← All fact checksAfternoon run · November 16, 2026 · #8 by reach
PoliticsConflictTechnologyAI-generated media

Did Trump post an AI-generated video of US troops kneeling in prayer?

Does the video Trump posted on Truth Social on 15 November show real US troops kneeling in prayer beside C-17 transport planes?

True, but…

Trump did share the video — but the footage is AI-generated, not real troops.

Yes. Donald Trump shared an AI-generated video of US troops kneeling in prayer on Truth Social at 8:58 p.m. ET on 15 November 2026. The same clip first appeared on TikTok on 5 April 2026 with an AI label, and Lead Stories, Newtral and VerificaRTVE had already rated it fake before Trump reposted it.

AI-generated
The post as it circulated, labelled by our media check.Video · 0:12 · 1320×1752

Claims

✓ 3 true

Donald Trump posted a fake AI-generated video of US troops kneeling in prayer on Truth Social.1238

✓ TrueCentral claimHigh confidence

The event is confirmed from several independent directions. Outlets including Raw Story, Yahoo News carrying a Daily Beast report, and IBTimes UK all report that Trump shared the clip on Truth Social on 15 November 2026 at 8:58 p.m. ET, and that it showed service members kneeling in prayer on a tarmac beside C-17 aircraft. Thai and Turkish coverage on the same and next day reports the same post from their own newsrooms, so this is not one article being copied — it is multiple outlets independently recording the same public post. The screen recording in Krassenstein's own post matches that description: the soldiers, the tarmac, the transport planes. Just as important, the 'fake AI-generated' half is settled separately: Lead Stories fact-checked this exact clip months earlier and rated it false, and Spanish fact-checkers Newtral and VerificaRTVE reached the same conclusion, with VerificaRTVE finding a Google SynthID watermark tying it to Google's Gemini model. Both halves of the sentence — that Trump posted it, and that it is AI — stand on independent evidence.

The video showing troops kneeling in prayer on a runway is not real footage but was created with generative AI.456

✓ TrueHigh confidence

This is the strongest part of the case, because it was decided by specialists before the post was even written. Lead Stories fact-checked this exact clip in April 2026 and rated it false and AI-generated. Newtral concluded the images were produced with Google's AI tools. VerificaRTVE went further: it found a Google SynthID watermark, which is an invisible marker the model itself embeds at the moment of generation — not something a later editor can easily add or forge. On top of that, our own visual review of the keyframes shows warped and smeared faces, bodies that merge and morph, hands lost in torsos, and figures that dissolve into the tarmac. These are the artefacts generative video leaves behind. Our automated detector did not add evidence here: it scored the clip as 0.996 'not AI generated' for the overall image classification, a result that conflicts with everything else, while returning a separate deepfake score of 0.60. We did not treat it as decisive. The watermark and the three independent fact-checks do the work.

Still open: Our automated detector produced a conflicting result on this clip; we relied on the SynthID watermark, the fact-checks and visual review instead.

The clip was originally uploaded to TikTok in April with an AI label and had already been debunked by fact-checkers before Trump shared it.345

✓ TrueHigh confidence

Lead Stories traced the clip to a TikTok account, @spectacularsights, which posted it on 5 April 2026 with TikTok's 'Creator labeled as AI-generated' disclaimer. That account routinely posts AI-generated religious scenes. VerificaRTVE independently records the same first-appearance date and the same AI label, so this is two separate fact-checking teams arriving at the same origin — not one blog repeating itself. Outlets reporting on Trump's November repost explicitly reference the earlier April fact-check, which means the clip's synthetic nature was a matter of public record roughly seven months before the president shared it. The 'already debunked' element is therefore supported by the fact-checkers themselves and by the news coverage that cites them.

Trump's posting of the AI video is disrespectful.

– OpinionHigh confidence

This is a value judgement about the president's conduct, not a checkable fact. Words like 'disrespectful' and 'beyond words' express how the poster feels about an act; reasonable people who agree on every fact here can disagree about whether the act deserves that label. We do not rate opinions true or false, and we did not treat this one as evidence of anything else. The important thing for a reader is that the opinion sits on top of an accurate factual description: the post does not get the facts wrong to reach its conclusion. The surrounding context the criticism points to — that the repost came hours after a CBS 60 Minutes report on Iranian strikes on US military and diplomatic sites — is reported by the outlets covering the post, and readers can weigh the timing for themselves.

What was posted

X

Brian Krassenstein posted on X a screen recording of a Truth Social post from Donald Trump, showing rows of camouflage-clad service members kneeling on an airfield in front of C-17 transport aircraft. The clip's soundtrack is a religious song. Krassenstein's caption calls the video "FAKE AI-generated" and states Trump posted it, framing the act as disrespectful — an opinion layered on top of an accurate description of what happened.

“Why the HELL is Trump now posting FAKE AI-generated videos of our troops kneeling in prayer. This is so DISRESPECTFUL. I’m BEYOND WORDS HERE.”

Posted on X by Brian Krassenstein, an account that regularly posts political commentary. The post shows a recording of Trump's Truth Social post rather than the file itself.

Who posted it

X

Account check

Looks like a genuine account38%
Likelihood of automation30%

Estimates from public signals — account age, posting pattern, audience — not a judgement of the person behind the account, and separate from whether this post is true.

What stood out

  • Handle resolves to Brian Krassenstein, a real, publicly identifiable individual (twin brother of Ed Krassenstein), a former #Resistance personality with ~1M+ followers — not a sockpuppet or anonymous persona.
  • Accounts were permanently suspended by Twitter in May 2019 for 'operating multiple fake accounts and purchasing account interactions'; reinstated December 2022 after Musk takeover. Brothers denied wrongdoing.
  • High-volume, engagement-farming style: rapid-fire sensational posts, ALL-CAPS 'BREAKING' framing (e.g., Nov 2026 FAA nationwide ground-stop post), heavy emotional commentary — typical of monetised influencer behaviour rather than sourced reporting.
  • Documented false/exaggerated claims: Oct 2026 'Trump waved to nobody' Marine One video fact-checked by Lead Stories as missing context; the Ed-linked Feb 2026 Bad Bunny halftime 'Liam Ramos' claim was debunked (child was actor Lincoln Fox) and left up rather than deleted.
  • Family business history (TalkGold/MoneyMakerGroup forums, DOJ civil asset forfeiture ~$450k, no charges) and accusations of monetising outrage weaken credibility as a source.
  • Consistency issue: announced 'officially leaving X' in March 2026, then continued posting prolifically through November 2026 — attention-driven rather than principled positioning.
  • Some content is accurate (e.g., criticism of Trump's mislabelled eagle photo, flagged AI troop-prayer video) but usually without citations; posts function as commentary, not verifiable reporting.
  • Profile metadata supplied is empty (no bio, no follower count, isplatformverified=false), so signals were reconstructed from external coverage and post aggregators; prior snapshots show the account was blue-verified.
Signals (8)+
Account identity
Brian Krassenstein (real named individual)
Follower scale
~1M+ (down from ~1.5M peak in 2025-26)
Posting cadence
Very high, multiple posts daily, breaking-news style
Platform enforcement history
Permanent suspension 2019 for fake accounts / purchased interactions; reinstated 2022
Fact check history
Lead Stories (Oct 2026, Marine One wave claim), Bad Bunny/Liam Ramos false claim (Feb 2026), PolitiFact check (Oct 2023)
Notable news coverage
NY Post, Daily Beast, Newsweek, Fox News, PJ Media, National Post
Commercial incentive
Engagement-monetised political influencer; denies paid by PACs/DNC
Source citation practice
Rare; commentary-first framing

Account checked November 16, 2026, 15:00 UTC

The media

Video
AI-generated· The claim depends on it

✓ What is genuine

The Truth Social post itself is real: Trump did share this clip on 15 November 2026, and the screen recording in Krassenstein's post accurately shows that share. The soldiers, the airfield and the prayer scene are not real.

The tells

  • ●●● Strong signal A Google SynthID watermark is embedded in the footage, indicating generation with Google's Gemini model.VerificaRTVE
  • ●●● Strong signal Soldiers' bodies warp and morph, faces repeat and smear, hands merge into torsos, and figures dissolve into the tarmac.our visual review
  • ●●● Strong signal The first upload on TikTok carried the platform's 'Creator labeled as AI-generated' disclaimer.Lead Stories
  • ●○○ Weak signal Our automated detector scored the clip as overwhelmingly 'not AI generated' on its overall classification while returning a separate deepfake score of 0.60 — an internally inconsistent result we did not rely on.our detector

Synthetic-media detector

Detector ran

The claim stands or falls on whether the media is real, so the detector ran.

AI-generated likelihood5%
Deepfake likelihood60%

Scores are the detector’s estimate for the file as uploaded. A low score does not rule out editing, a spliced caption, or footage shown out of context — those are judged above. Model v3.

How we checked the media

  • Reviewed the video's keyframes by eye for the visual signatures of generative video.
  • Compared the footage against the three existing fact-checks of the same clip.
  • Ran our automated AI-content detector on the video.

Sources

8 sources reviewed

Further reading

The full story

The thing to hold on to here is that two separate facts are both true. The first is that Donald Trump really did share this clip. Outlets including Raw Story, Yahoo News carrying a Daily Beast report, and IBTimes UK record the Truth Social post at 8:58 p.m. ET on 15 November 2026, and he followed it with a reshare captioned "May God protect our troops say AMEN", per IBTimes UK. Thai and Turkish broadcasters reported the same post independently.

The second is that the footage is not real. It first appeared on TikTok on 5 April 2026, posted by an account that routinely shares AI religious scenes, and it carried TikTok's own 'Creator labeled as AI-generated' disclaimer. Lead Stories fact-checked it that April and rated it false; Newtral and VerificaRTVE reached the same conclusion, with VerificaRTVE finding a SynthID watermark from Google's Gemini model.

That is what makes the story more than a routine correction. The clip was already publicly documented as synthetic when it was shared again seven months later. The repost also came hours after a CBS 60 Minutes report on Iranian strikes on US military and diplomatic sites, which is the context the criticism in Krassenstein's post is aimed at.

Why did it feel plausible? Because praying troops are an ordinary, emotionally charged image, and the scene is technically mundane — no explosions, no impossible physics. The plausibility came from the setting, not from the pixels.

Kernel of truth

Trump genuinely shared an AI-generated video of troops kneeling in prayer on Truth Social — the post is real, the prayer scene is not, and the clip had been fact-checked months earlier.

Timeline

  1. April 5, 2026

    The clip is uploaded to TikTok with the platform's 'Creator labeled as AI-generated' disclaimer. 4Lead Stories

  2. April 8, 2026

    VerificaRTVE publishes its check, finding a SynthID watermark from Google's Gemini model. 5VerificaRTVE

  3. April 2026

    Lead Stories fact-checks the clip and rates it false and AI-generated. 4Lead Stories

  4. November 15, 2026

    Trump shares the clip on Truth Social at 8:58 p.m. ET, then reshapes it with the caption 'May God protect our troops say AMEN'. 2Yahoo News (The Daily Beast)

How it was framed

Accurate

The headline claim — that Trump posted an AI-generated video of troops kneeling in prayer — is correct, and the post openly labels the footage as fake AI rather than passing it off as real. The 'disrespectful' element is opinion, not a factual misstatement.

  • Emotional language. "Why the HELL", "FAKE", "DISRESPECTFUL", "BEYOND WORDS HERE" — capitalised outrage framing around the factual core.
  • “Visual proof”. A screen recording of the Truth Social post is embedded so readers can see the share for themselves.
  • Missing context. The post does not mention that the clip had already been fact-checked as AI in April 2026, though that detail strengthens rather than weakens its point.

Posted on X by Brian Krassenstein, an account known for political commentary. The embedded video is a screen recording of a Truth Social post attributed to Donald Trump.

Bias

NoneSlightModerateStrong

Leans anti-Trump. The poster is a prominent liberal commentator and frames the factual point with strongly loaded language ('Why the HELL', 'so DISRESPECTFUL', 'BEYOND WORDS'). The underlying factual claim holds, but the post is packaged as partisan outrage against Trump.

Bias describes how the post frames its message — the slant of the language, selection and emphasis — and is judged separately from whether the claim is true. A true claim can still be framed with a strong slant.

How we checked

49 s
  1. 1

    Searched for coverage of Trump posting an AI video of troops kneeling in prayer on Truth Social.

  2. 2

    Searched for the earlier fact-check of the soldiers-kneeling clip on a runway.

  3. 3

    Reviewed the video's keyframes by eye for the signatures of generative video.

  4. 4

    Compared the footage against existing fact-checks from Lead Stories, Newtral and VerificaRTVE.

  5. 5

    Ran our automated AI-content detector on the video.

✓ Claim checked✓ Media checked✓ Sources checked

Why 91% confidence

High

We are highly confident because the two halves of the claim are each independently established: multiple newsrooms recorded Trump's Truth Social post, and three separate fact-checking teams identified the footage as AI, one of them via a SynthID watermark the model embeds at generation. Our automated detector returned a contradictory result and we did not rely on it, but the remaining evidence is strong enough that it does not change the verdict.

Limits. We did not open the full text of the source pages in this pass, relying on the search results and snippets. We could not independently read the SynthID watermark ourselves; that finding comes from VerificaRTVE.

What we don’t know

  • Whether Trump or a staff member selected and shared the clip.
  • Whether Trump knew the footage had already been fact-checked as AI in April 2026.
  • What prompted the timing of the repost beyond its coincidence with the CBS report.

Spot it yourself

Spot the fake

Looks suspicious, isn’t proof

  • The poster's alarmist, all-caps tone. Outrage is a reason to slow down, not a reason to assume the claim is false.
  • The partisan reputation of the account. Who posts something tells you nothing about whether the underlying facts check out.
  • The video is a screen recording of another platform's post, which feels second-hand. Many genuine stories reach you exactly that way.
  • The scene looks 'too clean' or 'too perfect' in places — plausibility of a setting is not evidence either way.

The actual evidence

  • Lead Stories had already fact-checked this exact clip in April 2026 and rated it AI-generated.
  • VerificaRTVE found a Google SynthID watermark tying the clip to Google's Gemini model.
  • Newtral independently concluded the footage was created with Google AI tools.
  • Visual inspection of the keyframes shows warped bodies, repeated faces and figures dissolving into the tarmac.
  • Independent newsrooms in several countries reported Trump's Truth Social post, confirming the share itself happened.
  1. 1

    Check whether the clip was fact-checked before

    Search a distinctive phrase like 'soldiers kneeling prayer runway AI' plus the word 'fact check'. A clip that has circulated for months usually has a paper trail.

  2. 2

    Look at the visual details, not the message

    Pause on the figures. Watch hands, faces and feet. Generative video tends to blur or morph them, and bodies sometimes melt into the ground.

  3. 3

    Check for a content credential

    Google's SynthID is an invisible watermark that survives some editing. Fact-checkers can often read it; general readers usually can't, so look for published findings instead.

  4. 4

    Ask where the clip first appeared

    If a fact-check names the original upload and the date, treat that origin story as the key fact — it tells you whether the clip was born as AI or later altered.

  5. 5

    Separate the share from the footage

    Ask two questions in order: did the person actually share this, and is the footage real? The answers can differ, as they do here.

Takeaway

A post can be factually right and still show you something fake. Separate the question 'did this happen?' from the question 'is this footage real?', because the same post can pass one test and fail the other.

Questions people ask

Did Trump really post a video of troops kneeling in prayer?+

Yes. Outlets including Raw Story, Yahoo News carrying a Daily Beast report, and IBTimes UK report that Trump shared the clip on Truth Social at 8:58 p.m. ET on 15 November 2026, followed by a reshare captioned "May God protect our troops say AMEN". The post is real; the footage is not.

Is the troops-kneeling-in-prayer video AI-generated?+

Yes. Lead Stories rated the clip false and AI-generated in April 2026, Newtral concluded it was made with Google's AI tools, and VerificaRTVE found a SynthID watermark from Google's Gemini model. The keyframes also show warped bodies and faces dissolving into the tarmac.

Where did the video come from originally?+

Lead Stories traced it to a TikTok account that posted it on 5 April 2026 with TikTok's 'Creator labeled as AI-generated' disclaimer. The account routinely posts AI-generated religious scenes. It was already debunked months before Trump shared it.

Does this mean the post criticising Trump is false?+

No. Krassenstein's post accurately says Trump shared a fake AI-generated video, and it openly labels the footage as fake rather than presenting it as real. The words 'disrespectful' and 'beyond words' are opinion, which we do not rate true or false.

Why does an AI watermark matter?+

Some AI models, including Google's Gemini, embed an invisible marker called SynthID at the moment of generation. Fact-checkers can sometimes detect it. It is stronger evidence than visual clues alone because it is attached to the file itself rather than judged by eye.

Go deeper

SynthID
An invisible digital watermark that Google's AI models embed into generated images, video and audio so they can later be identified as AI-made.
Generative AI video
Footage created by a model from a text prompt or source images rather than filmed with a camera; it often contains tell-tale warping and morphing.
Kernel of truth
The real fact or event that a false or misleading claim is built around, which is often what makes the claim feel believable.
Content credentials
Metadata and labels attached to a file that record how it was made and edited, such as a platform's 'AI-generated' tag.
For educators+
Separating the authenticity of a post from the authenticity of the media inside itHow content credentials and watermarks like SynthID work in practiceThe difference between a factual claim and a value judgement in the same postHow fact-checks build a paper trail that later coverage can cite
  • Krassenstein's post is accurate and outraged at the same time. Does emotional language make a factual claim weaker? Why or why not?
  • The clip carried an AI label on TikTok and a hidden watermark, yet it still reached a world leader's account. What does that suggest about how labels travel when content is re-shared?
  • If a public figure shares AI imagery of real people — here, service members — what responsibilities should apply, even when the image is not presented as documentary footage?

About this check

Platform
X
Posted by
@krassenstein
Posted
November 16, 2026
Age when checked
Under 24 hours old when checked
Where it spread
United States
Audience
US audience
Language
EN
Status
● Post is live
Verified
November 16, 2026, 15:01 UTC
Published
November 16, 2026, 15:01 UTC

Reach. Trump shared the clip at 8:58pm ET 15 Nov (Daily Beast); Krassenstein post 3.4K likes; Raw Story and Daily Beast pickup 16 Nov; Meiselas also called it fake AI

This verdict was generated with AI from the sources shown. AI can make mistakes — check the sources before relying on it.

Check anything yourself, in seconds.

Share any social post or video to What the Fact and get the verdict, the confidence, and the receipts.