AI in Content Writing: Strategies to Protect Content Authenticity

08 November 2024

Key highlights

• Gen AI is improving the efficiency and scalability of B2B content production but posing a challenge in maintaining the authenticity of a company’s voice.
• Content writers and content buyers need to carefully assess how to use AI tools and protect themselves from trust and reputational issues, biased and erroneous content, and content getting incorrectly flagged as ‘AI-generated’. 
• Writers need to go back to the basics and humanize their content by avoiding sentence structures, phrases, and words that AI tends to love.

Panchalee Thakur & Avinash Kumar Nayak

The idea of a computer beating a human at chess once seemed absurd. In 1997, IBM’s Deep Blue defeated Grandmaster Gary Kasparov, proving AI’s ability to surpass human intellect.

Today, AI has gone beyond niche skills. It has made its way into content creation, offering new possibilities for efficiency and scalability. But as Generative AI (Gen AI) mimics human writing, content writers and publishers are facing challenges to maintain the authenticity and originality in their work. On the other hand, organizations employing B2B content writing services are finding it tough to establish whether a piece of work was co-created with AI, and if it was, then to what extent. 

Then there is the question of plagiarism. According to a recent analysis by Copyleaks, 60 percent of outputs generated by GPT-3.5 contained some form of plagiarized content. Reports of lawsuits on copyright infringement in AI-generated copy show the risk writers and businesses face.

As a team, we have spent time understanding how best to use Gen AI in content writing to improve our deliverables, without it impacting the originality of our work. We have realized that even original pieces of content produced without AI assistance may get flagged incorrectly as AI-generated. We have worked on deconstructing the ‘AI-generated text score’ tools such as Grammarly and ZeroGPT offer and examine their reliability. 

This blog discusses why certain types of content attract a high ‘AI-generated text’ score. We provide actionable tips for content writers and employers and strategies to improve content authenticity and protect their credibility.

Gen AI in Content Writing is Here to Stay

AI continues to evolve and refine its capabilities, adding a high level of sophistication such as content personalization. This progression could help AI write complex content pieces, such as thought leadership articles, and build creative narratives that resemble human-authored content. 

Check this example of AI-generated text that succinctly describes the data management services of a fictitious company. This is the prompt we gave. “Create a landing page in a paragraph that highlights comprehensive data management services of an organization that focuses on data transformation, data strategy, and data analytics.” The response ticks all the boxes. It reads human-like and authentic, provides information that is useful to a reader, follows a logical flow of thought, and has a tone of voice consistent with marketing copy.

Media houses have found several practical applications of AI in writing. From improving the workflow of content creation to business content strategies, AI tools for writers have already made a significant impact.

• Washington Post designed Heliograf, an AI tool to enable data-driven coverage of major news events and automate news story creation. The tool has allowed them to publish more stories and reach a broader audience, leading to higher readership and engagement.

• The Associated Press (AP) is one of the first news agencies to have deployed an AI tool. Automated Insights creates news stories, without the need for a large team of reporters. 

Experts believe AI will soon be able to make real-time predictions of content trends and user preferences. AI tools will then be able to craft highly targeted content and boost audience engagement. 

Imagine Netflix-level personalization. Netflix examines user data to forecast preferences and generate personalized content that appeals to a specific demographic. Now, with machine learning methods, companies can create similar models to predict which topics or content formats will have a higher possibility of engaging a specific user group.

It is Not All Good: Mind the Gap

Gen AI tools can assist in writing. But the technology has not yet reached the level where it can replace human creativity, emotional depth, and thought. AI has its limitations, and both content producers and buyers must apply caution and balance AI with human creativity.

Trust issues and reputation risks: Content writers, reviewers, and buyers today rely on AI writing and detection tools, such as Grammarly, ZeroGPT, GPTZero, and Copyleaks, to verify whether the content is human authored. But these tools are far from reliable, with human-authored content often being incorrectly labelled as AI-generated. This creates mistrust between content creators and clients, affecting writers’ morale and harming business relationships.

A recent LinkedIn post by a writer from Washington who lost a client after being accused of using AI tools to write an article shows the risks involved. The client flagged her work as ‘76 percent likely to have been written by AI’. The client used AI detection tools like GPTZero and Copyleaks and decided to terminate the contract with her. The writer’s defence that she had authored the article without AI help was rejected. Following this, she now adds a clause to her contract addressing issues around AI usage and detection.

Erroneous and biased information: Many AI tools lack the ability to judge the accuracy of the information they collect from the web. The data they use may be from unreliable and biased sources, increasing the risk of errors and inadvertently furthering stereotypes and prejudices.

The potential SEO impact of AI content: Google prioritizes quality content with up-to-date information that follows E-E-A-T guidelines. E-E-A-T that stands for experience, expertise, authority, and trustworthiness. It determines how Google judges the quality and legitimacy of an organization’s web content. Google wants the content to be written for humans and not for algorithms. 

Content generated using AI may not follow the standard  E-E-A-T guidelines which can potentially bring down a website’s search ranking. Moreover, AI adds the risk of duplicate content. Thousands of writers use similar prompts to generate an article and directly incorporate the AI responses into their content. Search engines are likely to flag such articles as unoriginal or lacking depth and expertise.

The Muddled World of AI Detection Tools

AI detection tools claim to accurately identify whether a piece of content is AI-generated, human-written, or has a mix of both. However, our experience shows that the claims do not hold up in reality. According to Originality AI, an AI content detection tool, “The world needs reliable AI detection tools, but no AI detection tool is ever going to be 100% perfect.”

We tested an article–written without any AI assistance–on three different tools and got varying scores. While the article got a zero percent ‘AI-generated text score’ on GPTZero, it received a 70 percent score on ZeroGPT, and 100 percent on Grammarly.

Strategies for Content Buyers and Reviewers

Organizations that employ writers and content writing agencies need to find common ground by defining acceptable uses of AI tools and minimize the risks.

Create a space of trust and open communication: Conduct due diligence on which AI detection tool is the most reliable and stick with that since no two tools are likely to produce a similar score. Next, decide on a score that is acceptable and communicate that to your writers. Have an open discussion with the team on what is acceptable or not. For example, you may encourage them to use AI to conduct research or write email subject lines. But openly communicate to your writers that you will not accept articles co-written with AI. 

Look for style consistency: AI is not yet that sophisticated as to mimic someone’s writing style. And that can be your yardstick to judge the originality of a piece. Look at previous articles to see if a piece of work reflects the writer’s style and your organization’s style guidelines.  

Add your organization’s flavor: Content created on the basis of secondary research alone lacks depth and individuality. Give writers access to your subject matter experts (SME), case studies, and other in-house resources. Assets that include SMEs’ insights and showcases the organization’s own experience will create better engagement and shine through among a sea of similar content on the web.

Check the research trail: Ask your writers to document their research process. Writers usually keep a record of notes or bookmarks of websites and videos they have checked online. That is a strong demonstration of the effort gone into collecting and synthesizing information.

You can also request for their rough drafts, which can be shared in case of any doubts about human authorship. Ideas develop over a couple of drafts and show the natural progression in a writer’s thought process. For example, Google Docs stores all the versions of an article and is a great place for such checks.

Strategies for Content Writers

To ensure legitimate human-written content is not labelled as AI-generated, writers need to employ a number of tactics.

Bring out the storyteller: AI can be good in presenting facts in a coherent manner, but it lacks the ability to narrate stories. A recent study published in the journal, Scientific Reports, compared how humans and Large Language Models (LLM) like ChatGPT retell stories. It showed that humans use new words and concepts with each iteration marked by emotional depth, whereas LLMs rely on probabilistic reasoning, lacking deep domain knowledge. 

Writers bring their content, including B2B thought leadership work, alive with a creative narration. For example, use a client story or a hypothetical scenario to set the context in the introduction instead of high-level thoughts or data points.

Avoid cliches and jargons: Writers have over the years been conditioned to write robotically, liberally using jargon and overusing certain words. Hence, an AI detector cannot differentiate between a copy written by machines and humans. 

We have found certain words that trigger the ‘AI-generated text score’. Some words to avoid: foster, embarking, delve, tapestry, realm, resonate, additionally, navigate, holistic, revolutionize, and landscape. AI also loves adjectives, much like writers. Some adjectives to avoid are adept, ever-evolving, innovative, robust, seamless, thought-provoking, and transformative. No surprises there. It is time for writers to end their love affair with cliches. 

Here is an example of text that is likely to get flagged as AI-generated because of the number of cliches in it.

Use simple sentence structures: AI text detectors tend to flag content pieces with rigid sentence structures. Be mindful of wordy and complex sentence structures and unusually formal tone in text. Use a mix of short and long sentences in a paragraph.

Here is an example of text with its corresponding score before and after sentence restructuring.

If you are struggling to bring down the ‘AI-generated text’ score in spite of rephrasing the text, there is AI to help you humanize your content. Humanizer.org helps detect clichés, so you can rewrite the text to sound more human-like.

Is AI Killing or Reviving Writing?

The roles of different stakeholders in the content creation process are today getting redefined. Content creators, reviewers, and those that employ content writing agencies need to make themselves aware of the challenges that AI poses. They must be open to experimenting, learning, and adapting at AI evolves.

For example, when time is a constraint, writers can take the help of Gen AI to conduct their preliminary research. Microsoft Co-pilot is a great tool to gather all the relevant information for an article, including web links. But they must use their own creativity and knowledge to synthesize the information and write the article. With this approach, writers can speed up work yet retain the authenticity of their content.

Ironically, the emergence of AI is forcing writers to go back to the basics. There is a need to make a mental note of words and phrases that are being used too often, and look for alternatives. Writers must avoid complex sentences that make for a poor reading experience. Short and crisp sentences deliver ideas and thoughts with better clarity and have a greater impact on the reader.  

So, is AI harming content writing? Or is it challenging writers to sharpen their skills? The jury is still out on this debate.

Want to work with us to create original, authentic content for your organization? Write to us at info@purpleiriscomms.com.