Election Integrity Demands More Than Scary Numbers. Texas Deserves Real Answers.

There is a version of the Unite4Freedom report on the 2026 Texas primary elections that is genuinely alarming. In that version, someone with privileged access to the Texas secretary of state’s voter participation system spent the better part of three weeks injecting fake ballots, erasing real ones, and shuffling voter records in and out of the official dataset, all while election officials either looked the other way or participated. In that version, 66,146 Republican ballots were erased from existence, and the 26,000-vote margin separating John Cornyn from Ken Paxton in the U.S. Senate primary is not a real number at all, but a manufactured one. In that version, a national security emergency is already underway, and certification of the primary results would be an act of institutional fraud against the voters of Texas.

That version is not supported by the evidence Unite4Freedom has actually presented. That conclusion is not a defense of Texas election administration, which deserves scrutiny. It is not a dismissal of election integrity concerns, which are legitimate and worth pursuing with rigor. It is a methodological finding, and it matters enormously, because the difference between a compelling accusation and a provable one is exactly the difference that determines whether Texans get answers or just more noise.

The core of the Unite4Freedom analysis is a technique that sounds straightforward. Researchers downloaded repeated snapshots of the Texas SOS voter participation file during the early voting period, starting from the opening of early voting through and including election day, March 3, 2026. By comparing successive snapshots, they identified voter IDs that appeared, then disappeared, then reappeared, which they label “oscillations.” They identified voter IDs whose ballot count changed upward after they had already been recorded as having voted, which they label “injections.” They identified voter IDs present in early snapshots but missing from the final snapshot, which they label “deletions.” The numbers they report are large. 87,336 Republican voter oscillations. 66,146 Republican voter history deletions. 35,772 injected Republican ballots. On the Democrat side, 37,039 oscillations, 10,558 deletions, and 35,169 injections.

The problem is not with the counting. The problem is with what is being counted, and what those counts are understood to mean.

The Texas SOS voter participation file is not a ballot ledger. It is a live administrative roster, updated continuously by 254 separate county election offices, each operating on its own upload schedule and each processing ballots through a system that includes in-person early voting, mail ballot requests, mail ballot receipt, mail ballot acceptance or rejection, provisional ballot issuance, and provisional ballot resolution. Every one of those stages can generate a record change, a correction, a deletion of a duplicate, or a reappearance after a prior entry was flagged and removed. None of those changes, by themselves, represent the addition or removal of a vote from a certified tally. Treating them as if they do is the analytical error that inflates every number in the Unite4Freedom report from a data observation into an accusation.

Consider what happens to a single mail voter. She requests a mail ballot, and her voter ID appears in the file. The ballot is sent. She returns it, and the file is updated to show a ballot received. A county clerk reviews the envelope and finds a missing signature. The entry is flagged or temporarily removed pending cure notification. She cures the ballot. The entry reappears. That is one oscillation, and not a single vote was manipulated. Multiply that by the volume of mail balloting in a statewide primary, and oscillation counts in the tens of thousands become entirely explicable without any appeal to malicious actors inside the system.

The same logic applies to the “deletions.” When 254 counties upload data at different times on different schedules, an afternoon snapshot may include a batch of records uploaded by Harris County at noon, while the evening snapshot reflects a correction that Harris County pushed at 6 p.m., removing duplicates generated by a scanner logging the same in-person voter twice at two different poll books. The voter’s record is not deleted. The duplicate is. The snapshot comparison, however, records this as a deletion event. Unite4Freedom’s report acknowledges that county and precinct details were not consistently available in the data, which means the analysis cannot distinguish between a real record removal and a data synchronization artifact produced by routine county-level corrections. Without that distinction, the deletion count is not evidence of voter suppression or ballot erasure. It is a count of all record changes, regardless of cause.

The February 17 chart above is the most visually striking element of the Unite4Freedom presentation, and it is also the most straightforwardly misleading. The chart tracks, across dozens of snapshots taken from February 19 through March 3, how many ballots were recorded as having been cast on February 17. The count moves up and down seven times across those snapshots. Unite4Freedom presents this as proof that the vote count went backwards seven times during the voting period, which they argue is mathematically impossible unless someone inside the system is altering records in real time.

What it actually shows is that the recorded attribution of ballots to specific cast dates in a live multi-county administrative file takes time to stabilize. A mail ballot postmarked on February 17 may not arrive at the county clerk’s office until February 20. It will then be logged with a cast date of February 17, because that is when the voter cast it, but the log entry will appear in a snapshot taken on February 22. That is not a manipulation. That is postal transit. Similarly, a county that corrects an upload error on February 21, removing a batch of records that were accidentally double-posted with a February 17 date, will cause the February 17 count to drop in that snapshot. That is not a deletion of votes. That is data hygiene. The chart proves that the February 17 count fluctuated. It does not prove why, and the why is everything.

The Bernie Reyna table above illustrates a separate but related confusion. The table shows election night vote tallies for three candidates in Texas’s 10th Congressional District as reported incrementally, alongside the final count from the Austin-American Statesman. At 65% of precincts reporting, candidate Marshall shows 10,013 votes. At 68% reporting, Marshall shows 17,969, a jump of nearly 8,000 votes with only 3% more precincts checked in. At 99% reporting, Marshall’s total falls to 12,131. These fluctuations look suspicious on paper, but they are a well-documented feature of partial-precinct election night reporting. Large precincts report all at once. When a heavily populated precinct checks in, it can shift totals dramatically in a single update. When a subsequent correction adjusts the total expected precinct count upward, the apparent % of results in can drop even as the raw vote total climbs. None of this is evidence of fraud. It is evidence that partial election results are inherently unstable and should not be read as final until certification.

This brings us to the most consequential question the Unite4Freedom report raises, even if it raises it badly. The reported margin between Cornyn and Paxton is approximately 26,000 votes out of roughly 2.1 million cast in the Republican primary. The report’s figures, if taken at face value, suggest that the volume of anomalous record changes is larger than that margin. The conclusion Unite4Freedom draws, that the results should not be certified, does not follow from those figures for the methodological reasons described above. The conclusion that follows is more modest, but it is still important: a margin this narrow, in a race this consequential, in a system this complex, warrants a formal and thorough investigation, not by volunteer researchers working from public snapshots, but by the law enforcement and election administration bodies with actual subpoena power and system access. It should be noted that I am a Paxton supporter.

Texas law is clear on this point. The Texas Attorney General is legally obligated to investigate credible reports of election fraud or irregularities referred to his office. That obligation does not require the referring party to have already proved fraud. It requires only that a complaint be filed, that it be based on identifiable data, and that it present a reasonable basis for inquiry. The Unite4Freedom report, whatever its methodological limitations, presents a body of identified data anomalies in the official SOS voter participation file. That is a reasonable basis for referral.

Both the Republican Party of Texas and the Texas Democratic Party should file formal complaints with Attorney General Ken Paxton’s office, attaching the Unite4Freedom analysis as supporting documentation and requesting a formal inquiry into the specific anomalies identified. This is not an endorsement of Unite4Freedom’s conclusions. It is a recognition that the proper institutional response to documented anomalies in a live election dataset is a formal investigation, not a social media declaration that the results are fraudulent, and not a dismissal of the concerns because the methodology is flawed.

What the investigation should actually include is worth specifying, because the Unite4Freedom analysis, to its credit, has identified genuine questions even if it has answered them incorrectly. Any credible formal inquiry should begin by establishing a baseline for expected record change volumes in a live multi-county administrative file of this size across a comparable historical period. Without knowing how many oscillations, corrections, and deletions are normal for the Texas SOS system in an election of comparable size, the numbers Unite4Freedom reports are impossible to evaluate. They may be extraordinary, or they may be routine. No one can say from the report alone.

The investigation should then examine the directionality of all changes, meaning whether the deletions and injections systematically benefited one candidate across the relevant precincts and counties. If this were a deliberate manipulation, the changes would not be randomly distributed. They would cluster around competitive precincts in ways that advantage a specific outcome. If no such clustering exists, the manipulation hypothesis becomes dramatically less plausible. If clustering does exist, it demands explanation.

Third, investigators should cross-check the final certified vote count against any independently verifiable totals, including precinct-level paper records, county canvass reports, and voting machine tapes. The question that matters is not whether the administrative file changed during the early voting period. It is whether the final certified tally is accurate. Those are different questions, and the Unite4Freedom methodology conflates them.

Fourth, the investigation should verify each candidate against any unfamiliar or recently constituted organizational entities whose role in the process is not yet well understood, ensuring that no external access to the SOS data system occurred outside of normal administrative channels. Fifth, investigators should examine the specific instances where the SOS data file format changed during the early voting period, which Unite4Freedom notes and which deserves explanation. Format changes to a live election data file during an active voting period are unusual, and the SOS should be required to account for them. Finally, a complete audit of mail ballot request, receipt, rejection, and cure records should be cross-referenced against the oscillating voter IDs to determine what fraction of those oscillations are explained by mail ballot processing workflow, and what fraction, if any, remain unexplained after that accounting.

This is what a rigorous investigation looks like. It is more difficult than downloading public snapshots and counting record changes. It requires institutional access, cooperation from county election officials, and independent forensic expertise. It is also the only investigation that would actually answer the question Texans deserve to have answered, which is whether the certified results of the 2026 Texas primary elections accurately reflect the will of the voters who participated.

The Unite4Freedom report, and the 𝕏 posts amplifying it, treat the appearance of anomaly as proof of crime. That is not a conservative principle. Conservatism at its best is skeptical of hysteria and respectful of the difference between a serious claim and a proven one. The serious claim here is that the Texas SOS voter participation file generated anomalous-looking record changes during the 2026 primary early voting period, and that those changes have not been publicly explained by election officials. That claim is worth pursuing. The proven claim, meaning what the available evidence actually establishes, is considerably more limited. It establishes that a live administrative database processed records from 254 counties over three weeks, and that successive snapshots of that database reflected changes consistent with routine mail ballot processing, duplicate correction, and county upload sequencing, as well as potentially something else, something that a real investigation with real access could determine.

Texas voters, Republican and Democrat alike, deserve that investigation. They deserve it conducted by institutions with the authority to compel records and the expertise to interpret them. They do not deserve a premature verdict rendered by a methodology that cannot distinguish between a manipulated ballot and a corrected duplicate entry. And they do not deserve the corrosive certainty that the election was stolen before any competent authority has examined the actual evidence. That certainty, broadcast across 𝕏 and amplified by well-meaning patriots, makes a real investigation harder to conduct by poisoning the well before the water is even tested.

File the complaint. Demand the investigation. Insist on a full methodological accounting that includes baselines, directionality analysis, certified tally cross-checks, and a complete audit of mail ballot processing records. That is the path to an answer. It is slower and less satisfying than declaring a national security emergency on a Tuesday afternoon, but it is the only path that actually leads somewhere, and Texas Republicans, headed into a May 26 runoff that will determine the Republican nominee for US Senate, cannot afford to walk any other one. P.S. Transparency demands a few additional disclosures, because the argument above is stronger for them, not weaker.

First, I am not a disinterested observer of Texas election administration. I am a supporter of Ken Paxton, and I recognize that the anomalies Unite4Freedom has identified, whatever their actual cause, tend to generate a political narrative that benefits him. I am aware of that. I am not willing to amplify unfounded claims simply because they advantage my preferred candidate. The methodological critique in this piece stands on its own terms, and I would apply it identically if the numbers appeared to favor Paxton at Cornyn’s expense. Intellectual honesty is not a partisan convenience.

Second, my skepticism of the Unite4Freedom methodology does not extend to a general confidence in the Texas Secretary of State’s office. I have significant, independent concerns about Jane Nelson’s tenure that have nothing to do with snapshot databases or oscillating voter IDs. Nelson is currently spending millions of Texas taxpayer dollars fighting the Republican Party of Texas in federal court over whether Democrats may participate in GOP primaries. She is the Republican Secretary of State, and she has hired Washington lawyers, reportedly from the never-Trump wing of the legal establishment, to litigate against the party she nominally serves. That is a remarkable posture for a Republican SOS, and it deserves far more scrutiny than it has received.

Third, and more troubling on the election integrity front specifically, I have grave and documented concerns about Texas’s use of the Help America Vote Verification system, commonly known as HAVV. That system is used to cross-check voter registration data against Social Security Administration records. The data I have reviewed contains serious red flags that I cannot yet fully characterize publicly, because every attempt I have made to obtain clarification has been met with stonewalling or silence. When I contacted the SOS’s office directly, I received no substantive response. When I filed open records requests, those requests were simply ignored, and it appears that under current Texas practice, the SOS’s office is not even required to notify a requestor that it has decided not to respond. The Social Security Administration, to its credit, did at least respond to my inquiry, informing me that its correspondence with the Texas SOS regarding HAVV irregularities cannot be disclosed. That response tells me the correspondence exists. It does not tell me what it contains, and I believe Texans are entitled to know.

These are the reasons I believe the referrals to Ken Paxton’s office are not merely warranted but urgent, and why they should come from the Republican National Committee, the Democratic National Committee, and Unite4Freedom simultaneously. A bipartisan referral forecloses the argument that this is partisan score-settling. It puts the full institutional weight of both parties behind a demand for answers. And it places the legal obligation to investigate squarely where it belongs, with the Attorney General’s office, which has the subpoena power, the system access, and the statutory duty to act. As of this writing, it is my understanding that none of those three organizations has filed a formal complaint with the AG’s office. That is an omission that should be corrected immediately, by all of them, and without waiting for the other to go first.

If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe.

Sponsored by the John Milton Freedom Foundation, a nonprofit dedicated to helping independent journalists overcome formidable challenges in today’s media landscape and bring crucial stories to you.

READ NEXT: The Left’s SAVE Act Scare Campaign Doesn’t Hold Up