Artificial intelligence is a godsend for many —it can generate blog title ideas, fix your grammar, simplify your sentences, summarize long-form content, and so on. But it can also be a backstabbing friend, especially when its ‘hallucinations’ hit.
In this article, we’ll outline steps to fact-check AI-generated content, explore fact-checking techniques, and how you can find reliable and good sources without breaking a sweat.
Why AI Can Be A Tricky Friend
We’re not going to sugarcoat it — AI can generate false information.
Don’t believe us? Here are some numerous AI mishaps:
- Google’s AI overview said to eat glue and rocks
- A chatbot that lied about a discount
- Two lawyers used fake citations generated by ChatGPT
…and that’s just the tip of the iceberg!
So, why is AI giving wrong information? Poor understanding of context, lack of critical judgment, and outdated or insufficient data can contribute to the so-called ‘AI hallucinations,’ a term given when large language models (LLM) like ChatGPT provides fictional answers and fictional sources.
And the consequences of using fake content can be severe. You risk harming your credibility as a writer or, worse, endangering the lives of your readers.
How to Fact Check AI?
The good thing is that you don’t have to abandon AI completely for your writing projects. You can prevent misinformation by tapping into your inner ‘journalist” and do a fact check. Here are four steps you should follow when fact-checking AI-generated citations and content.
Step 1. Be A Skeptical Detective and Question Everything
ChatGPT can be convincing, especially when it generates a well-structured sentence or paragraph to verify the validity of the claims. But don’t be fooled!
Here are some clear identifiers that AI is tricking you:
Unfamiliar or suspicious citations
AI usually forms citations based on patterns in its training data. So, there’s always a high probability that it might provide what looks like a legitimate scholarly source but is actually non-existent or misattributed.
For instance, you might find a citation like “According to a 2020 study in the Journal of Advanced Technology…” only to discover that the study or journal doesn’t exist.
Generic or Vague Language
If the AI offers general statements like “Experts agree that…” without naming the experts or organizations, treat it as unverified. Reliable information always comes with clear, identifiable sources.
Mismatched Context or Logic Errors
Every now and then, you’ll notice AI tools blend information that does not make sense. For example, it might say that “Mars has a breathable atmosphere for humans.” When that happens (and you know it isn’t true), trust your instincts and stand your ground. If you are unsure, you can always fact-check first.
Step 2. Fact-Checking Tools and Resources Are Your BFFs
If you are skeptical about the provided ai-citation or if you suspect the provided information is false, you can use these non-AI sources:
- Google Scholar: Helps you find peer-reviewed articles and credible research papers. Fact-checking AI using Google Scholar is pretty easy, too —just type your query in the search bar, and it provides you with related journals or studies.
- PubMed: Focuses on medical and scientific literature for accurate information on public health.
- Statista: For statistics in a wide range of topics.
- Government websites: This will depend on the topic that you are writing about. For instance, if you are writing about climate change, you can head on over to the United States Environmental Protection Agency and check their resources.
How about Wikipedia? Anyone can edit a Wikipedia post, so you should only use it as a starting point for your research. Go to the bottom of the page (footnotes), in the ‘references’ section, and you will find the post’s sources. Use those and still do a fact-check.
Remember, Wikipedia is a secondary source of information, and you don’t quote the middleman! Always go straight to the source material.
When in doubt, remember that established academic journals, scholarly sources, and peer-reviewed research papers are trustworthy compared to random websites and social media posts.
Step 3. Cross-Reference Like a Pro
Just like an inquisitive house buyer, don’t just unquestioningly trust a listing (or, in this case, a source).
Dig deeper and check different sources before deciding who to trust.
To effectively cross-check;
- Define the claim you want to cross-check
- Use at least three credible sources to compare facts and views
- Confirm publication date
- Find the primary source; it tends to be accurate
- Asses for bias
You should never skip this step because not all sources are created equal. Some may leave out parts of the story, get some facts wrong, or even publish blatant lies.
Before using any source, here are some questions to ask yourself:
- Who wrote it? Is the author or organization a recognized authority?
- When was it published? Is it up-to-date?
- What is its purpose? Is it to inform, persuade, or merely to entertain?
Step 4. Don’t Make Assumptions: Connect the Dots
Lastly, information gathering is not all about collecting facts; it’s more about understanding, comparing, and putting them into perspective.
For example, if you want to explore the effects of drinking coffee and AI indicates that coffee is a “cure” for some diseases, a deeper analysis may show otherwise. You may notice that the study is purely correlational, as coffee may reduce the risk of the disease but not “cure.”.
Don’t Let AI Fool You—Fact Check It!
We won’t deny it — writing content is now easier with AI, but trusting it 100% with whatever it tells you can put you and your work on the line. So, don’t take AI-generated content and citations at face value — you owe it to yourself and your readers to always fact-check.
Let us know in the comments how often you fact-check AI and share your personal stories of how ChatGPT or any other AI writing tool almost got you in trouble.