How To Detect AI Generated Content

“Responsible AI use has the potential to help solve urgent challenges while making our world more prosperous, productive, innovative, and secure. At the same time, irresponsible use could exacerbate societal harms such as fraud, discrimination, bias, and disinformation; displace and disempower workers; stifle competition; and pose risks to national security.”‍–President Biden’s Executive Order on AI signed Oct. 30, 2023

Why Is It Important To Identify AI-Generated Content?

Your content represents your brand. If that content is plagiarized, inaccurate, misleading, or biased, your brand may run afoul of coming laws. President Biden’s October 2023 executive order has made it urgent for marketers to pay more attention to how to detect AI-generated content, as the order establishes the authority of government agencies to monitor and regulate AI development and use. You can expect new regulations from the Federal Trade Commission soon.

Aside from legal concerns, you have your customers to consider. If you publish poorly executed AI content, your audience may perceive you as untrustworthy and irresponsible. And perception is everything in content marketing.

You work extremely hard to build a brand identity that resonates with your target audience. Putting out just one piece of AI generative content that contains inaccuracies can become a brand-destroying PR nightmare with severe legal and financial implications.

Unfortunately, AI generative content often contains hallucinations. AI is also extremely good at creating manipulative content to generate responses. Because of the danger of AI combining these aspects, you need to proactively examine AI content before hitting "Publish." You can get ahead of the coming regulations by developing a framework that will detect AI writing and guide your organization in using AI content creation productively and responsibly.

In a recent Compose.ly webinar, Maddie Lambert — from leading AI detection solution Orignality.ai — discusses how to detect AI-generated content. Lambert also shares highly useful tips on how content marketers can use generative AI writing tools responsibly and safely. The principal takeaways are:

• Content integrity is critical: “It’s important for content to be accurate, reliable, and free of hallucinations or manipulations.”
• Trust, reputation, credibility: “There has been a 400% increase in AI-generated online reviews. [...] People are less likely to make decisions based on reviews. [...] I’m not going to put my trust fully into generative AI.”
• Human brain: “Make sure you're analyzing or reading AI content in a critical way. [...] Leverage AI to improve content but don’t lose human voice.”
• Editorial process: “There are tools that will help you craft content that sounds human, is factually accurate, and is not plagiarized.”
• Expert content review: “It's crazy that somebody can hit 'Publish' on a piece of content that is so wrong.”
• Replacing humans: “These tools are being relied on to replace humans rather than being used to enhance humans.”
• Who should be responsible for AI integrity? “The leadership team needs to have a stance on what's acceptable and what's not with a very clear threshold.” 

Maddie Lambert works with generative AI daily as a marketing and sales executive. Her profound insights into the current state of AI and content marketing come from direct experience. The entire 30-minute webinar with Compose.ly is packed with actionable information.

How To Detect Artificial Content From Generative AIs

“AI content can sound so great and so compelling that it's easy to lose that human touch.”
–Maddie Lambert

Your human analytical ability is your greatest asset when trying to detect artificial content. Once you know the tell-tale signs, you can consistently spot AI-generated content, whether it's an image, video clip, or text.

Discover the Tell-Tale Signs of AI Images

AI-generated images can be challenging to distinguish from actual photos, but some tell-tale signs can help you spot them:

• Blurriness, pixelation, distortion, or mismatched lighting or shadows
• Asymmetric facial features, unnatural expressions, misaligned eyes or teeth, or unrealistic skin texture or hair
• Missing background details, illogical scenes, or incongruent objects or people
• Repetitious patterns like identical faces, clothing, or poses in a crowd or a group of images
• Metadata revealing the source, date, location, or author

Using an AI image detector can help, and companies like Amazon, Google, and Microsoft all have AI image detectors. An excellent standalone tool is Optic’s AI or Not, which gives you 100 free image checks to get started.

Spot AI Videos From a Mile Away

AI videos can be powerful and creative tools, but they can also pose ethical and social challenges. Because of this, you have to be critical and cautious when you share them. Here are some clues to help you spot AI videos:

• Inconsistencies in lighting, shadows, reflections, or colors
• Blurry or distorted edges
• Unnatural movements, expressions, gestures, or lip-syncing
• Mismatched audio and video quality
• Background noise that does not match the scene
• Lack of context, source, or verification for the video or its creator

Taking three actions will help you spot AI videos: 

1. Use metadata analysis to check the date, location, device, or software used to create the video. 
2. Do forensic analysis using a tool to detect signs of AI content and ChatGPT plagiarism. 
3. Fact-check websites or organizations to verify information or claims in the video.

Pinpoint AI-Written Texts Like a Pro

Artificial intelligence blog writing has telltale signs that are easy to spot when you know what they are:

• Logical flaws, factual inaccuracies, or contradictions
• Generic or repetitive language and grammatical structure
• Insensitive or offensive ethical, moral, or social content
• Absence of human-like nuances or opinions
• Lack of specific tone or style
• Omission of readability best practices

It’s critical to hone your AI detection skills because of the difficulties in finding reliable tools to detect AI text. Your perceptions are your best defense against publishing potentially harmful AI hallucinations or insensitive content.

Does a Reliable AI Detection Tool Exist?

No. Not only are AI detection tools notoriously inaccurate, but writers can also use tools like GPTMinus or Undetectable.ai to modify AI-generated content and fool detectors. Writers can train even AI models to copy their style. Perhaps because of these realities, Google Search has decided not to base rankings on content sources but rather on content scores using its E-E-A-T guidelines.

However, even with their limitations, the best AI detection tools can be highly useful to editors and publishers to:

• Analyze sentiment, tone, and style
• Detect plagiarism, paraphrasing, and spinning
• Evaluate the quality, coherence, and readability of a text
• Compare and contrast different texts on the same topic

The most accurate AI detector for you depends on your purpose.

• Originality AI is designed for web publishers who must quickly check for AI-generated content and factual accuracy. It has an AI plagiarism checker, and you can upload individual pieces of content or check a whole website.
• Copyleaks targets enterprises that need to check for AI-generated content and plagiarism. This software can also scan entire websites and enforce enterprise-wide generative AI governance and compliance to protect data.
• GPTZero features a dashboard that makes it easy for academics to check student work for AI use and plagiarism. It’s an AI paper checker and AI detector for essays that can handle document batches.

False positives are a common occurrence with all AI detection tools. Human vetting of flagged material is critical to avoid embarrassing mistakes that can harm reputations needlessly. 

Ensure Human-Written and High-Quality Content With Compose.ly

“AI makes it really easy to create content that’s not great.”
–Maddie Lambert

Microsoft MSN republished an article claiming San Francisco Supervisor Dean Preston resigned from his position after Elon Musk criticized him. The story was completely false. AI had selected and published the content without human oversight.

At Compose.ly, you get great, original, fact-checked content from a vetted human writer backed up by human editors. Compose.ly chooses the top 1% of applicants based on writing skills, industry knowledge, and SEO know-how to get your content ranked on search engines. You can add another layer of security by using one of Compose.ly's subject matter experts to authenticate content accuracy and boost your E-E-A-T score.  

Get started on upping your content marketing game with informative, engaging, human writing. Request a sample now.

‍