You may be wondering about how tools such as ChatGPT and other generative AI tools relate to fact checking. AI can be an immensely powerful tool assisting with a variety of tasks, from brainstorming ideas to summarizing large amounts of information. Fact checking may seem like a natural task for this technology. Like Wikipedia, AI tools can assist with basic background research. However, AI is not a perfect tool. Just as it can be used to fact check and verify the accuracy of information, it can also be used to produce misinformation. Sometimes this is intentional, when AI is employed by groups interested in disseminating false claims. But even where there isn’t ill intent, AI’s algorithms are imperfect, and the information it produces can be inaccurate, or lacking the full context necessary to truly verify information.

Fact-checking and journalistic organizations are working on developing reliable AI tools that can be used for fact-checking, but there is currently no substitute for verifying information from the original source using the tools of fact-checking described in this tutorial. This doesn’t mean you must entirely avoid tools like ChatGPT, but use them with caution (and within the boundaries of your instructor’s policies). Do not automatically assume that the information you receive from AI chatbots is accurate. Think of AI as an assistant you don’t know very well yet. They may be able to help you with your work, but they cannot reliably do it for you. As the technology continues to evolve, and organizations and individuals grow more accustomed to using it, best practices for how AI can be used most effectively and reliably will also continue to evolve. For now, as a fact checking tool, chatbots should always be used in conjunction with additional sources and verification techniques.

Evidence Ratings

You may come across evidence rating systems as part of analyses published by professional fact-checking departments. These ratings more accurately capture the degree to which a fact is accurate in a specific context. Rating scales follow this basic scale:

True -- Mostly true -- Mostly false -- False -- Fabricated

The Fact Checker by The Washington Post created the Pinocchio test which rates evidence on a scale from one to four Pinocchios, with four being the highest rating for inaccuracy. The Fact Checker also makes use of the “verdict pending” rating for facts that are still being investigated. This rating system helps readers better understand how and why a story is accurate.

Lateral Reading

Like a professional fact-checker, you can use the technique of lateral reading to evaluate the sources you come across. Lateral reading is a research strategy that involves examining a source using multiple outside sources. For example, if you are fact-checking a website cited as a source in an article, you can use lateral reading to critically analyze its reliability as a source. Before you do a deep analysis of the site’s contents, look for an About page to get a sense of the site’s mission. Next, open a new browser tab and see what other sources say about the website and its coverage of important issues. These insights will help you evaluate the reliability of the website and prepare you to take a closer look at the way it presents facts.

Continue to take a short quiz to check your understanding of fact-checking.