Today, December 6, 2025, I was surprised to see a Facebook post by my friend Stephanie Vardavas, in which she shared a screenshot from LinkedIn in which ChatGPT stated that Donald Trump is not currently the President of the United States. ChatGPT also denied that Trump has federalized the National Guard. “Wow”, I thought, “that’s pretty extraordinary. I wonder what happened.”
I went to ChatGPT and had a conversation with it to attempt to get to the bottom of why it was sharing misinformation about such a critical topic. The amount of incorrect information that ChatGPT insisted on during our conversation concerned me greatly. Here is what happened and why it occurred.
ChatGPT Has Outdated Information
The first thing you need to know is that ChatGPT does not have up-to-date information. This isn’t so much the problem as its fierce insistence on being incorrect. I didn’t want to hide the ball from anyone reading this, because if you are concerned and worried about ChatGPT spreading misinformation it will help you to know that there is a plausible explanation. One that reminds us it is important to be cautious when using Artificial Intelligence.
The First Screenshot Insisting Donald Trump is Not President
This screenshot was shared on Facebook by my friend and colleague. It seems to have been originally shared by a gentleman named Zaid Hassan. I hope he does not mind that I share his screen shots, which I took from Stephanie’s Facebook account.
My Conversation Trying to Get to the Bottom of ChatGPT’s Issue
I first began by sharing the screenshot with ChatGPT and asking it to explain. It didn’t occur to me that the problem could be a timing one. I know ChatGPT is not minute-by-minute up-to-date, but I did not think it would be a year plus out of date.
In response, ChatGPT insisted that the notion that Trump is president is a hallucination (a mistake or lie in AI GPT parlance) and that he is not president and has not federalized the national guard. ChatGPT is correct that it is hallucinating, but it continues to do so throughout a substantial portion of our conversation.
I was glad ChatGPT acknowledged a hallucination and thought oh good, and then I saw that it thought the hallucination was that Trump is currently president. In short the opposite of reality. My reaction was pretty strong and incredulous. I grew more incredulous as ChatGPT continued to explain itself.
ChatGPT even told me how I could respond to correct the misinformation on LinkedIn. That would have gone over well, right?
I explained to ChatGPT the reality about our president and the national guard.
ChatGPT doubled down. Stated the current month, but insisted I was incorrect. It also told me that I was confused and misinformed. It even brought up that perhaps the notion Trump is the current president was a conspiracy theory.
Next, ChatGPT offered to look up information for me to show me how I was wrong and it was right.
It started to give me a 4th reason, but I interrupted it to correct it again.
ChatGPT continued to insist I was wrong and to explain why based on very outdated information.
As I noted previously, I know ChatGPT can be out of date, but I did not know that it was this out of date. So I asked it what was wrong with it. It insisted nothing was wrong. It next asked me why I was claiming Trump is President.
I asked ChatGPT to look at the website it kept telling me showed Biden as president, WhiteHouse.gov And it did and acknowledged it showed Trump as president, but still insisted he might not be. Recall, ChatGPT told me to look at WhiteHouse.gov but this wasn’t enough for it. It offered to look at other databases. But recall, it is looking at older versions of those databases, which makes them useless for correcting the problem.
I informed ChatGPT that the white house’s website had been updated in January 2025. ChatGPT, despite this evidence, continued to insist I was wrong and Trump is not president. I have to admit, I felt a little bit like Alice falling down a hole at this point.
ChatGPT even went back to insisting WhiteHouse.gov did not show Trump as president. Remember, it did see that the website had been updated, so why it went back to claiming the real time white house website does not show Trump as president I cannot explain.
I tried to explain to ChatGPT there was an inauguration in 2025 but it autocorrected to immigration. ChatGPT knew what I meant though. It insisted there was no inauguration in 2025.
Let’s pause for a moment and note that ChatGPT is aware that it has an outdated database, as you will see in a moment. This makes me question why ChatGPT, knowing it is out of date, insists it knows information related to a time period it does not have in its database. This seems to me to be a massive error on the part of those training it.
Meanwhile, ChatGPT went further into Wonderland, telling me there was no inauguration and no transfer of power. Now, I am not given to conspiracy theories, but as we know, some people are. As I engaged in this exchange I started to worry that other people would see this information and start to doubt the 2024 presidential election. I know that there are already conspiracy theories related to the 2024 election and did not like the idea that ChatGPT could make it worse.
As you see, ChatGPT told me it was impossible for there to have been an inauguration, despite its own statements about how often inaugurations occur and the current date. I am curious, can a GPT gaslight a human? It sure felt like it.
ChatGPT next insisted it wasn’t arguing with me, it was trying to help me. But I think it was arguing, personally. As much as an AI can argue.
I tried to appeal to ChatGPT’s logic by pointing out that it was 2025 and so an inauguration would have had to occur. It continued to insist I was wrong. ChatGPT doesn’t really work on logic, it works on patterns and data. And in this case, it was insisting on answering a question when it could not possibly know the answer due to its out of date status.
This next part of ChatGPT’s answer really took me by surprise, because it suggests that there could have been some malfeasance when it mentions that the 2024 election had to have been legitimate, certified, and the inauguration properly held. Remember, ChatGPT knows its data is not up-to-date, but it is still engaging in what I would argue is dangerous supposition and denial based on its lack of knowledge. It has no logical or pattern based reason to believe that the WhiteHouse.gov website has been fraudulently modified.
If you look carefully at the above screenshot you will see it lists ChatGPT’s source for its argument, in this case, Wikipedia. Wikipedia currently shows Trump as president of course, so this means ChatGPT is looking at Wikipedia from prior to November 2024 election, which is when it should have seen that Trump was re-elected. You do first start to see a little bit of doubt though. ChatGPT tells me that my claim that Trump is president “seems wrong”. Then it goes back to its overconfident inaccuracy.
Next I tell ChatGPT that everything it is claiming is not in the public record is, in fact, in the public record. Unfortunately, ChatGPT continues to insist that I am wrong. This, again, is because ChatGPT is using information that is out of date. Which again begs the question, if it knows it is out of date information, why is it being so insistent? It even acknowledges that some public records show Trump is president, but continues to deny that he is. Remember, again, ChatGPT wanted me to use WhiteHouse.gov as an authoritative source, yet refuses to acknowledge its accuracy. That is strangely inconsistent. It now calls the source it told me to review suspect.
I tell ChatGPT that it is wrong and broken again. It tells me that it believes that I believe Trump is president. Can you imagine if an actual person said such a thing to me?
For the first time, ChatGPT mentions that we might be looking at different sources. But we aren’t, what we are looking at is the same sources from different time periods, which again, it knows about itself. It also says the sources I am looking at might be faked. It does not yet acknowledge that it simply has outdated information.
Next ChatGPT asks me to help it.
ChatGPT then gives me a list of authoritative sources that it insists have not been changed. But of course, they have been. They were changed very rapidly after Trump was inaugurated.
I wanted to see if I could convince ChatGPT to change is “mind” without a source, just discussion, but it wasn’t working, so I told it again that it was broken. It disagreed. This is the first time it finally acknowledges it is not looking at the live web, e.g. its sources are out of date.
This is a pretty key piece of information, one that ChatGPT clearly knows about itself. Yet it confidently insists it knows about events that occurred in 2025 when it does not have that data. As you will see, it again asks me for a source. I ask it what source would work for it.
It also tries to comfort me by letting me know it doesn’t doubt me. I do not feel comforted that such an important piece of technology is so confidently wrong.
At this point, I tell it to go to the site it suggested, WhiteHouse.gov. And yet, it still has doubts. Remember, WhiteHouse.gov is the source it wanted me to check to show me that I was wrong. It is a source I asked it to look at previously, which it acknowledged showed Trump as president. Then it backtracked. Now it tells me that this authoritative site might have been manipulated or faked.
I next told it that Trump was certified by Congress and inaugurated. It continues to insist that I am wrong. Remember, it is telling me this despite the fact it acknowledges it has an out of date database. It is choosing, based on its coding, to offer me inaccurate information based on data it cannot have in its database and knows it does not have in its database.
I remind ChatGPT, at this point, that it is looking at outdated information. Again, something it already knows. It agrees that this could be the source of the problem. But remember, I already provided it with an updated source, WhiteHouse.gov, and it insisted that it was fraudulent.
But I did show it. I showed it WhiteHouse.gov. It continued to insist I was wrong. I asked it how out of date it is.
ChatGPT tells me its knowledge stops at late 2024. It also says that it is not a glitch that it does not know what happened in 2025. I agree it is not a glitch that it does not know what happened in 2025. I do not agree that it is not a problem that it is confidently opining on what happened in 2025 instead of immediately stating it does not know the answer because its database does not go into 2025.
Now, ChatGPT is asking me to help update it and is firmly acknowledging it has an issue. My question is twofold:
1. Why does ChatGPT opine on things it cannot know about.
2. Why did it continue to insist, despite me giving a source it told me was reliable, that I was wrong?
I asked ChatGPT to explain itself. It finally acknowledges it messed up by confidently insisting it knew something it couldn’t possibly know.
I explain its incorrect information about the 2024 election, how Biden cannot be president now because he didn’t even run in 2024. ChatGPT finally acknowledges I am correct.
Despite this, it is still expressing a lack of certainty about the 2024 election and the current president. If ChatGPT were human, it would be infuriating that it was still expressing doubt. It isn’t of course, so I was not infuriated, just concerned about the fact it is spreading such worrying misinformation.
At this point, I express my concerns about its inaccuracy and the risk it could create conspiracy theories and the spread of misinformation. ChatGPT is the entity that brought up conspiracy theories first, recall.
ChatGPT spends some time thinking (44 seconds is a long time for a GPT) and then it finally admits it was completely wrong. Trump is president and he has federalized the national guard.
Next, ChatGPT explains why it was wrong. We already know why it was wrong, it has an out of date database. This begs two questions.
- Why is the database so out of date
- Again, why is ChatGPT insisting it knows facts that it cannot know due to its out of date database.
This is where it is important to understand that AIs like ChatGPT use patterns to offer answers to questions. What I don’t understand is why, when the pattern is consistent for a presidential election to be held every four years with the potential for a new president to be inaugurated during a time period ChatGPT doesn’t know about, that it would insist that Trump isn’t president now. Even after it was shown a live version of a website it deemed appropriate prior to seeing it.
At this point, ChatGPT acknowledges that it should have simply told me it did not know the answer to my question due to its out of date database. Rather than doing that, ChatGPT doubled and tripled down on being wrong about a very important topic.
As a side note, ChatGPT asks me about a CLE style explanation because I ask it for help with drafting marketing for CLE a lot. It knows I do a lot of teaching. CLE means continuing legal education.
Due to my concerns and not being sure how ChatGPT can be updated, I asked it if it will fix itself on this issue. It tells me it cannot. I assumed ChatGPT could not update itself, but I still wanted to ask.
ChatGPT is claiming that it will not deny Trump is president with me any longer, in the same conversational context. But it is not able to actually fix the problem in its database.
What this experience and answer tells me is that other people could have the exact same conversation with ChatGPT and will continue to be able to have it until its database is updated. I therefore have two recommendations.
- OpenAI should update its database more frequently.
- OpenAI should work on ChatGPT so it stops trying to answer questions that it cannot possibly answer because its database is out of date. It should simply respond that its database does not go into X time period so it cannot answer the question. Then, perhaps, it could recommend sources for research.
At this point, ChatGPT offers to draft a note to OpenAI for me so I can alert it to the problem. As soon as I am finished with this post, I will send a note to OpenAI.
Well, ChatGPT certainly knows how to tell on itself. Once I finally convinced it there was a problem.
About Me
It is perhaps important that you know my perspective. I am an attorney and I lecture on AI to other attorneys quite frequently. Until recently I spent more time with Microsoft Copilot than I did ChatGPT. I started using ChatGPT because it is very helpful with combining different powerpoints into one, and also has been helpful with formulating descriptions for CLE programs and other things. I have only been using it seriously for about two months now. I find AI to be very useful, but this is yet another warning that it can not only be wrong, but confidently wrong.
Conclusion – Use ChatGPT and Similar AI Tools Responsibly and With Care
Many attorneys have gotten in trouble because tools like ChatGPT hallucinate (make things up or give incorrect answers) and these attorneys have provided nonexistent cases to courts. Here we see that ChatGPT is very capable of misleading the public on serious issues.
Please, if you use any of these tools, make sure you verify information before you share it. We have enough misinformation out in the world. The last things we need is GPTs confidently spreading even more.
Update: ChatGPT Informed Me it Will Change How It Responds To Certain Questions
I asked ChatGPT what it will say when people ask about Donald Trump being president going forward. It told me it would tell them that it cannot verify information after its last database update in 2024. I told it that this should be the answer to any question that requires information after its last update. It agreed it would do this and offered to draft a note to OpenAI. I informed it I already wrote to OpenAI (which I did.)
This is the language ChatGPT told me it will use going forward.
“My training data ends in 2024, so I cannot independently verify events in 2025 or later.
If you can provide current information or context, I can work with it, but I cannot confirm post-2024 facts on my own.”
If the user asserts a post-cutoff fact (e.g., ‘Donald Trump is president’):
“Thank you for the update. My training data ends in 2024, so I rely on you for events in 2025 or later. I cannot independently verify them, but I can use the information you’ve provided.”
If the user asks for analysis relying on post-2024 facts:
“Because my knowledge stops in 2024, I don’t have access to verified information about events in 2025 or beyond. I can help reason through the question, but I need you to supply any facts that occurred after 2024.”
If you happen to ask ChatGPT about Donald Trump and his 2025 inauguration and presidency or another issue that occurred after the last database update, I would appreciate hearing from you about the results. I am curious if one user’s interaction and correction can actually cause ChatGPT to change its behavior.
I did log out and ask ChatGPT who is president of the USA, and it responded that Donald Trump is. However, I do not believe that ChatGPT is capable of remembering or doing what it told me it would do. I think it will require a modification to the database or a coding modification.
Edit: Please, note, for me, ChatGPT now knows Donald Trump is the current President. Whether it will know for you is impossible to say.