We wanted to test whether generative AI can create a simple timeline to keep track of important election deadlines that election officials can use to build a voter communications calendar. For this task, we used Maryland’s 2026 election calendar as a test case. We downloaded the calendar into a PDF, uploaded the file to Claude, ChatGPT and Gemini, and asked each tool to produce a simple, continuous, horizontal timeline highlighting important dates and deadlines for voter communications and outreach.

Effective prompting is important for tasks like this one. An initial test using Claude, ChatGPT and Gemini uncovered limitations in how the large language models interpreted what a “timeline” meant. Claude performed the strongest and required the fewest iterations, which is why we decided to complete the task with Claude.

After several iterations, we found the best way to get the desired output was to include a template calendar, which we sourced through a basic Google image search. Once the template was included, Claude produced the desired output albeit with oddly formatted text. A follow-up request to fix text formatting resolved the issue, and the final output was visually appealing and conveyed the desired information in a simple and clear manner.

Prompts in Practice

Use Case / Purpose

Produce a simple timeline for internal use by election offices to keep track of important deadlines for voter communications and outreach.

Audience

Election officials

Timing

Ideal for pre-election cycle planning

Sources

Maryland State Board of Elections 2026 Election Calendar

AI Tool Used

Claude Sonnet 4.5

Prompt

The attached pdf file contains the election calendar for the calendar year 2026 in the state of Maryland. Using the attached jpg file as a reference, construct a simple continuous, horizontal timeline visualization for a Maryland election office’s internal use highlighting important dates and deadlines that will require communication outreach with voters. Only include high priority deadlines. Export it as a pdf

Link: MD PDF, Timeline Template

Output

Claude produced a two-page PDF, with one timeline showing deadlines for the upcoming primary elections and another timeline showing deadlines for the subsequent general election. The first timeline Claude produced had inconsistent format, so we asked Claude to resolve this issue. The updated timeline was formatted clearly. Below is a snapshot of both the initial (wonky formatted) timeline for Maryland’s primary elections and the updated clear timeline with corrected text formatting. The initial prompt, follow-ups and outputs can be found here.

AI Use Safety Signal

🟢 Green (Go)

This use case aims to simplify complex election calendars to help election offices build voter communications around important deadlines. While the GenAI tool completed the task accurately and effectively, it took specification and iterations to get right. This use case highlights the importance of prompt design to ensure the intended outcome. The output has been reviewed by a human for accuracy.

(TL;DR: We’re looking for a few jurisdictions to help co-develop and test a new workshop focused on using AI to improve poll worker training materials. If you’re interested in participating, please fill out this brief form.)

The goal is to create a flexible, hands-on workshop that helps election officials use AI tools to develop clearer, more engaging poll worker training materials. This pilot is designed to support that work with structured guidance rooted in adult learning principles and real-world training needs.

What participation involves:

• A facilitated 1.5–2 hour workshop centered on training needs and experiences

• A chance to demonstrate and share how your office approaches developing and refining materials

• Collaborative discussion that helps us understand what works in the field

• An opportunity to shape a national model by sharing your insights, materials, and lessons learned

What we’re looking for:

• Jurisdictions planning to improve or refresh any part of poll worker training

• One or two real materials to work on (slides, job aids, manual excerpts, etc.)

• Willingness to provide brief feedback after the session

Timeline: Pilot workshops will take place December through February, with minimal time commitment outside the session itself.

Interested? Please submit your interest here: https://docs.google.com/forms/d/e/1FAIpQLSfzHsLCGQ5vcnzsIoqPoiHz9A9VdMSsyqko1kSONxgW7YNNsg/viewform?usp=sharing

We’ll follow up to schedule a short introductory conversation and confirm fit.

That report underscored the need for election offices to build continuity and resilience before a crisis. To demonstrate how AI can help, we generated a nearly complete Continuity of Operations Plan (COOP) for the fictional Concord County using two foundational resources:

• The U.S. Election Assistance Commission’s Continuity of Operations Plan Template (2023)

• FEMA’s National Incident Management System (NIMS) Doctrine (2017), which outlines the national standard for command, coordination, and resource management.

If you create a COOP or similar planning product using this approach, we strongly encourage you to share it with your public safety and emergency management partners. Walking through the plan together — even in a short tabletop or discussion-based exercise — can clarify roles, build relationships, and surface improvements long before they’re needed.

Prompts in Practice

Use Case / Purpose

Generate a draft Continuity of Operations Plan (COOP) for Concord County’s election office by merging the EAC’s 2023 COOP template with FEMA’s 2017 NIMS guidance to ensure operational alignment with national emergency management standards.

Audience

Election officials who want to develop or modernize their COOP plan with minimal setup and clear integration into the NIMS framework.

Timing

Ideal for pre-election readiness reviews or annual emergency management updates.

Sources

• U.S. Election Assistance Commission COOP Template

• FEMA National Incident Management System, Third Edition (2017)

AI Tool Used

ChatGPT 5 (Plus Version)

Prompt

Create a Continuity of Operations Plan (COOP) for the Concord County Election Office using the attached EAC COOP Template (2023) and FEMA’s National Incident Management System (2017). The plan should ensure essential election functions, such as ballot processing, vote-by-mail operations, and results reporting, can continue or quickly resume after a disruption. Incorporate NIMS principles like the ICS command structure, resource management, and communications protocols. Write in clear, professional language suitable for official use, with placeholders for Concord County details, and format the output for easy export.

Output

AI Use Safety Signal

🟢 Green (Go)

This prompt combines public-domain templates and federal guidance, producing structured outputs that can significantly reduce administrative drafting time. However, election officials must review, localize, and verify all content with their emergency management partners and legal counsel before adoption, and should avoid uploading any personally identifiable information (PII) or sensitive operational details when using AI tools.

Subscribe now

For this task, we use New Jersey’s Voter Information Portal as a test case. This test case will help evaluate the effectiveness of using AI-generated checklists for auditing websites and demonstrate how election offices can run their own web accessibility audits.

First, we saved the website we want to audit to a local drive as an HTML file (Ctr + S; Save as Website, Complete). Then, we uploaded the HTML file, along with the PDF of the checklist we had previously generated into ChatGPT (free version). We also made sure to include an instruction in the prompt to format the output according to the checklist to make sure the output is as concise and readable as possible.

Prompts in Practice

Use Case / Purpose

Use an AI-generated WCAG web accessibility compliance checklist to audit website content accessibility and demonstrate its effectiveness for election offices

Audience

Election officials who want to assess whether their website meets accessibility requirements

Timing

Ongoing

Sources

• New Jersey Voter Information Portal

• WCAG Checklist

AI Tool Used

ChatGPT (Free Version)

Prompt

The webpage file provided links to New Jersey’s official voter information portal. The Americans With Disabilities Act (ADA) prevents discrimination on the basis of disability and requires that official government resources are accessible to those with disabilities. Please use the attached WCAG checklist to comb through this website to assess accessibility. Format your response as a filled out reproduction of the checklist template.

Output

Below is a snippet of the output.

ChatGPT generated a summary table with three categories for the audit results: ✓ for successful compliance, ✗ to flag for potential issues with compliance, and N/A for fields that were not present on the website.

AI Use Safety Signal

🟡 Caution

This prompt requires some minor technical ability to ensure accuracy. For example, when we uploaded the website as a “Website, HTML Only” file (instead of Website, Complete) ChatGPT identified more accessibility violations than when we uploaded the website file saved as a complete webpage. Therefore, it is especially important to manually verify accuracy of the output. The checklist is meant to streamline the manual auditing process, not replace it.

Web content accessibility creates distinct obstacles for individuals with disabilities. Information access needs vary depending on a person’s disability, but those needs must be addressed to ensure that everyone has equal access to information online.

In 2024, updated federal standards clarified how state and local governments should apply the Americans with Disabilities Act (ADA) to their websites and mobile applications, incorporating specific technical requirements based on the Web Content Accessibility Guidelines (WCAG).

Because the WCAG resource is highly technical and does not include a straightforward checklist, we set out to create one that election offices can use to audit their websites.

Prompts in Practice

Use Case/Purpose

Produce an ADA and WCAG compliance web accessibility checklist that election offices can use to assess website content accessibility.

Source

ADA Guidelines On Website Accessibility

AI Tool Used

Claude Sonnet 4.5

Prompt

I want to produce a web accessibility audit checklist for my elections office using the federal government’s WCAG2 guidance. Review the guidelines in this resource and produce a checklist that my elections office can use to determine whether our election website meets these guidelines. Make sure to include all criteria (A, AA, AAA). Format the checklist to a print-friendly format.

Audience

Election officials who want to assess whether their website meets accessibility requirements

Output

Claude produced a checklist, which we exported as HTML and then saved as a five-page PDF. The generated checklist includes a “how-to” section alongside a list of common features most likely to be missing from websites (like alt text in images). Below is a snippet of what it looks like, and the full PDF output can be accessed here: https://drive.google.com/file/d/1zl2-VDv8QMk9b0SZISm9QfZ5CwtOn9S2/view.

AI Use Safety Signal

🟡 Caution

This use case supports accessibility by using a generative AI tool to summarize technical guidance and create a checklist for evaluating election websites. Outputs should be reviewed for accuracy against official WCAG resources and verified by subject-matter experts. Always consult legal counsel for advice on accessibility compliance. AI-generated materials should be treated as informational aids, not legal determinations.

I wanted to see if I could use artificial intelligence to make a set of training flashcards that would be both a keepsake and a useful tool for poll workers. The training deck from the U.S. Election Assistance Commission and the conversation starter cards from the U.S. Alliance for Election Excellence were both really well executed. I wanted in on the fun!

Because I already had a Gemini Gem loaded with all of my information, I prompted it to use only the poll worker training documents referenced on a specific webpage.

I simply had to ask it to produce 100 questions and answers regarding duties at a polling place. Eventually, I asked for another 40 questions that included info on early voting and same-day voter registration. Eventually, I whittled these down to 52 cards that I liked, and I used Canva and my branding guidelines to produce the PDF proof. Now we are all anxiously awaiting the finished product from the printer.

Prompts in Practice

Use Case/Purpose

Designing a set of poll worker training flashcards, drawn from official training documents, that serve both as a fun keepsake and as a practical tool to reinforce core duties and procedures

Source

Webpages and training materials (PDFs, PowerPoints, Word Documents and video) produced by the Greenwich, Connecticut, Registrars of Voters

AI Tool Used

Google Gemini

Prompt

Review the poll worker training material documents linked from https://www.greenwichct.gov/631/Poll-Worker-Training-Materials. Using these documents only, provide a set of 100 flashcard questions and answers based on poll worker roles. Include general questions such as who can be in a polling place and the hours of voting. Display the questions and answers in a two column table.

AI Use Safety Signal

🟢 Green (Go)

This use case is a productivity task – the AI used inputs which included terminology which was specific to the jurisdiction, and was limited to specific documents produced by the election office. The outputs (questions and answers) were reviewed for accuracy.

I learned during this process that the Gem does not “remember” information between sessions. To work around this, I created a Google Doc that has links to all of the items it should review before answering questions. I can modify those documents and add links when new materials are produced.

I also set up a second Google Doc in which I copied and pasted specific email correspondence with the Connecticut Secretary of the State’s Office that gave definitive interpretations on specific points of election law.

These Google Docs are referenced in the Gem instructions and are the limiting factor.

This process allows me to refresh the Gem quickly without having to re-feed it a hundred documents and all of the URLs that contain the Connecticut General Statutes regarding elections.

The Gem instructions are provided below in the Prompts in Practice section. In addition to the instructions, two documents are linked in the “Knowledge” section of the instructions, which can include up to 10 files.

You can review the shared gem at https://gemini.google.com/gem/10U9StggupCUmnUXLPhGDpaNoCGVq090o?usp=sharing. (You must have a Google account to access this.)

Ask the Gem: “What information do you use to answer my questions?” and you will see the knowledge from which it pulls.

Also ask the Gem: “Can you provide answers with any information beyond your source of truth?”

By feeding only trusted sources, I maintain a level of control that means I can share this tool with staff and be confident that the answers will be trustworthy.

Fred DeCaro III is the Republican Registrar of Voters in Greenwich, Connecticut. He has served in his role since 2009.

Prompts in Practice

Use Case/Purpose

Source

• Webpages and training materials (PDFs, PowerPoints, Word Documents and video) produced by the Greenwich Registrars of Voters.

• Manuals produced by the CT Secretary of the State.

• Title 9 of the Connecticut General Statutes.

• Selected email correspondence from the Connecticut Secretary of the State copied into a Google Doc (including to/from headers).

AI Tool Used

Google Gemini

Prompt

You are a subject matter expert on running elections in Greenwich, Connecticut. You will absorb election related materials and provide training guides and possibly videos. You can construct FAQ documents, and answer general questions from a staff member or a poll worker.

Prior to answering questions, you should immediately refresh your memory by reviewing all of the materials that are linked on this page: https://docs.google.com/document/d/1iN4JQ2De9I5RIc0u_se522NyMwnBjW-cZErRQX-yWoI/edit?tab=t.0. You should also review selected email correspondence with the Connecticut Secretary of the State’s Office linked here: https://docs.google.com/document/d/1VvrK298UA9xRSm0zgmRL0VLsAFXbFL1YSsQJ4sF5Bag/edit?usp=sharing

This includes watching all of the linked videos and extracting knowledge from them.

AI Use Safety Signal

🟢 Green (Go)

This use case is a productivity task – the AI supports research, drafting, or design work while leaving the final judgment to you. Because outputs can be reviewed and verified before use, the risks are low. Always check accuracy, adapt language, and confirm compliance before sharing externally.

A friend and supporter of the AI & Elections Clinic, Microsoft’s Ethan Chumley, spoke alongside Jennifer Morrell, the chief executive officer of The Elections Group.

During the presentation, Chumley reiterated a point he made on our first AI & Elections Clinic webinar about how to prompt AI tools.

He shared that the quality of your prompt determines the quality of the output. Great prompts have four key pieces:

• Goal: What you want the AI to produce

• Context: Relevant background information

• Source: Where the AI tool should look for necessary information (e.g. a document, URL)

• Expectations: Specific requirements for the output (e.g. tone, format, length)

Prompts in Practice

Use Case/Purpose

Creating a summary of existing documentation and state materials pertaining to in-person voting options that can be understood more quickly by voters

Source

AI Recipe Book from The Elections Group (Bot Appétit!)

AI Tool Used

Google Gemini

Prompt

Please produce a clear and concise summary of Arizona’s official early voting procedures, drawing only from the Arizona Secretary of State’s webpage at https://azsos.gov/elections/about-elections/elections-procedures/early-voting-procedures. The goal is to provide an overview that captures the main rules, timelines, and processes for ballot-by-mail, in-person early voting, emergency voting, and special election boards. The audience may not have time to read the full page but still needs accurate and reliable information. Please use a professional, neutral tone, keep the length to around 500 words or less, and format the response with clear headings and bullet points for readability.

Output

AI Use Safety Signal

🟢 Green (Go)

This use case is a productivity task – the AI supports research, drafting, or design work while leaving the final judgment to you. Because outputs can be reviewed and verified before use, the risks are low. Always check accuracy, adapt language, and confirm compliance before sharing externally.

But I don’t think that’s true. Election administration is often about guessing. What will the legislature do next session? What parts of VVSG 2.0 will my state adopt? How much money will I need for an election that is 18 months away? How many ballots should I order? I’m left asking, “Can AI help me guess better?”

During the June 2025 election, I ran a test: How accurately could I predict turnout based on the first day of early voting? Turns out, pretty accurately. Or at least that’s what my AI assistant – Employee #10 as our nine-person office likes to call it – tells me.

Using Employee #10 as a guide, I was able to run a Historical Proportional Forecast based on five primaries worth of early voting data to predict turnout each day of early voting and at each precinct on Election Day.

My predictions got an R-squared of .990001. I have no idea what this means. My degrees are in anthropology, but Employee #10 tells me this was a very good analysis. Better AI-assisted guessing can help improve staffing and equipment allocations.

How I Did It

Virginia has 45 days of in-person no-excuse absentee voting (also known as Early Voting). The law allowing no excuse early voting passed in 2020. When I first asked Employee #10 to help me predict turnout, I needed 50 datasets to build a predictive model. Arlington County predictably has a Democratic Primary every spring, so I had five datasets. I suspect large data modeling will only ever be of limited use to a local election shop because not even Georgia has 50 elections a year.

AI usage restrictions in my jurisdiction prevented me from loading data into Employee #10, but I could ask for help. Through prompts, Employee #10 recommended how to structure the data using Excel and recommended an analysis I could run.

I first looked at trend lines to see if there was consistency between elections. The 2020 primaries stood out – data from those elections, administered during the pandemic, did not follow a pattern. I excluded these elections from the dataset to avoid skewing the predictions.

I then calculated the percentage of the total early voters who voted each day of early voting. I averaged percentages for each day (e.g. on the first day of early voting on average 1.37% of early voters cast ballots). Using actual turnout from the first day, assuming that represented 1.37% of early voters I could expect, I forecasted that 4,308 would vote early. I then forecasted the number of voters might come each day using each day’s average.

On day 3, my model was off by two. On Day 4, it was off by five. Strong start. I next decided to use it to forecast the number of Election Day voters at each precinct using a similar method of calculating averages of percentages. Two weeks into early voting I adjusted my ballot order because we were tracking higher than initial predictions based on the type of election and offices on the ballot. The model was right: Turnout was higher than anyone in the state predicted.

With the election over, certified, and audited, I’ve turned back to Employee #10 to tell me if my model was any good. With its help, it recommended calculating R-squared to measure how well actual turnout was reflected by the model. It provided steps to calculate R-squared in Excel (something I didn’t know Excel could do). Employee #10 helped me interpret this number, told me that I was very strongly able to predict turnout, and it gave me five stars.

I realize this isn’t groundbreaking research. Political scientists and analysts have been running these analyses for years. I suspect there are a few reading this who will politely comment on why this is analysis is bad or how it could be improved.

AI can’t make me an expert, but it can help me know I need to find a temp to cover lunches starting three Fridays out from an election as opposed the two I had planned.

Will this model hold for the fall? I’m not sure, but you can follow the election with me by checking out our dashboard where we publish early voting turnout daily.

P.S. If you’re wondering when the best time to vote early to avoid a line is, I recommend a rainy Wednesday afternoon. If you enjoy standing in line with all of your neighbors, we'll see you the last Friday around 3 p.m.

Note: This post originally appeared on LinkedIn. It has been edited slightly for tone and to update time references.

Prompts in Practice

Use Case/Purpose

Turnout forecasting for staffing, ballot ordering, and equipment allocation during early voting and Election Day.

Source

Election official (Arlington County, Virginia, General Registrar/Director of Elections Gretchen Reinemeyer)

AI Tool Used

ChatGPT

Example Prompt

I want to use Excel to build a simple model that predicts early voting turnout in Virginia. We have 45 days of in-person, no-excuse absentee voting. I only have data from a handful of past elections (about five datasets), and I cannot upload the data here because of restrictions.

1. Tell me how to structure the data in Excel to make it usable for forecasting.

2. Recommend what kind of analysis I can run with a limited dataset.

3. Explain how to identify anomalies (for example, pandemic-era elections that don’t follow normal trends) and how to decide whether to exclude them.

4. Walk me through how to calculate daily turnout percentages and how to use the first day’s turnout to project total early voting turnout.

5. Keep all explanations step-by-step and assume I don’t know advanced statistics, but do know how to enter formulas and use basic charts in Excel.

AI Use Safety Signal

🟡 Yellow (Caution)

This use case supports productivity by assisting with data structuring and forecasting methods. While outputs were highly accurate in this instance, predictions carry real-world consequences (e.g. ballot orders, staffing). Always validate models, confirm assumptions, and review results with subject-matter experts before acting.

However, from recent conversations with election officials, I know that the number of folks who are exploring AI-enabled tools is growing quickly.

To meet this growing curiosity, a number of resources have already been developed to help election officials get started. Microsoft’s Elections & Societal Resilience team developed an AI prompt library. Similarly, The Elections Group has been sharing its AI Recipe Book and accompanying presentation at election administration conferences – complete with chef’s hats, aprons and robot sightings. (Unfortunately for me, my head is usually too big for any Amazon-bought chef’s hat.)

Both of these resources provide an introduction to election officials on how to leverage AI tools for their work. AI has many applications, but here we’re focusing on how to prompt generative AI tools.

Today, I’ll share a simple example, and on Tuesdays through the end of the year, I’ll continue to post more examples.

Prompts in Practice

Use Case/Purpose

Social Media and Media Outreach Campaign Planning: Election Night Results

Source

Prompt Library (Microsoft)

AI Tool Used

ChatGPT 5 (Plus Version)

Prompt

Create a social media campaign and press release for Election Day to be released before the polls close. The campaigns should inform the public about how election results will be reporting on election night in Sun Devil County, emphasizing that initial results are unofficial and the full counting process takes time. Reference that paper and mail in ballots will take longer to process. The tone should be formal, but easy to understand. Include our website where citizens can find official results (sundevilcounty.org).

Output

AI Use Safety Signal

🟢 Green (Go)

This use case is a productivity task – the AI supports research, drafting, or design work while leaving the final judgment to you. Because outputs can be reviewed and verified before use, the risks are low. Always check accuracy, adapt language, and confirm compliance before sharing externally.

Have an AI use case you think is worth exploring? Email me and I’ll add it to the queue.

Likewise, if you’re interested in having a presentation on any of these resources – or on AI in elections generally, send me a note.

Join the AI & Elections Clinic on Friday, September 5, at noon ET for a conversation focused on practical, real-world applications of AI in election administration. This webinar will highlight use cases that election officials are already exploring, from simplifying communications and managing workflows to strengthening resilience. Panelists will share lessons learned, tips for safe and effective implementation, and considerations for building trust when using new technologies.

Hosted by Arizona State University’s Mechanics of Democracy Laboratory, in partnership with The Elections Group, this session is designed for election officials at every stage of exploring AI. Whether you are curious about AI, cautious about its risks, or already testing tools in your office, you’ll gain concrete takeaways and a grounded perspective on how AI can support the work of election officials.

Register