can’t Localize Content

All of the AI content tech we use today was developed in the US, and it’s pretty stubborn about using US English. That means if you want to create content that will appeal to English speakers in the UK, South Africa, Australia, etc., you’ll need a human editor.

Check out what happens when we try to get ChatGPT to write in UK English:

This simple example shows that ChatGPT can’t even use the proper UK spelling of a word like “colour.”

In other cases, the AI made some UK-appropriate changes, like changing the word “trunk” to “boot” (as in, the trunk of a car). But it went way overboard and drastically changed the tone and register of the passage.

It’s also noteworthy that the AI cannot convert currencies accurately.

5. AI Content Quality is Too Low for SEO

Google is not against AI content as long as it meets its standards of quality known as EEAT (Experience, Expertise, Authoritativeness, and Trustworthiness).

But it’s become clear that raw AI output does not meet these standards. In fact, Entrepreneur and Forbes recently argued that an overreliance on AI content will not help you rank.

There are a few reasons for that:

AI content is not original: Google rewards content that is relevant and original. Good content adds something new to the conversation that appeals to a specific audience. All AI content is based on previously written content, so it can only rehash what’s already been done.
AI content can’t communicate effectively: Human writers and editors communicate with readers by satisfying their search intent and understanding their needs. AI will never be able to write to an audience.
AI content lacks human experience: Experience is a recent addition to EEAT, and AI doesn’t have it. You need human editors to infuse human experiences into your content. This also gives your content more authority (another EEAT metric).
AI content is not relevant: OpenAI’s “brain” doesn’t know anything past 2021. This can cause it to hallucinate or provide false answers in content that requires modern context. This could lead to penalization by Google for publishing inaccurate or outdated information.

There are stories of SEOs ranking their unedited AI content on the first page of the search results, but these successes are often shortlived, as Google quickly identifies the content as having “little-to-no value.”

If you want your content to achieve longevity in the search results and protect your brand from backlash, editing is a must.

6. AI Content Doesn’t Engage Readers

If your goal is to convert readers to customers (or at least get them to stay on-page for longer than 10 seconds), raw AI content won’t accomplish that.

AI content is very basic, and it can’t appeal to a specific target market or audience. If real human experience and expertise don’t come through in your content, readers will not gain the confidence and trust needed to take action (such as making a purchase).

Marketing and SEO experts have said this time and time again.

Elisha Gada, Marketing Manager for Meta, said, “AI tools could be a good starting point for ideation. However, for any professional content creation, it could be hard for AI to replicate brand voice…. Use it as a starting point but don’t forget to add your personal touch to it.”

Learning prompting skills can help improve AI output. But to add personalization, experience, brand voice and expertise (aka engagement), you need to humanize your AI content with proper editing.

Take a look at two introductions to the same article. One is raw AI output and the other is human. Which one do you find more engaging?

The Consequences of Publishing Raw AI Content

AI’s benefits are so enticing that many businesses have been lured into going all in with AI—and much like Icarus flying too close to the sun, the results have not been pretty.

Major tech media outlet CNET decided early on to start generating its articles with AI, only to discover soon after that over half of its articles contained factual errors. CNET had to apologize and go on damage control, reviewing dozens of articles and issuing corrections.

The fantasy publisher Tor also faced backlash for using AI-generated art as a cover for a new novel. While this wasn’t illegal, it was tone-deaf to fans who respected and supported the company’s promotion of artists.

Other businesses decided to publish raw AI output in order to rank faster on Google. But their pages were soon deindexed en masse.

Harnessing the beneficial aspects of AI can help businesses improve their processes and cut costs. But they must be aware of AI’s numerous shortcomings.

Raw AI Output is the New Floor

AI content might not seem that terrible. In fact, it might seem pretty good. But now, literally anyone with a phone produce that level of content quality at zero cost.

Moz CEO Rand Fishkin explained that “AI content is the new floor.”

Technology like ChatGPT has raised the bar on overall content quality. But that also means that anyone publishing raw, unedited AI content today is producing the lowest-quality (and lowest-effort) content possible.

That’s why Fishkin also said, “If your content isn’t better than what AI can produce, it’s not worth making.”

Luckily, you can still harness the advantage of AI while producing high-quality content. By using AI responsibly and developing a reliable human editing process, you can engage readers and maintain your position in the search rankings.

Wolfestone Group: AI Editing and Proofreading Services

Wolfestone Group has been a leader in linguistic services for nearly 20 years, and we pride ourselves on being at the forefront of AI technology. We employ a team of trained AI proofreaders and copy editors who specialize in turning your raw AI content into accurate and engaging copy.

Our AI copy editing team ensures that your AI content is: