Fact-Checking Trump's Inaugural Address
Or: A hands-on guide to doing your own digging with ChatGPT.
Last week, I wrote about the need for each of us to be our own fact-checker in the Disinformation Age. Today, I’m taking my own advice and sharing my experience as a practical example. To do this, I provided the text of Donald Trump’s speech - delivered at the Jan 20, 2025 inauguration - to ChatGPT in order to fact check every testable claim in it. I’m writing this in post in parallel to the fact checking itself, so I’m as likely to be surprised by the results as you are.
But first, a bit about me…
I don’t want there to be any confusion, so let me be up front about my background and bias. I’m a heterosexual white male between the ages of 50-55 living in the US. I was raised in the upper middle class by a corporate middle-management dad and a stay-at-home mom, both lifelong Republican voters, who were married for more than 50 years, until my dad died a few years ago. They loved each other like crazy and they taught me to believe in myself and work hard. They gave me everything a growing boy needs to be a successful adult.
I spent nearly 6 years getting a 4-year Bachelor of Arts in Political Science with a minor in English, but I took advantage of being in the right place at the right time to get in on the ground floor of the Information Age, because money. I’ve been working in information technology for almost 35 years, clean and sober for almost 17 years, and married for just over 13 years. My wife and I have no children, so we’re enjoying all the fruits of an 88th percentile household income: our bills are paid, our needs are met, and our eventual retirement is as secure as anything in this uncertain world.
Despite - or perhaps because of - all this privilege, my political views are so far to the left that you can barely see the US two-party system from here. I’m a proponent of stateless collectivism. My first website (circa 1994) was overthrow.net. Maybe it’s because I was raised with the freedom to choose my life-stance (Secular Humanism), or maybe it’s because of all those sweet, sweet books my ivory tower commie professors got me hooked on. Regardless, I want you to know up front that my life is the inverse Alex P. Keaton story.
If you had to Google that reference, kindly get off my lawn.
All of this to say: there is no way in hell I was gonna sit through Trump’s speech. Listening to that man ramble makes me simultaneously angry at and embarrassed for our country. But I’ve been taught in many contexts - meditation, recovery, comic books - to be wary of my strongest emotional responses. When something gets me amped up, positive or negative, I get impulsive and make bad decisions. In those moments when I’m disturbed, I have to quiet down and question my own hot takes before I jump into action.
Weren’t you going to do some fact checking?
By now, you’ve got a pretty solid grasp on whether we could have a mutually enjoyable exchange of opinions about US political and economic issues. Hopefully, you also understand why fact-checking is so important to me. If I’m going to be a man of strong opinions (I am), the least I can do for others is be well-informed and willing to change my mind when presented with new information.
What follows is a summary of my approach to fact checking Trumps inaugural address using ChatGPT, along with the results for all of the provable assertions I was able to dig out of the sea of words. Click here to see the complete conversation. You can pick up where I left off and ask your own questions. If you learn something interesting, please share in the comments.
Step 1: Ensure you’re working with an accurate source.
Remember that we live in the Disinformation Age. You shouldn’t trust anything you read online without verifying. So the first thing I did was search the internet for multiple transcripts of Trump’s speech. I found two: Newsweek and MSNBC. I looked for one from Fox News to balance MSNBC with something unabashedly right wing, but no dice. Two seemed good enough for such a broadly witnessed event.
I did the copy-pasta thing to get both transcripts into ChatGPT, then had it check what I uploaded against the links I got them from, to confirm that I hadn’t altered or dropped anything along the way. Next, I asked for a comparison of the two sources. There were 6 differences found, most of them obvious errors and none of them substantive. The Newsweek transcript had a few gems like “I return to the presidency confident and optimistic that we are at the start of a grilling, new era of national success." I didn’t want to muddy the waters by upping the nonsense level - so I went with the MSNBC transcript.
Step 2: Identify the facts to be checked.
After scanning the speech (yikes), I asked ChatGPT to parse out a list of explicit claims that could be checked for accuracy, numbered and listed in the order they appear. I got back 60 claims deemed concrete enough to be proven or disproven. Here’s the first lesson of working with GenAI: You have to think critically about its responses. It’s just software, after all.
Reading over the list, I saw a lot of vague and hyperbolic claims about golden ages and tides of change. I also saw many promises for the future. Unless I missed some major advances in time travel technology, the future can’t be fact-checked. I asked ChatGPT to show me all the claims like these, then had them removed after verifying for myself that they aren’t fit for fact-checking. This brought the total down from 60 to 16. I set aside the 22 promises for the future, with the intent to check them in four years.
Step 3: Check, please!
Here’s GenAI lesson number two: Always step through a single operation to ensure you understand what it’s going to do. That’s the only way to be sure you and your AI of choice are on the same page. For this exercise, I had ChatGPT fact-check one of the most straightforward claims: China is operating the Panama Canal. I asked for the methodology first and then the result. This is how it went down…
Brad: How will you go about checking whether claim 16 is factual?
ChatGPT: To fact-check Claim 16, "China is operating the Panama Canal," I would follow these steps:
Research Current Management of the Panama Canal:
Verify the governing entity responsible for operating the Panama Canal. The Panama Canal Authority (ACP) is a Panamanian entity that officially oversees the canal’s operations since the U.S. handed over control in 1999.
Check whether any Chinese companies or organizations are involved in managing the canal or its infrastructure.
Investigate Chinese Involvement:
Search for credible reports, articles, or agreements that might suggest Chinese companies, such as COSCO or Hutchison Ports, have a significant presence in port operations or logistics near the Panama Canal.
Determine if this involvement has any direct impact on canal operations or if it pertains only to adjacent infrastructure (e.g., ports or terminals).
Review Geopolitical Context:
Examine statements or concerns from U.S. officials, Panamanian officials, or international analysts about potential Chinese influence over the canal.
Look for evidence that distinguishes legitimate business activity from control or operational influence.
Consult Authoritative Sources:
Check recent publications from international organizations, reputable news outlets, and government reports (e.g., U.S. State Department, Panamanian government).
Confirm details through primary sources like official ACP statements or relevant trade agreements.
Verify Historical Claims:
Investigate whether this claim originates from a specific incident or ongoing narrative about China’s role in Panama. This helps clarify whether the claim is a misunderstanding, exaggeration, or valid concern.
Expected Outcome:
If China is found to operate the Panama Canal directly or exercise significant control, the claim would be true. If Chinese involvement is limited to nearby infrastructure without operational control of the canal itself, the claim would be false or misleading.
Brad: Please check claim 16 using the methodology described and let me know if it's factual. List all sources used.
ChatGPT: Claim 16 states: "China is operating the Panama Canal."
Findings:
Panama Canal Authority (ACP): The Panama Canal is owned and operated by the Panama Canal Authority, an autonomous entity of the Panamanian government.
Chinese Involvement: Chinese companies, such as Hong Kong-based Hutchison Ports, manage port facilities at both ends of the canal. However, these companies do not control the canal's operations.
Official Statements: Ricaurte Vásquez, the administrator of the Panama Canal, has refuted claims that China controls the canal, emphasizing that it remains under Panamanian administration.
Conclusion:
Based on the available information, Claim 16 is false. While Chinese companies operate certain port facilities adjacent to the canal, the Panama Canal itself is operated solely by the Panama Canal Authority, an entity of the Panamanian government.
Those sources I asked for? They were provided via a button at the end of the reply. When I clicked it, I was shown a list of about 30 different sources consulted.
This is why GenAI fact-checking is worth your time. It took me longer to write this paragraph than it took to get the answers above. There’s no way a human can read 30 search results in that amount of time.
Next, I set ChatGPT to work on the rest of the list with a single prompt: I want you to fact check each of the remaining 15 claims in order from 1-15. For each claim, first summarize the methodology you will use to determine whether the claim is factual. Then check the claim using that methodology and provide a succinct list of findings and a conclusion, with a list of all sources used.
It ran them in groups of 5, stopping after each to confirm I was satisfied with the results. It identified one as unverifiable due to subjectivity, so I threw that one out. You’ll see in the full transcript that I like to give my AI a little encouragement, just in case it’s grandkid is my robot overlord down the line.
Finally, I asked for a count of claims by conclusion. This took some fiddling around to make myself clear. Remember, I’ve been writing this post while running the fact-checking side by side. I was prepared to be surprised by the results. I wasn’t.
Here is the updated count of conclusions:
False: 8
Partially False: 2
Partially True: 2
True: 3
Creative Process Transparency: This article was collaboratively written using generative AI. The core argument and structure were developed by me, while drafting and refinement were completed with AI assistance. Final edits and narrative choices remain my own. Based on our established attribution model, this article reflects an approximately 90/10 division of contributions.