How a False Pentagon Explosion Became Breaking News

A column of dark smoke rose beside a pale building. The image circulating on Twitter on May 22, 2023, looked like evidence of an emergency. Its captions supplied the location: the Pentagon. There had been no such explosion. The local fire department publicly said so, and a Pentagon spokesperson separately confirmed that to ABC News. ABC News, May 23, 2023.

The interesting part is the distance between those two versions of the morning. One consisted of a picture and an alarming claim. The other involved checking whether the event had happened. To understand how a misleading story travels, follow the places where the claim appeared and ask what evidence, if any, each appearance added.

The beginning we can actually document

The original creator of the image has not been established by the sources reviewed here. AFP identified a post from the account CBKNEWS as the earliest tweet its investigators found carrying the picture. It explicitly left the image’s original source unresolved. That qualification is essential: finding an early public copy does not establish who made the image or whether it had circulated somewhere else first. AFP, May 22, 2023, updated April 29, 2024.

That gap should remain visible in any reconstruction. An investigator’s earliest result depends on what survives and what can be retrieved. A deleted post, a private message or a copy on another platform could change the chronology. Here, the evidence is sufficient to follow public circulation without assigning an inventor, a sponsor or a motive.

The picture acquires more publishers

By that morning, several accounts were circulating the image. Lead Stories preserved a May 22 post from WhaleChart with the caption “BREAKING: Explosion near Pentagon.” AFP documented amplification by RT and Zero Hedge, along with copies across social platforms. These records show multiple appearances of the false claim; they do not establish every handoff between the accounts. Lead Stories, May 22, 2023; AFP.

Some amplifying accounts carried blue checkmarks. ABC’s reporting placed the episode in the context of Twitter’s paid verification system, under which users could purchase a check. That symbol could offer a familiar appearance of credibility without verifying the alleged explosion. ABC News.

This is where a reader needs to count evidence differently. Five accounts carrying the same image may provide five opportunities to see it. They do not automatically provide five observations of the event. Before treating a second account as corroboration, check whether it supplies an independent witness, another camera angle or reporting from the location. A new caption can leave the underlying evidence exactly where it started.

The words used to frame a post deserve the same scrutiny. “Breaking” describes urgency. It supplies no information about how a claim was checked. An initial report can be worth investigating, but the label cannot do the investigation on the publisher’s behalf.

How the correction entered the chain

One revealing turn came from an account that had helped circulate the claim. In a surviving archived post, OSINTdefender said it had deleted its Pentagon-explosion post because the reports appeared to rest on one picture, with no other evidence of the event. The statement records an amplifier recognizing the weakness of the available support. It does not establish where the image originated. Archived OSINTdefender post, May 22, 2023.

At 10:27 a.m. on May 22, Arlington Fire & EMS posted a more decisive correction, identifying both itself and the Pentagon Force Protection Agency. It said there was no explosion or incident at or near the Pentagon reservation and no immediate danger to the public. A reproduction of that public statement, including its timestamp, survives on printed page 121 of a House hearing appendix. Arlington statement reproduced in the congressional record.

There was also reporting from the actual location. AFP said its journalist at the Pentagon saw neither the smoke depicted in the circulating image nor an emergency response after the rumor spread. The agency obtained a Defense Department statement rejecting the report. AFP.

The correction had a different evidentiary basis from the circulating picture: a statement from the responsible institutions, supported by a reporter’s observations at the location. The observations alone would have limits; a reporter cannot see every possible incident. Together, the checks addressed the specific claim being circulated rather than relying on how convincing the image looked.

Keep the remaining questions separate

The question of whether an explosion happened can be resolved more firmly than the question of how the picture was made. ABC described visual inconsistencies consistent with AI generation. Lead Stories tested two near-identical versions with an AI-image detector and reported substantially different scores. The safe description is a fabricated image that appeared to be AI-generated; those findings do not identify a particular generator or its operator. ABC News; Lead Stories.

Likewise, circulation by multiple accounts is insufficient to establish a coordinated operation. This reconstruction does not determine whether particular publishers were deceived, careless or acting deliberately. Each explanation would require additional evidence. The falsehood of the explosion claim does not settle those separate questions about responsibility and intent.

A useful check before the next share

The Pentagon case suggests a short sequence for readers encountering an alarming image:

• State the claim precisely. What supposedly happened, where, and when? A dramatic image needs a verifiable account of what it depicts.
• Trace the source behind the source. Follow links beyond the account that brought the story to you. Record where the trail stops.
• Look for independent evidence. Another copy of the same picture adds distribution. A separately obtained observation may add corroboration.
• Check relevant local reporting and official channels. Treat their statements as evidence to evaluate alongside observations and records.
• If you shared an error, publish a visible correction with the supporting source. Deleting your copy cannot update copies other people already saved or shared.

These are recommendations drawn from the case, rather than a guarantee that every fast-moving report can be settled immediately. Sometimes the responsible conclusion is that the evidence is insufficient and sharing should wait.

The enduring lesson of this episode is how little a claim has to gain, evidentially, while gaining publishers. Following the chain makes that visible. Ask at each stop what was independently checked. When the answer is only that somebody else posted it, the central verification task is still unfinished.