Fact check: How do election officials verify the integrity of voting systems?

1. Why Michigan’s post-election scrutiny made headlines — and what it actually showed

Michigan’s statewide audit and recount process in 2024 combined a state-led recount, a statistical statewide ballot audit, and targeted procedural reviews to test whether voting systems produced accurate outcomes. The official reports conclude the systems were “secure and accurate,” noting that among nearly 120,000 randomly sampled ballots the hand counts diverged from machine tabulation by 33 votes (about 0.03%) — a margin election officials describe as consistent with expected operational variance rather than systemic failure ^{[2] [1] [4]}. These findings were published in mid-October 2025 and framed by state officials as validation of existing controls, while critics urged broader contextual data about chain-of-custody and precinct-level anomalies be published.

2. The toolkit: pre-election tests, logic-and-accuracy checks, and public demonstrations

Election operations hinge on a sequence of technical checks designed to catch errors before Election Day: pre-election equipment acceptance testing, logic and accuracy (L&A) tests, and public zero-precinct or mock election runs. Guidance and local practices vary; jurisdictions like Boulder County publicly invite observers and bipartisan teams to participate in tests to bolster transparency and trust, showing ballots marked by election judges are counted consistently by machines ^[5]. These procedures are standard across jurisdictions and are intended to reveal configuration errors, ballot-definition mistakes, or hardware failures prior to live voting.

3. Post-election audits: statistical sampling and hand counts that measure real-world performance

Post-election verification typically uses statistical audits — including risk-limiting audits (RLAs) — and hand-count validation of samples to quantify discrepancies between machine totals and physical ballots. Michigan’s audit is an example of a statewide statistical approach producing a precise discrepancy metric (0.03%), illustrating how sampling can detect small differences without recounting every ballot ^[1]. Proponents argue this balances accuracy and cost; opponents and watchdogs insist audits must be comprehensive, public, and include chain-of-custody documentation to rule out localized irregularities that aggregate statistics might obscure.

4. Federal and state roles: cybersecurity, standards, and policy leadership

Federal agencies and state actors share responsibilities: the Cybersecurity and Infrastructure Security Agency (CISA) provides cyber guidance, risk assessments, and coordination, while state secretaries of state and election security programs fund tools and interagency cooperation to counter interference and misinformation ^{[6] [7]}. Policy analysts recommend that state leaders strengthen resiliency through sustained investments, training, and statutory updates to match evolving threats ^[3]. Tensions arise when national advice meets diverse state laws and procurement policies; the federal role is advisory and facilitative rather than prescriptive, which shapes uneven adoption.

5. Testing certification: lab approvals, state acceptance, and evolving technical standards

Voting system certification involves independent testing laboratories recognized by federal standards and state-level acceptance processes that require testing for performance, security, and usability before deployment. The U.S. Election Assistance Commission’s test-lab model and state acceptance testing create a layered approval path intended to catch design flaws early ^[8]. However, certification timelines, differing state standards, and the pace of cyber threats mean certification is a baseline, not a continual guarantee; jurisdictions rely on routine testing cycles and post-deployment monitoring to maintain integrity over a device’s service life.

6. Transparency and public confidence: why observation and openness matter

Public demonstrations, bipartisan test teams, and published audit results are central to building confidence because technical assurances alone do not persuade all voters. Boulder County’s practice of inviting community members to watch equipment tests exemplifies how procedural transparency addresses skepticism ^[5]. The Michigan audit reports’ publication in October 2025 provided factual metrics but also sparked requests for more granular data and clearer communication about what small discrepancies mean, demonstrating that trust hinges on accessible explanations and responsiveness to critiques.

7. Divergent narratives and the limits of single-source claims

Summaries claiming that audits “confirm” absolute security can overstate what testing achieves; audits measure accuracy in sampled or audited contexts and reduce risk but do not eliminate every possible failure mode. Michigan’s 0.03% difference is evidence of functioning controls in sampled ballots ^[1], while federal guidance emphasizes ongoing investment to address emergent threats ^[3]. Observers with political agendas may selectively highlight audit results or procedural gaps to support broad conclusions; assessing system integrity requires synthesizing certification, testing, audit data, and transparency practices across jurisdictions.

8. Bottom line: layered defenses plus public scrutiny produce the best assurance

Verification of election-system integrity is neither a single act nor a one-time report; it is a multi-layered process combining pre-election testing, independent certification, public and bipartisan procedures, statistical post-election audits, and federal–state cybersecurity coordination. Recent state reports and local practices illustrate the strengths and limits of these layers — small discrepancies can appear and are measurable, but sustained resilience depends on state leadership, transparent reporting, and continuous adaptation to new threats (p1_s1, ^[6],