Election Notice:  Election Night Returns
IN-PERSON SERVICES FOR THE OFFICE OF THE TEXAS SECRETARY OF STATE HAVE MOVED TO 400 W. 15TH STREET. PARKING DETAILS AND MORE INFORMATION

Election Advisory No. 2026-30

To: Election Officials
From: Christina Adkins, Director of Elections
Christina Adkins signature
Date: November 4, 2026
RE: Risk-Limiting Audit Procedures

A Risk-Limiting Audit (RLA) is a type of post-election audit that is designed to review a statistically significant number of paper ballot records to verify that the outcome of the election was correct. 

In Texas, this process is conducted by selecting a statewide contest (or contests) to be audited, identifying a random selection of ballots across the state that will be included in the audit, pulling those ballots from their original ballot boxes, counting the ballots, and then aggregating the results of the count from those ballots across the state to determine whether there is strong evidence that the reported outcome of the election is correct.

Under Chapter 127, Subchapter I of the Election Code, the Risk-Limiting Audit process applies to any election occurring after August 31, 2026 that contains a statewide race or measure and in which an auditable voting system is used.  Accordingly, the first official statewide RLA in Texas will be conducted for the November 3, 2026 election.

Our office has been engaged in the pilot program for the RLA process with counties since Senate Bill 598—passed during the 2021 Legislative Session—created this program.  While we have discussed these procedures with counties in the course of many pilots, regional trainings, annual conferences, webinars, and other hands-on training opportunities, this document is designed to consolidate those resources and provide written guidance on the procedures for conducting an RLA in Texas.

After election day, our office will provide more specific training and instructions for counties on how to use the RLA software (Arlo).

General Background

A risk-limiting audit is a type of post-election audit that has been adopted in several states as an additional method of verifying the accuracy of electronically tabulated election results.

While there are several different types of RLAs that can be conducted, the RLA method that is used in Texas is a ballot polling audit (PDF).  This method involves ballots being randomly selected and retrieved, then counted throughout the entire state, thus providing statistical certainty that the outcome of the election is correct.

One key feature of RLAs is that they are transparent and repeatable.  A dice roll is used to create a seed number that is used by a pseudo-random number generator to determine which random ballots will be reviewed for the audit.  Notably, pseudo-random number generator seeds will generate the same sequence of random ballots every time.  This repeatability allows for independent verification that the correct ballots were reviewed during the RLA.  For detailed information, see “How the Random Seed is Generated” below.

The Secretary of State’s Office will provide each county with a specific list of ballots to retrieve from their ballot boxes.  Because the ballots are randomly selected, some counties may not be assigned ballots to pull.  On the day designated by the Secretary of State, counties that have been assigned ballots will retrieve those ballots and report the results to the state.  Results from all counties will be aggregated to determine whether the audit was successful or an additional round of counting is required.  If the audit is not successful after multiple rounds, the Secretary of State's Office may determine that a full hand recount be conducted.

Timeline of the Risk-Limiting Audit

The Risk-Limiting Audit will follow these steps:

Dates and Deadlines for the Risk-Limiting Audit for the November 3, 2026 General Election

The following dates and deadlines will apply to the Risk-Limiting Audit for the November 3, 2026 General Election:

The Secretary of State will certify the completion of the RLA once the audit is concluded, and will publish the results of the RLA on the Secretary of State’s website no later than three days after the audit is completed.

Creating a Ballot Manifest

Each county must create a ballot manifest.  A ballot manifest is a record of how the ballots are stored by identifying the number of ballots in each container.  This process allows a specific ballot to be located and retrieved.  When ballots are randomly selected, they will be identified by the container that they are located in, and the number in which they are sorted in the stack. 

Here is a basic example of a ballot manifest:

Batch Name Number of Ballots
Main Early Voting Location – Box 1 2591
Main Early Voting Location – Box 2 1202
Courthouse (Election Day) 3207
City Library (Election Day) 502

And here is an example of how randomly selected ballots would be identified:

In other words, the batches that are used to select and identify the ballots that are included in the audit will be based on the ballot manifest that the county provides.  A more detailed and specific ballot manifest will make it easier for the county to pull their ballots for the audit because it will provide more specific direction about how to find the ballot that is being used for the audit.

We encourage counties to review our RLA Handbook for additional information on how to properly prepare your ballot manifest and submit it to our office to make the audit process as efficient as possible.

The deadline for a county to submit the ballot manifest for the RLA for the November 3, 2026 general election is Monday, November 16, 2026.

Appointing Audit Teams

The county election officer is responsible for appointing the audit board members to perform the audit.

The audit board will include the individuals who are retrieving ballot boxes, the individuals who are participating in pulling and counting the ballots, and the individuals who manage the reporting process with our office.

The audit board can include one or more audit teams that are assigned to specific ballot boxes or to specific tasks in the audit process.

The Election Code does not provide specific eligibility requirements for the members of the audit board.  We recommend that the audit board be made up of individuals who are either registered voters of the county that would satisfy the eligibility requirements for election judges and clerks, or individuals who are permanent employees of the county that are registered to vote in Texas.  The general custodian is considered a member of the audit board.

To the extent possible, the county election officer should maintain party balance in appointing members of the audit board.  If one or more audit teams are created, then we recommend party balance in designating those team members.

Procedures for Conducting the Audit

Each county will download a set of audit files through the RLA software (Arlo), including a list of the randomly selected ballots from that county to include in the audit.

The county will retrieve the ballot boxes that contain the ballots that are included in the audit and will verify the serial numbers on the ballot box seals against the chain of custody paperwork.  The county will then unlock the ballot box and begin the process to retrieve the audited ballots from their containers.

For example, if the selected ballot is Main Early Voting Location – Box 1, Ballot 521, then the county will open the ballot box labeled “Main Early Voting Location – Box 1”, and will pull the ballots out of the box.  The audit team assigned to that ballot box will count down the stack of ballot until they find the 521st ballot in the stack.  The audit team will then read and report the voter’s selection on the ballot for the race (or races) that are included in the audit through the RLA software (Arlo).

The counting process is different from the traditional tallying process that is used for recounts and other post-election audits.  Those results will not be tracked on a tally sheet, but instead the audit team will indicate the voter’s selection on each selected ballot in the RLA software (Arlo).

For example, if the ballot that was randomly selected is “Main Early Voting Location – Box 1, Ballot 521”, and the voter selected Candidate A for the audited race on that ballot, then the audit board will indicate in the reporting software that Candidate A was selected on “Main Early Voting Location – Box 1, Ballot 521”.

Reporting the Audit Results

As each ballot is counted, the audit board will enter the selection for that ballot into the RLA software (Arlo).  Please keep in mind that these submissions are not raw totals of the number of votes received for each candidate, but should state the candidate that was selected on each individual ballot that was included in the audit.

Aggregating Results and Determining the Success of the Audit

After the audit is completed and the counties finalize their submissions in the reporting software, the results submitted by the counties are aggregated by the state to determine whether the audit was successful.  The audit is considered successful if the results of the audit demonstrate that the outcome of the election was correct within a pre-defined threshold of statistical certainty (the “risk limit”).

If the audit is not successful after the completion of the first round, then the RLA will proceed to a second round.  The second round will follow the same steps as the first round, beginning with the random selection of ballots and ending with the aggregation of the results.

If the audit is not successful after the completion of multiple rounds, then the Secretary of State has discretion to determine whether a full hand recount of the race will be conducted to determine whether the outcome was correct.

Completion of the Audit

Once the audit is determined to be successful, the Secretary of State will certify the completion of the RLA.  The results of the audit will then be posted on the Secretary of State’s website no later than the third day after the audit is completed.

Notice Procedures

The county election officer is required to post notice on the county’s website of the date, hour, and place of the RLA.  The law does not prescribe a specific timeframe for when this notice must be posted, but we recommend doing so at least 48 hours before the start of the RLA.

Poll Watcher Procedures

Each candidate in the election may appoint one watcher for the Risk-Limiting Audit.  When the watcher reports for service, they must deliver a certificate of appointment to the general custodian of election records. 

The certificate must be in writing and must include the printed name of the watcher, the election subject to the audit, and the printed name and signature of the candidate that appointed the watcher.  Form 4-30 (PDF) can be used for this purpose.

Post-Election Hand Count Audit and the Risk-Limiting Audit

The RLA and the Post-Election Hand Count Audit (PEHCA) are two different types of post-election audits that are legally and conceptually distinct.  However, both audits involve counting randomly selected ballots in the election to verify the accuracy of the reported results.

The PEHCA is a jurisdiction-specific audit that is conducted after every election.  The PEHCA is conducted by using randomly selected polling places (and ballot by mail precincts) and counting all of the ballots cast at the selected polling places and precincts.  The PEHCA verifies that the electronic voting systems used to count the ballots in the election in that jurisdiction were working properly and accurately tabulated the results.

The RLA is a statewide audit that is only conducted for elections in which there is a statewide race or measure on the ballot.  Unlike the PEHCA, which is independently conducted by each county (or entity) that held an election, the RLA is a coordinated effort that is conducted by all of the counties across the state on the same day.  The RLA is designed to verify that the outcome of the audited races are correct across the entire state, regardless of whether the ballots were originally hand counted or were counted using electronic voting system equipment.

In an election where an RLA is conducted, the deadlines to conduct the PEHCA are modified as follows:

How the Random Seed is Generated

The Secretary of State’s Office will perform a dice roll to determine the random seed that is used for the RLA.  The dice roll will be open to the public and livestreamed.  Notice will be posted on the Secretary of State’s website at least 48 hours before the dice roll occurs.

Here is an example of how this process works:

Ten ten-sided dice are rolled.  The results of the roll are as follows:

Dice 1 Dice 2 Dice 3 Dice 4 Dice 5 Dice 6 Dice 7 Dice 8 Dice 9 Dice 10
8 3 9 1 4 4 8 7 8 2

Based on this dice roll, the random seed that is entered into the RLA software is 8391448782.  Using this random seed and the ballot manifests provided by the county, the software will randomly select specific ballots from counties across this state. 

The number of ballots that each county will count is not pre-determined.  The randomization process looks at all ballots across the entire state to determine which ones are selected.  Because larger counties typically have more ballots cast in the election, they will also typically have more ballots included in the audit.

There are a few important things to keep in mind about this process:

If you have any questions about this advisory, please contact our office at 1-800-252-8683 or at Elections@sos.texas.gov.

CA:CP