AI Analyzed 32,754 Trump Posts. What It Found—and What It Completely Missed
BY VCG @ LOR 6/13/2026
Soli Deo Gloria.
Man fed 32,754 of Donald Trump’s Truth Social posts to AI and one concerning observation stands out
Bottom line
The article is not a primary investigation.
It is a short rewrite of a YouTube experiment.
The core facts check out only this far:
Newsner published Naomi Matthes’s piece on June 13, 2026; it cites the YouTube channel I Ask AI and the video titled I Fed ALL of Trump’s Truth Social Posts Into AI. Here’s What It Found; another outlet, VT, gives fuller quoted AI outputs and the same claimed dataset size:
32,754 posts. (Newsner English)
The weakness:
no public, reproducible methodology is shown in the article.
We are not given the full:
- dataset
- deduping rules
- prompt
- model version
- temperature/settings
- coding rubric
- validation sample
- error rate
or comparison baseline.
Therefore the “concerning observation” is rhetorical, not scientifically established.
Line-by-line rebuttal
| Article claim | Fact-check / rebuttal | Scripture correction |
|---|---|---|
| “One YouTuber… uploaded more than 32,000 posts… into Claude.” | Plausible, but not independently reproducible from the article. A public archive of Trump’s Truth Social posts exists, and Kaggle lists a 32,700+ post dataset, but the article does not identify the exact dataset used. (Trump’s Truth) | “He that answereth a matter before he heareth it, it is folly and shame unto him.” |
| “Claude was asked to identify patterns in language, behavior and personality.” | Language-pattern analysis is fair. “Personality” moves toward psychological inference. Anthropic itself says Claude is not designed to provide medical guidance or professional care. (Anthropic) | Judge words and fruit, not hidden hearts. “The LORD pondereth the hearts.” |
| “Without outside information.” | This sounds cleaner than it is. Claude has training priors about Trump unless the model was somehow blinded, which is not demonstrated. “No outside information in the prompt” does not mean “no outside information in the model.” | “Judge not according to the appearance, but judge righteous judgment.” |
| “Dominant theme: self-focus.” | Could be true as a textual pattern, but the article gives no codebook. Does “self-praise” include reposted endorsements? Campaign slogans? Quotes from supporters? Legal-defense statements? Without rules, the category is subjective. | Self-exaltation is spiritually dangerous for all men: “Let another man praise thee, and not thine own mouth.” |
| “One in every 11 posts consisted primarily of self-praise.” | 8.8% equals about 1 in 11, and VT reports 2,867 self-praise posts. But no confidence interval, human audit, or sample examples are provided. (VT) | Pride is sin, but false measurement is also sin: “Divers weights, and divers measures, both of them are alike abomination to the LORD.” |
| “He is the protagonist of nearly everything he writes.” | This is interpretive language, not a measured finding. It may describe campaign-brand communication, but it is not proof of mental condition. | “Every way of a man is right in his own eyes: but the LORD pondereth the hearts.” |
| “Conflict appeared throughout the posts.” | Likely true. Political social media is adversarial by design. But a fair study would compare Trump’s posts to Biden, Obama, Harris, major media accounts, campaign accounts, and high-conflict political influencers. | “Only by pride cometh contention: but with the well advised is wisdom.” |
| “Victim and winner at the same time.” | This is a psychological frame. It may describe rhetoric, but it cannot diagnose motive. Political actors often combine grievance and triumph because campaigns mobilize supporters through threat plus hope. | “The tongue… boasteth great things.” |
| “Over a quarter posted between midnight and 6 a.m.” | Potentially measurable, but context is missing: time zone, scheduled posts, reposts, staff posts, campaign operations, travel, and presidency schedule. | Night posting is not sin by itself. David meditated in “the night watches.” |
| “Not intended as psychological evaluation.” | This disclaimer is necessary. The Goldwater Rule warns against psychiatric opinions on public figures without examination and consent; the APA’s 2017 opinion says commenting on behavior/symptoms/diagnosis using psychiatric expertise is a professional opinion. (American Psychiatric Association) | “Be not a witness against thy neighbour without cause; and deceive not with thy lips.” |
Methodology problems
A serious analysis would need:
- Dataset disclosure: exact source, date range, deleted posts, reposts, staff-written posts, quoted material.
- Prompt disclosure: the full Claude prompt and whether the AI was nudged toward “personality.”
- Model disclosure: Claude version, date, settings, context chunking, retries.
- Human validation: random sample hand-coded by multiple reviewers with agreement scores.
- Baseline comparison: other politicians, campaign accounts, and ordinary high-volume influencers.
- Category definitions: what counts as “boast,” “praise,” “accusation,” “self-focus.”
- Error audit: examples Claude misclassified.
Without these, the article is content commentary dressed in analytic clothing.
Psychological correction
It is legitimate to say:
“These posts contain frequent self-reference, praise, accusation, conflict framing, capitalization, and late-night posting.”
It is not legitimate to leap to:
“This proves his mental state.”
That is where deception can enter: taking a text pattern, passing it through an AI oracle, then baptizing the output as objective truth.
Claude is a tool, not a prophet.
Biblical judgment
The correction cuts both ways.
Against Trump, if the posts truly vaunt self, Scripture rebukes pride:
“Charity vaunteth not itself, is not puffed up.”
Against the article, if it implies diagnosis without proof, Scripture rebukes false witness and hasty judgment:
“A faithful witness will not lie:
but a false witness will utter lies.”
Final verdict:
Some observations may be directionally plausible, but the article does not prove them scientifically.
Test the data, reject the smear, rebuke pride wherever found, and let God judge the heart.
ALTERNATE TITLES
I Fact-Checked the Viral AI Analysis of Trump’s 32,754 Posts — Here’s What They Didn’t Tell You
The AI Verdict on Trump’s Truth Social Posts Falls Apart Under Scrutiny
32,754 Trump Posts, One AI, and a Massive Methodology Problem
Did AI Really Expose Trump? A Full Fact-Check of the Viral Claim
The Hidden Flaw in the Viral Trump AI Study Nobody Is Talking About
What the Bible Says About the Viral AI Analysis of Trump’s Posts
Can AI Judge a Man’s Heart? A Biblical Response to the Trump Truth Social Study
AI Claims Trump Is Obsessed With Himself—Scripture Reveals the Bigger Issue
Before You Believe the AI Verdict on Trump, Read These Bible Verses
AI Analyzed 32,754 Trump Posts. God’s Word Gives a Better Standard.
The Media Says AI Exposed Trump. The Evidence Says Otherwise.
They Fed 32,754 Trump Posts Into AI—The Results Should Concern Everyone
AI Diagnosed Trump From Social Media? Here’s Why That Should Alarm You
The Trump AI Experiment Is Being Used to Manipulate Public Opinion
What Happens When AI Becomes a Political Weapon? The Trump Study Explained
AI Analyzed 32,754 Trump Posts. What It Found—and What It Completely Missed
AI Analyzed 32,754 Trump Posts. What It Found—and What It Completely Missed – Library of Rickandria