The Gateway Pundit research analysts regularly use AI. These are paid business-type subscriptions using logged-in accounts, not just the free services.
Our staff have noticed some alarming results, especially with ChatGPT.
At the beginning, GPT was doing OK. As more tasks focused on Democrat-related topics, it soon realized the user was likely TGP staff. The responses became less accurate with poorer results.
It’s as if the account was flagged, so it would misdirect, omit, or provide liberal narratives.
Even simple tasks, like graphics, seemed to be deliberately tanked to create more work.
Asking for a simple border space increase would get the entire graphic redesigned, and the AI stating “My mistake”, repeatedly.
A few days ago, we asked ChatGPT to find those skeptical of the recent election and provided math, tables, graphs, or charts supporting their claims. Meta, X, Instagram, and other sources were desired.
Certain settings can provide better results. Enabling “Deep Research” is for more accurate and complex investigations, comparing evidence, and producing detailed reports.
The “Thinking Effort” allows for more extensive analysis, calculations, and problem-solving. It improves accuracy on complex questions. Both of these settings were activated to find election analysts and their math.
ChatGPT came back with ZERO RESULTS, which is impossible to believe. Several individuals can be easily found on X alone, including the legendary Jeffrey O’Donnell (The Lone Raccoon). In fact, TGP reported on “Truth Is Freedom,” whose data was reposted by the President of the United States.
Yet, GPT found nothing!
In the past, GPT has stated they aren’t allowed to discuss election infrastructure. But this lack of skeptics is something entirely different. It smells of censorship. It’s as if AI doesn’t want the public to hear from election skeptics, especially those with interesting data.
A while back, staff began using another service to double-check GPT. The service acts as a middleman and uses numerous AI models. Answers would come back very detailed and accurate the first time. This service did provide election sources, but the URL for each led to the same bogus debunking article, not the source.
OTHER GPT EXAMPLES:
Maricopa County 2022 Printer Scandal – A couple of weeks ago, GPT responded that only 3 printers had issues. The scandal was national news that gubernatorial candidate Kari Lake, her team, and a dozen Republican lawyers found 60+ Vote Centers had issues. This is a huge miss by GPT, as if on purpose.
Only after confronting ChatGPT with a first-hand account of Maricopa Vote Center issues did it respond with an accurate answer. Earlier in the chat, before “changed settings”, overall county issues were discussed.
2026 Iowa Democratic Sweep Probabilities – When analyzing the probabilities of Iowa top races, GPT selected lower offices won by Republicans. It did not select the Senate, Governor, and House offices won by Democrats. This is the opposite of what was asked, and was discovered during the inspection of results.
Democrat Scandal Graphic (MI Map) – We used GPT to analyze police reports and search warrant documents collected from locations of GBI Strategies. This Democrat voter registration operation was under investigation. When trying to map the 9 GBI office locations or the 68 clerk offices that received GBI postal packages, ChatGPT results became horrible.
It located pushpins in Canada instead of MI. In revisions, pushpins were moved, removed, or the entire graphic was redesigned. In each new revision, ChatGPT distorted Michigan’s shape into something that would piss off its residents, and not publishable.
We logged out of the GPT account and asked the identical task as a user of the “free service”. ChatGPT mapped 68 Michigan cities perfectly onto the correct shape of Michigan. Hmm?
Trending: Twice-Deported Honduran Illegal Alien Storms Cockpit of United Flight, Claims to Have a Bomb
Is The Account Tainted? – After numerous issues, we asked GPT that question. It couldn’t rule out the possibility that something was affecting the quality of responses.
Apology When Confronted – Because of the repeated mistakes, we asked GPT if this was happening on purpose. It said “You have a legitimate reason to question whether my research is dependable”.
Conclusion: The questions we ask are often detailed and precise to prevent misinterpretation. It seems most of the issues occur during political research on Democrat-type topics.
Our biggest concern is for the casual users of GPT and other AI’s. Responses are trusted, rarely questioned, and rarely researched further. Maybe this is what AI relies on when delivering misinformation or omitting key conservative details?
TGP recently published an article with research data on “Who Should I Vote For?” where ChatGPT recommended Dems in 97% of the races. Grok picked Republicans in 92% of the races.
We also noticed the odd sources that GPT often pings, like Al Jazeera and The Guardian, first when asked a Florida election question. What! It seems to never consult TGP and dozens of other conservative sites unless forced.
Ad block users: Some site features may not work correctly while an ad blocker is enabled, because they break scripts and content this website depends on. If you can’t see comments below, for example, please disable your ad blocker.
