As content creators or content marketing managers, your role is crucial in eliminating the spread of misinformation in AI-generated content. Your efforts in vetting sources can significantly reduce the risk of inaccuracies or fabricated information.
The Rise of AI in Content Creation
AI gives brands the capability to create content at an incredibly rapid rate. This increased accessibility introduces new challenges with accuracy.
The Speed vs. Accuracy Trade-off
AI-driven responses are designed to prioritize speed and efficiency, delivering information to requestors almost instantaneously. AI platforms are programmed to analyze vast data sets and generate content within seconds. But unsurprisingly, producing content using shortcuts can come at the cost of factual reliability.
AI fills in gaps in its “knowledge” with educated guesses. When it doesn’t “know” something, AI algorithms typically rely on mathematical probability to provide a likely answer. But if we don’t fact-check that and just accept the answer as truth, we are going to run into factuality issues.
Risks of “Hallucinations” in AI Outputs
AI-generated content can contain many inaccuracies, ranging from misinformation to false citations and non-existent sources. When AI produces such content, it is referred to as an ‘AI hallucination.’ This term describes the generation of false but plausible information.
Hallucinations occur for a variety of reasons, including overfitting and training data bias or inaccuracies. AI developers lack transparency in disclosing training data, making it hard to discern exactly what the programs are exposed to. The algorithms that power AI are optimized for relevance and coherence, not fact-checking. So, content can sound very authoritative but be completely incorrect. AI hallucinations can sometimes be difficult to identify because they often appear well-written. A reader uninformed on the topic might not question the information.
For instance, AI-generated content with hallucinations could lead to legal issues, depending on the misinformation. An extreme example would be if AI were used to write instructions on a medical label. Imagine if it filled in a knowledge gap with a guess about dosage – this could cause great harm in practice.
AI hallucinations can also have a negative effect on SEO, as search engines prioritize reliable and accurate content. If AI-generated content is not properly vetted, it could lead to a decrease in search engine rankings and visibility.
Why Vetting Sources is More Important Than Ever
As AI-generated content becomes increasingly commonplace, the need for rigorous source verification is increasing. Otherwise, we risk widespread misinformation in available content and research.
The Cost of Misinformation
Numerous documented examples exist of how unverified information can have potential repercussions. On a lower stakes end of the scale, content creators risk damage to credibility, audience trust, and SEO rankings.
However, there are also higher stake risks that could have a more widespread effect.
For instance, when the errors stem from bias. Data used to train AI models may contain information with biases related to race, gender, and political persuasion, among others.
AI-generated content may then amplify existing biases, both on an individually harmful level and regarding customer preferences or market trends. Any data presented in AI-generated content needs to be sourced and verified to avoid perpetuating accidental bias by content creators.
Protecting Brand Integrity
Brands must carefully vet information to avoid AI-generated inaccuracies. Failing to do so can result in a loss of authority.
A lot of consumer behavior is based on emotional trust. While many are excited about the capabilities of generative AI tools, there is still a lot of trepidation surrounding them. Any misstep on the part of brands in using AI will cause long-lasting damage to reputation and trust among consumers. By using AI responsibly—including using transparency and verifying information provided by any AI-generating platform—brands minimize this risk.
Best Practices for Finding Reliable Sources
It’s also important to locate credible and verifiable sources to boost the reliability of AI-driven content. There are many best practices for locating reliable sources, but make sure at least some of these are being incorporated into the process.
Seek Reputable Publications and Primary Sources
Start with trusted repositories and databases as primary sources. Use reputable platforms like Google Scholar and JSTOR for academic and scientific references. For industry-specific publications or databases, websites like Statista, Bloomberg, and Adweek are good places to begin.
Favoring established publications, academic journals, or primary sources over lesser-known sites or AI-generated content is safest.
Use Cross-Referencing
Don’t just check one source. Compare information from multiple (reputable) sources to confirm consistency and avoid relying on a singular source. Cross-referencing helps identify discrepancies.
Assess if the source cites credible references, reinforcing its own reliability. Double-check that all referenced data is traceable back to original studies, real reports, or verifiable statistics.
Techniques for Vetting Sources as AI Saturates the Internet
Several other methods can be used to evaluate source credibility when using AI-generated content. As AI saturates the internet, it’s important to make sure that the sources used are not also AI-generated.
If they are, you could accidentally contribute to an AI-generation feedback loop. In this loop, the information you are trying to verify might be “confirmed” by other AI content repeating the same misinformation.
Fact-Checking Against Authoritative Sources
As an extra step to protect against misinformation in AI-generated content, employ lateral reading. Step outside the original source and investigate its credibility. Explore who produced it, their expertise, and any potential biases they might have. Doing so will help detect inconsistencies and ensure the information you’re relying on is well-supported and trustworthy.
Assess if the source itself cites credible references, reinforcing its own reliability. Double-check that all referenced data is traceable to original studies, real reports, or verifiable statistics. Verify AI-provided facts against known or authoritative databases or websites. Lateral reading can help you figure out which sources are the trusted field experts. Analyze everything from the publication’s reputation and authorship to the inclusion of transparent references.
Evaluating Author Expertise and Source Relevance
When evaluating authorship, assess the following criteria to ensure the author is credible and if their content is relevant to the topic:
- Credentials and expertise: Check the author’s professional background, academic qualifications, or industry experience to confirm their authority on the subject.
- Relevance of other work: Look at whether the author has published other works in the same field or has previously demonstrated expertise on the specific topic at hand. Look for previous articles, books, or papers authored by the individual or organization to gauge the depth of their knowledge.
- Affiliations, bias, and objectivity: Identify the organization or institution the author/content is associated with. Consider the tone and intent of the content. Ensure the author presents balanced, fact-based information rather than opinion-driven or biased narratives.
- Evaluate the source’s authority: Research the authorship and the organization behind the content. The most reliable sources will be field experts, organizations with authority, or peer-reviewed publications. Also, verify the date of publication to ensure the source is up-to-date and relevant.
- Leverage digital tools for source verification: Platforms like Grammarly Insights can help identify biased or unsupported claims. Grammarly and Plagiarisma can also detect plagiarism to help ensure originality.
- Website metrics and reputation: Tools like Moz or Ahrefs can analyze a source’s domain authority. Avoid sources with excessive ads or clickbait headlines; these typically indicate low reliability.
A combination of some or all of these will provide a clearer picture of an author or source’s qualifications.
Spotting Red Flags in AI-Generated Information When Vetting Sources
There are generally some easily spotted, common red flags in AI-generated content. Easily recognizing these will help identify potential inaccuracies when vetting sources.
Watch for Generic or Ambiguous Statements
Generic or ambiguous statements often signal unreliable AI-generated content because they lack the specificity and depth of an expert on the topic.
Vague wording, such as “many experts agree” or “research shows,” without source citations or verified numerical statistics, are also indicators. Broad claims and overly simplistic explanations can also indicate a lack of research or critical analysis, pointing to potential AI content.
Reliable content is typically characterized by detailed, actionable insights supported by verifiable data and precise language. For example, I researched a question and wrote an answer. I then asked AI (Chat-GPT) the same question.
Here are the results:
|
What percentage of the population is left-handed? |
|
| My Response |
Chat GPT Response (copied and pasted exactly as provided) |
| Researchers estimate around 10% of the global population is left-handed. There is also around 1% of the population who has no clear preference and are therefore considered to be ambidextrous.
This number has increased in recent decades, mostly likely due to the dismissal that left-handers were historically considered “evil” within many cultures. Children are no longer being forced to be right-handed. Experts are still unsure about exactly what causes right-handedness vs. left-handedness. However, ongoing studies are always conducted to try to determine the cause. One such study was performed on the protein-coding part of the genus. They found that rare coding variants within TUBB4B had a 2.7 times higher rate in left-handed participants than in right-handed participants. |
Approximately 10-12% of the global population is left-handed. This proportion has been fairly consistent across different studies and regions. Left-handedness appears to occur slightly more frequently in males than females.
While the exact reasons for left-handedness are not fully understood, it is believed to result from a combination of genetic, biological, and environmental factors. It is not tied to any particular demographic or cultural group and occurs naturally in humans worldwide. |
Notice how the human-generated response uses trusted sources (BBC, National Library of Medicine, etc.), critical thinking, and verifiable numbers. The AI-generated response provides no source for the numerical statistic and uses broad, generalized statements.
Spotting these kinds of anomalies will narrow down what is likely AI-generated content versus reliable information.
Recognize Patterns of Overly Positive or Overconfident Language
AI models, as they are not human, have no reason to be uncertain of themselves. Therefore, it may present information with a tone of certainty, even when the underlying data isn’t verified.
This phenomenon is particularly notable when the model encounters ambiguous or incomplete contexts. Since the system can’t think critically, it fills in the missing context with the most probable answer.
Watch for exaggerated or definitive yet unproven claims. AI might describe concepts or data as absolute without acknowledging potential exceptions or nuances. It also might fail to recognize uncertainties or different interpretations of the information.
The tone of any AI-generated writing may also overstate the reliability of a source or data. For example, “This study proves…” when the study only suggests a possible correlation.
But most importantly, use your own human judgment. If information or a source seems questionable, biased, overtly generic, or just generally “off,” it’s probably worth some extra research.
Vetting Sources: The Human Role in Safeguarding Content Integrity
AI is a powerful tool, but it’s only as reliable as the humans guiding it. Take your time when vetting sources to prevent misinformation and set the standard for content integrity.
About the Author: Rhianmôr Thomas
I am a freelance writer living in Los Angeles. With an undergraduate in Screenwriting and an MBA, I write for a variety of mediums.
Most days, you can find me working on client blogs, articles, website copy, product descriptions, and more. Currently, I write on a range of topics, specializing in B2B & B2C marketing, fashion, film, social media, higher education, and ecommerce.
I always aim to write copy that makes the reader think – about a better way to do something, the future of an industry or just about their opinion on the topic. When you can blend the goals of the client with the ability to make the intended (or unintended!) audience think – that’s successful copywriting.
To learn more about Rhianmôr, check out her writing profile here: Rhianmôr Thomas