fact-checking

  • |

    The Vote-Flipping Rumor Is Having Another Election-Day Dress Rehearsal

    ByHolden McGroin

    Election Day has brought a familiar internet production: a voter reports a touchscreen selection that does not look right, and the rumor machine rushes in wearing a trench coat labeled NATIONAL CONSPIRACY. The reported mismatch matters. The leap from “a screen may have shown a selection incorrectly” to “votes are being deliberately switched everywhere” is a different claim, and it needs different evidence.

    What the reports establish—and what they don’t

    Recent coverage and fact-checks describe complaints involving touchscreen selections in multiple states, including Tennessee and Texas. FactCheck.org and PolitiFact reviewed reports, official responses and explanations from election experts. The reviewed sources found no evidence of deliberate or systematic vote flipping. That does not prove every machine worked perfectly. It marks the difference between a reported problem and a claim about its cause and scale.

    Officials and experts cited possible user error or calibration problems as explanations for some reported mismatches. Those possibilities should not be used to wave away a voter’s concern. A machine can be imperfect without being part of a national plot, just as a loose shopping-cart wheel is not evidence that the grocery store has declared war on ankles.

    There is a check before the rumor gets a crown

    FactCheck.org describes procedures that let voters review selections and correct them before casting a ballot, including review of a printed paper ballot in systems that use one. Processes vary by voting system, so this is not a promise that every machine works the same way or that safeguards eliminate every problem. It is a reason to treat a mismatch as something to flag and check—not as proof that a completed ballot was secretly rewritten.

    That distinction matters: a screen complaint deserves a clear response, but it does not establish that a ballot was cast incorrectly, much less that anyone deliberately changed it. The evidence cited in these fact-checks does not support turning individual reports into a nationwide account of systematic switching.

    PolitiFact also reports that some footage shared as evidence of current problems dates to 2018 and 2019. Old video can show an old incident; by itself, it cannot establish what is happening today. The algorithm wore a trench coat, found a clip in the attic, and tried to pass it off as a live witness.

    Follow the thread, check the knot

    Voters deserve systems they can inspect and officials who answer specific questions clearly. That means taking a reported mismatch seriously, explaining how review and correction work, and not asking the public to replace evidence with trust. It also means not treating a viral post as a forensic report just because it arrives with dramatic music and twelve exclamation points.

    The panic boutique sells certainty in both directions: either “nothing can ever go wrong” or “every glitch proves a plot.” Neither helps the person at a screen trying to make sure the recorded selection matches the one they intended. Check whether footage is current and whether a claim is supported by election officials or credible fact-checking. Keep the conclusion proportional to the evidence. The corkboard can stay; maybe calm down can keep the highlighter.

  • |

    The Ballot-Mark Rumor That Treats Routine Election Work Like a Plot Twist

    ByHolden McGroin

    The viral warning has a tidy little thriller plot: an election worker writes on a ballot, therefore the ballot is doomed, therefore somebody should put a red string across the kitchen wall. The problem is the jump from “I saw a mark” to “the vote is disqualified.” A pen mark may look mysterious when it appears without context. Mystery, however, is not a procedure—and a rumor’s confidence is not evidence.

    What North Carolina says the marks can mean

    North Carolina’s State Board of Elections addressed the claim on October 14, saying it had resurfaced in multiple counties during the current early-voting period. The board says election workers may make notations for routine administrative reasons, including sorting ballots by precinct and helping retrieve a ballot if needed. In North Carolina, a worker’s notation alone does not stop a ballot from being counted. That is an explanation of specific procedures, not a magic statement that every mark on every ballot everywhere is harmless.

    That distinction is where the rumor’s corkboard starts sneezing. An unfamiliar mark is something a voter might reasonably want explained; the sinister conclusion attached to it is a separate claim. Confusing the two turns uncertainty into a ready-made plot twist. It also asks voters to infer what a technical procedure means without first checking the rules and explanations from the officials responsible for administering it.

    There isn’t one rulebook for every jurisdiction

    The Associated Press reports that ballot procedures vary by jurisdiction and that election officials in multiple states reject the blanket claim that a poll worker’s mark automatically disqualifies a ballot. PolitiFact’s October 16 fact-check also examines the claim in North Carolina and other states. These reports do not establish that every marking is acceptable or that every state handles ballots the same way. They do undercut the all-purpose warning that any worker notation means a ballot will not count.

    For anyone asking about a particular ballot or local procedure, the useful answer is local: check the official guidance from the election office responsible for that ballot. Rules and processes can differ. North Carolina’s explanation is useful for North Carolina; it is not a universal decoder ring. That may be less thrilling than a rumor with a ring light, but it is a better way to get a reliable answer.

    Follow the thread, check the knot

    The panic machine skips the boring middle: what was written, why workers use a notation, and what local rules say happens next. The supplied reporting does not show that this rumor caused ballots to be rejected or that it began in North Carolina. What the sources do show is that a worker’s notation does not automatically mean disqualification, and that jurisdiction matters. The highlighter labeled “maybe calm down” can stay on the desk. When a post turns an administrative mark into a national plot, check the local facts before buying the premium string.

  • |

    A Real Rally Assault Got an AI-Generated Photo Makeover

    ByHolden McGroin

    At a Trump rally in Youngstown, Ohio, on October 12, a real altercation became attached to a fabricated visual. An image shared on X on October 14 purported to show a woman in a “Release the Epstein Files” shirt being punched. Fact-checks published October 15 reported that the image was AI-generated. The incident happened; that particular image was not a genuine record of it. The corkboard can keep both facts pinned up without tying them together with premium string.

    Real event, invented image

    Lead Stories reported that an OpenAI tool detected a SynthID watermark and that another AI detector flagged the image. Its fact-check also compared the image with coverage of the incident and an Associated Press photo. That combination matters: the conclusion was not simply “a detector beeped, case closed.” A detector result or watermark finding can be evidence to weigh, but neither should be treated as an infallible witness. Even the algorithm wore a trench coat, apparently, and still needed its paperwork checked.

    Gizmodo also reported that the image was fake, while noting that AI-checking tools produced confusing results. That is a useful brake on detector worship. A tool can produce a clue worth investigating without settling every question about an image’s origin or what it depicts. And a score, by itself, does not show where an image came from or establish whether its specific visual details match the event. Lead Stories’ comparison with reporting and an AP photo gave readers evidence to consider beyond a number floating alone in the fog.

    Check the picture, not just the story

    Contemporaneous coverage helps separate the event from the image attached to it. Al Jazeera published video coverage of the Youngstown rally altercation on October 13, before the October 15 fact-checks. That supports the basic distinction: there was a real incident to report, but the existence of an incident does not authenticate every dramatic picture later offered as proof of what happened. The video and the AP photo are useful context; they are not a magic stamp validating every image associated with the rally. The existence of a fire does not certify every photograph of the smoke.

    That is the little trap in this case. When people already know an event occurred, a fabricated image can borrow that knowledge and arrive wearing the event’s name tag. No need to speculate about who made or shared this one, or why. The practical check is narrower and stronger: compare the specific visual claim with reliable photos, video, and reporting from the time. Ask whether those sources support the details shown—not merely whether the underlying event was real. Follow the thread, but check the knot.

  • |

    The Numbers Entered the Fog

    ByHolden McGroin

    When a statistic steps up to the microphone sounding like the whole case, I reach for my tiny courtroom gavel—and then check the label on the measuring instrument. The Associated Press fact-checked claims in President Donald Trump’s September 22, 2026, address to the United Nations, examining figures involving munitions assessments, oil reserves and drug seizures. Those are real categories of measurement, but they do not become one universal proof just because they arrive in a confident voice.

    A munitions assessment concerns munitions; an oil-reserve figure concerns reserves. Neither automatically settles a broader claim outside what that measure can establish. AP’s review is useful here not because every number is made up, but because the argument attached to a number may travel farther than the evidence. My corkboard is willing to entertain a pattern. The highlighter marked “maybe calm down” insists we ask whether the threads actually connect.

    Drug seizures offer the clearest example of the leap. Seizure data records what authorities seized; by itself, it cannot tell us how much trafficking went undetected. That distinction matters. A recorded seizure is evidence of a seizure, not a complete inventory of everything that moved unseen. Treating the first as proof of the second is like counting the packages a mail carrier caught in the rain and announcing you know every letter that reached the neighborhood.

    That is the rhetorical fog machine: put unlike measures beside one another, give them the lighting of certainty, and let the audience feel that a much larger conclusion has been proven. The number may be accurate within its lane. The trouble starts when the speech quietly moves it into another lane and asks it to testify there. AP’s fact-check gives readers a reason to separate the stated measure from the larger claim being built around it.

    Public figures should be questioned, and evidence should be taken seriously—including evidence from institutions that have not earned automatic trust. But skepticism is not the same as treating every confident statistic as a master key. Ask what was counted, what the count can establish, and what remains unknown. That small audit is less dramatic than a conspiracy corkboard, but it leaves less room for someone else to sell certainty by the yard. The statistic may be real; the fog machine is doing the extra work.

    Sources

  • |

    When Twenty-Five Million Meets the Calculator

    ByHarlan Quill

    My library card objects to calling 25 million people a headcount when the paperwork contains encounters, expulsions, and estimated gotaways. Those are different categories, and some may overlap. An encounter is not automatically a unique person, an expulsion is not a new arrival, and an estimate requires the modest qualification that it is, in fact, an estimate. The calculator remains unimpressed by the volume of the announcement.

    This is how political number inflation works: collect every large-looking figure, place them in one bucket, and announce that the bucket is a population. The public gets a frightening total; the media ecosystem gets a louder segment; ordinary people get another argument built on accounting performed with a fog machine. The practical rule is simple: before accepting a giant number, ask what it counts, whether categories overlap, and whether anyone can produce a receipt. In this case, the number appears to have arrived wearing a costume and left without paying the bill.

  • |

    The 278,000-Voter Ghost Story Has a Spreadsheet Problem

    ByHolden McGroin

    My corkboard has seen some numbers in its day, but even it put on a little safety vest when the White House presented an estimate of approximately 278,000 alleged noncitizen voters on July 16. The number arrived dressed as evidence, and the national panic desk immediately stamped it PROOF in red ink. The trouble is that a large number is not automatically a large fact. Without the method, the matches, and the limits of the comparison, it is just a very confident number wearing a government lanyard.

    FactCheck.org reported that the Department of Homeland Security did not disclose the methodology behind the estimate. That is not a minor footnote; it is the part where the public learns what was actually counted. Were these registration records, database matches, people who cast ballots, or something else? Those categories are not interchangeable, unless arithmetic has joined the witness-protection program.

    Experts cited by FactCheck.org also warned that comparisons against commercial databases can produce large numbers of false matches. A name, address, citizenship record, or outdated file can collide in the machinery and emerge as a suspicious-looking human being. The administration presented the figure as evidence that elections had been compromised, but the public was not given enough information to test whether the estimate measured voting, registration, mistaken identity, or a blender full of all three.

    That is how the misinformation loop gets its premium string: an official statement supplies the authority, social media supplies the repetition, and every repetition makes the original uncertainty harder to see. Ordinary voters are then asked to fear a compromised election while the underlying matches remain offstage. Nobody is saying election records should never be audited. The point is that an audit requires inspectable work, not a dramatic number followed by a request for unquestioning patriotism.

    The responsible standard is painfully ordinary: show the method, define the claim, identify the matches, and distinguish registration from verified voting and fraud. Until then, the country has not been handed proof. It has been handed the authority of a federal spreadsheet without being allowed to inspect the spreadsheet. The national séance did not summon evidence; it summoned a spreadsheet nobody is allowed to inspect.

  • |

    The Internet Put Mitch McConnell on Life Support, Then Asked Us to Inspect the IV

    ByHolden McGroin

    My corkboard has one rule written in permanent marker: suspicious is not the same as proven. The internet recently replaced that rule with a magnifying glass and began auditing Mitch McConnell’s hospital photos like the republic had deputized every phone owner as chief surgeon. An earlier viral photo was identified as AI-generated through Google’s SynthID, according to TechCrunch. That was a real fact. Naturally, the panic machine treated it as a master key for every blurry tube, reflection, hand, and hospital-room detail that came afterward.

    Then came the later viral photo, where the evidence got less convenient. Lead Stories found no evidence that it came from McConnell’s staff, but also noted that suspected visual glitches did not conclusively prove it was AI-generated. That distinction is the whole story, unfortunately, because the internet hates a distinction. “Not verified” is too boring for the algorithm, while “definitely fake” arrives wearing sunglasses and carrying a ring light.

    Limited official information helped create the vacuum, but the vacuum did not remain empty. It filled with amateur medical detectives, conspiracy-minded commentators, and people who apparently believe an IV bag becomes a confession if you zoom in far enough. The public did not receive a clear health update; it received uncertainty with a suspicious bend in the tubing. Every ordinary viewer was invited into a forensic group chat where confidence counted as credentials and the loudest screenshot analyst became attending physician.

    There is a civic cost to this fog. When powerful institutions share little, people understandably look for signals, and platforms reward the most dramatic interpretation of whatever they find. That does not make every suspicion foolish, and it does not make every official silence respectable. It does mean the public deserves facts rather than a scavenger hunt, while the people selling the panic deserve scrutiny too. Confusion is profitable content: it keeps everyone refreshing, arguing, and mistaking engagement for evidence.

    So the pattern is not that every McConnell hospital photo was fake, authentic, or secretly carrying a diagnosis. The pattern is that one confirmed AI hoax gave the crowd a real foothold, then uncertainty turned that foothold into a trampoline. Follow the thread, but check the knot. Otherwise the internet will keep promoting itself to chief of surgery after being handed nothing more than an IV bag and a reason to panic.

  • |

    Read the data, not the scare line

    ByHolden McGroin

    That “MIGRANTS AND MURDER” chant doesn’t start as a dataset—it starts as a dopamine vending machine: “11,888 murderers… they allowed them into our country.” Then the brain goes, “Cool, I’ve got a villain origin story,” and nobody asks where the arrow ends and the accounting begins.

    But “FALSE” doesn’t mean “oops, vibes.” It means “CHECK THE SOURCE,” and the checks point out how the number gets read wrong when you flatten timing/convictions/ICE bookkeeping into one clean “allowed in = murderers” storyline. So yeah—if you want to be a patriot instead of a hostage to a scare caption, follow the boring thread: READ THE DATA, NOT THE SCARE LINE.

  • |

    Releases ≠ Crossings: Border “Zero” via Definition Hopping

    ByHugh Jass

    When the headline starts with “Not one illegal alien was able to get into our country,” it’s running a paperwork shell game: it treats “release” like “crossing,” so the conclusion looks spotless even though the process is doing separate, bureaucratic things. Definition hopping turns a messy border into a clean story—if you never change the form.

    The “truth” version stamps the logic as “That is an exaggeration.” It points out that CBP says encountered migrants weren’t released into the country, then admits that some people still evaded arrest at the border—and, crucially, that a “zero-gotaway day” had not happened yet. Bottom line: NO RELEASES ≠ NO CROSSINGS.

  • |

    Only His Credit’s Up

    ByJustin Jest

    “Factory construction is up” is the kind of sentence you say into a microphone right before reality starts yelling back. The claim wears a brave little grin, and then the monthly spending line goes down, down, down—so the whole thing gets stamped FALSE like a parking ticket for narrative lawbreaking. This is what happens when campaign power treats numbers like optional background music and assumes workers will applaud the key change anyway.

    The funniest part isn’t even the mismatch; it’s the credit laundering. If there was an earlier surge—allegedly under Biden in 2023—the system still tries to bill the current guy for the improvement, because in billionaire-candidate logic the only trend that matters is “my name goes on it.” Reality doesn’t have to cooperate. It just has to keep being inconvenient.

End of content

End of content