Back to Articles
    Sector-Specific AI

    VITA Sites and Free Tax Prep: Volunteer Training and Appointment Flow With AI

    Nonprofits that run Volunteer Income Tax Assistance (VITA) and Tax Counseling for the Elderly (TCE) sites compress a year of work into a few frantic months. Volunteers have to be recruited and certified every fall, clients have to arrive with the right documents, appointments and walk-ins have to be balanced against a limited number of certified preparers and quality reviewers, and every grant report has to be filed on time. AI can take real weight off the coordinators who hold all of this together, from study aids for new volunteers to reminder messages in a client's own language to queue planning built on last season's numbers. It also brings a hard set of limits, because taxpayer data and tax advice must never pass through a general AI tool. This guide walks through where AI fits in a free tax site, where it does not, and how to set it up before the season starts.

    Published: November 15, 2026•19 min read•Sector-Specific AI
    Volunteers being onboarded and trained, representing VITA free tax preparation site volunteers preparing for the filing season

    Free tax preparation is one of the most concrete economic benefits a community nonprofit can deliver. A family that files through a VITA site keeps the fee a paid preparer would have charged, and often receives refundable credits they did not know they qualified for. The IRS describes VITA as serving people who generally make about $70,244 or less, along with people with disabilities and limited English speakers, while TCE focuses on taxpayers 60 and older. Behind every one of those returns is a site run by a United Way, a community action agency, a library, a credit union, a university, or a coalition of local partners.

    The funding that keeps many of these sites open is competitive and closely watched. In January 2026 the IRS announced $53 million in grants to 315 VITA and 48 TCE grantees, after receiving 479 applications requesting more than $79 million. That gap between requests and awards means most programs supplement IRS matching grants with United Way allocations, bank and credit union sponsorships, and local foundation support, each with its own reporting expectations. Coordinators end up running a seasonal operation, a volunteer program, a compliance program, and a grants portfolio at the same time.

    That workload is exactly where AI can help, and exactly where it can cause harm if used carelessly. A free tax site handles some of the most sensitive data a nonprofit will ever touch: Social Security numbers, wage statements, bank account numbers for direct deposit, and dependents' information. It also operates under detailed IRS rules about who can prepare returns, how they are reviewed, and what kinds of returns are in scope. AI tools have no place in that core work. They do have a useful place around it.

    This article covers how a VITA or TCE season actually runs, the bright lines AI must never cross, and six areas where AI earns its keep: volunteer certification prep, appointment and queue planning, multilingual client communications, volunteer shift scheduling, EITC outreach, and site and grant reporting. It closes with a pre-season plan for putting a simple AI policy in place. Nothing here replaces IRS guidance or your SPEC territory office, and the publication numbers referenced should always be checked against the current versions on IRS.gov.

    How a VITA or TCE Season Really Works

    Before deciding where AI fits, it helps to map the operation. Most free tax programs follow the same arc. Recruitment begins in late summer or early fall. Training and certification run from November into January. Sites open around the start of the IRS filing season in late January, and the busiest weeks cluster in early February, when families expecting the Earned Income Tax Credit are eager to file, and again in the final weeks before the April deadline. Then the season ends, and the reporting begins.

    The volunteer pipeline is governed by IRS certification requirements. Every volunteer, including greeters and screeners who never touch a return, completes Volunteer Standards of Conduct training, described in Publication 4961, and passes the related exam. Most volunteers do this through the IRS Link & Learn Taxes online platform, where the standards of conduct exam must be passed before any other certification test. Preparers then certify at a tax law level, and anyone conducting intake, interviews, or quality review also certifies for that role. Volunteers sign Form 13615, the Volunteer Standards of Conduct Agreement, which records what they are certified to do. Site coordinators complete their own annual training as well.

    At the site, every return follows a defined path. The client completes Form 13614-C, the intake, interview, and quality review sheet, which a certified volunteer uses to interview the client and confirm details. The preparer works in the IRS-provided tax software. A certified quality reviewer checks the return against the client's documents and the intake sheet before it is filed. These steps, along with requirements for reference materials, site identification, security, and volunteer agreements, are spelled out in Publication 5166, the Volunteer Quality Site Requirements.

    Two other boundaries shape daily operations. Scope of service limits, set out in the Scope of Service chart in Publication 4012, the Volunteer Resource Guide, determine which tax situations volunteers may handle at each certification level. And privacy obligations described in Publication 4299 govern how partners, coordinators, and volunteers protect taxpayer information. Sites also choose among filing models: traditional in-person, drop-off (where the client leaves documents and returns later), and virtual models where some steps happen remotely, which require the client's signed consent on Form 14446.

    Pre-season (August to January)

    Where coordinators lose the most hours

    • Recruiting returning and new volunteers
    • Chasing certifications and signed Forms 13615
    • Building the site calendar and appointment slots
    • Printing outreach and document checklists

    In season (late January to April)

    Where bottlenecks hurt clients most

    • No-shows and clients missing key documents
    • Walk-in lines that exceed reviewer capacity
    • Volunteer burnout and last-minute shift gaps
    • Weekly counts owed to funders and coalitions

    The Bright Lines: What AI Must Never Touch

    Start with the limits, because they shape every other decision. A free tax site is a trust operation. Clients hand over documents they would not show most of their relatives, on the promise that the site follows IRS privacy rules and that the people helping them are certified. A general-purpose chatbot is not part of that promise. Whatever a vendor says about data retention, a consumer AI tool is not an approved place for taxpayer information, and putting it there exposes clients to identity theft risk and your organization to the loss of its sponsor relationship.

    The second limit is about advice. Volunteers are certified to apply tax law within a defined scope, and quality review exists so a second certified person checks the work. An AI tool that answers a client's question about whether a dependent qualifies, or tells a volunteer how to handle a situation, sits outside that structure. It can also be confidently wrong. Even the IRS's own privacy documentation for a VITA generative AI pilot focuses on helping volunteers search public IRS reference material, not on making tax determinations or handling return data. That is a useful signal for how cautious nonprofit sites should be.

    Non-negotiable rules for every site

    Put these in writing before the first volunteer logs in

    • No taxpayer data in AI tools. No SSNs, ITINs, names tied to tax details, W-2s, 1099s, intake sheets, photos of documents, bank numbers, or return data. Not even "anonymized" screenshots.
    • No tax answers for clients. AI does not tell a client whether they qualify for a credit, how to file, or what their refund will be. Those conversations happen with certified volunteers.
    • No return preparation. Returns are prepared in the IRS-provided software by certified volunteers and quality reviewed by certified reviewers. AI is not a preparer or a reviewer.
    • No scope decisions. Whether a return is in or out of scope is decided by certified volunteers and the site coordinator using Publication 4012, not by an AI summary of it.
    • No shortcuts on certification. AI can help someone study. It never takes a certification test for them or substitutes for the required training.

    A simple test helps volunteers apply these rules on a busy Saturday: if the text you are about to paste could identify a client or reveal anything about their finances, it does not go into an AI tool. If the question you are about to ask is about a specific client's taxes, it goes to the site coordinator or a certified reviewer instead. Everything in the rest of this article respects that test. The AI work happens on generic material, aggregate counts, and communications templates, never on a person's return. For a broader framework on handling sensitive information with AI, see our guide to data privacy and security in AI tools.

    Helping Volunteers Prepare for Certification

    Certification is the first place many programs lose volunteers. People sign up in September full of enthusiasm, open Link & Learn Taxes in November, and stall somewhere between the standards of conduct exam and the tax law test. Retirees who did their own taxes for decades may find the format unfamiliar. College students in accounting programs may know the concepts but not the VITA-specific procedures. First-time volunteers who want to be greeters or screeners may not realize which certifications their role requires.

    AI can serve as a patient study partner, as long as it works from official materials and stays in the role of tutor. The most reliable approach is to have the AI generate practice material from the scenarios and topics covered in the IRS training publications, then have volunteers check their answers against the official resources rather than against the AI's explanation. A volunteer might ask for ten practice questions about the steps on the intake sheet, or for a plain-language explanation of what quality review is meant to catch, then confirm the details in the relevant IRS publication.

    Good study uses

    Built on generic, fictional scenarios

    • Flashcards on Volunteer Standards of Conduct and privacy expectations
    • Role-play of the intake interview using made-up households
    • Plain-language summaries of a training chapter the volunteer has already read
    • A personal study schedule that fits around work and family
    • Translated explanations of training concepts for multilingual volunteers

    Uses to avoid

    These undermine certification itself

    • Pasting live certification test questions into a chatbot
    • Treating AI explanations of tax law as authoritative
    • Relying on AI for the current year's thresholds or amounts
    • Using real client situations, even from memory, as practice prompts

    The point about the current year matters. Tax figures change annually, and AI models are trained on data that may be a year or more out of date. A model may cheerfully quote last year's credit amount or an expired rule. That is one reason AI should support understanding of process and concepts rather than serve as a tax reference. The Volunteer Resource Guide and current-year training materials remain the authority.

    Coordinators can also use AI on their side of the certification process. A coordinator might draft a friendly week-by-week certification reminder sequence, write a one-page "which certifications does my role need" explainer for new volunteers, or turn a long training-day agenda into a clear checklist. These are communications tasks, not tax tasks, and they free time for the personal follow-up calls that actually get stalled volunteers across the finish line. Our article on AI-assisted volunteer onboarding covers more patterns that translate directly to a tax site.

    Planning Appointment Flow and Walk-In Queues

    The season's hardest operational problem is matching client demand to certified capacity. Preparers are only half the equation. Every return also needs a certified quality reviewer, and many sites find that reviewers, not preparers, are the true bottleneck on peak days. Add walk-ins who arrive before the doors open, appointment clients who arrive without a key document, and drop-off returns waiting for a preparer, and a site can quickly fall behind.

    AI is genuinely useful here because the inputs are aggregate and non-identifying. Last season's counts by day and hour, average time per return by complexity, number of no-shows, number of clients turned away, and number of volunteers on each shift are operational statistics, not taxpayer data. Export them as totals from your scheduling system or tally sheets, strip any names, and an AI tool can help you spot patterns and plan. It might show that Tuesday evenings consistently had more appointment capacity than demand while Saturday mornings turned people away, or that the second week of February needed twice the reviewers you scheduled.

    A queue planning routine using last season's totals

    Run it in October, then revisit weekly in season

    • 1. Export aggregate counts only. Returns filed per day and hour, appointment versus walk-in, no-shows, turn-aways, and volunteers present by role. No names, phone numbers, or return details.
    • 2. Ask for patterns, not predictions. Have the AI identify peak periods, underused shifts, and the ratio of preparers to reviewers on days that ran smoothly versus days that backed up.
    • 3. Draft a capacity plan. Translate the patterns into appointment slots, walk-in windows, and staffing targets for each role. The coordinator adjusts based on what they know about this year's volunteers and partners.
    • 4. Plan for the document problem. If a meaningful share of last year's clients had to come back because of missing paperwork, put more effort into the reminder and checklist work described in the next section.
    • 5. Review weekly. Update the totals each week in season and ask what changed. Shift resources between sites or models while there is still time to help.

    Filing models deserve a place in this planning. Drop-off and virtual models can smooth peaks by letting preparers and reviewers work through returns during quieter hours, but they come with consent requirements, document handling procedures, and security obligations that in-person models do not. AI can help a coordinator compare the staffing implications of each model, draft the client-facing explanation of how drop-off works, or outline the steps in a virtual workflow. Decisions about whether to adopt a model, and how to meet the Quality Site Requirements while doing so, belong with the coordinator, the partner organization, and your IRS relationship manager.

    Walk-in management benefits from clear, consistent communication more than clever technology. A coordinator can use AI to draft signage explaining how the walk-in line works, what time the site stops accepting new walk-ins, and which documents to have ready while waiting. Greeters can be given a short script, translated into the languages the site serves, so every client hears the same information. These small things reduce frustration at the door, which is where many clients form their impression of the program. For general scheduling approaches, our article on AI-powered program scheduling offers useful frameworks.

    Reminder Messages and What-to-Bring Checklists in Every Language

    The most avoidable failure at a tax site is a client who waits an hour only to learn they are missing a Social Security card for a dependent or a form from a second job. The IRS publishes a checklist of what to bring to a VITA or TCE site, including photo identification, Social Security cards or ITIN documentation for everyone on the return, wage and income statements, and a copy of last year's return if available. Most sites add their own items, such as bank routing and account numbers for direct deposit or documents related to health coverage.

    AI can turn that official list into communications clients actually read. A coordinator can take the IRS checklist plus the site's local additions and ask an AI tool to produce a short version for a text message, a fuller version for an email, a large-print version for TCE clients, and a version organized by life situation ("if you worked more than one job," "if you have children," "if you received unemployment"). None of this requires client information. The AI is working on your template, and your system sends it.

    Multilingual materials

    Translate templates, then verify with people

    VITA explicitly serves limited English speakers, and many sites serve clients in Spanish, Vietnamese, Chinese, Haitian Creole, Arabic, and other languages. AI translation makes it practical to produce checklists and reminders in all of them.

    • Have a fluent volunteer review every translated template before use
    • Keep official form names in English alongside the translation
    • Use IRS translated materials where they already exist

    Reminder sequences

    Fewer no-shows, fewer return visits

    • At booking: confirmation, address, parking, and full checklist
    • One week out: "start gathering these documents now"
    • Day before: short reminder with the top items people forget
    • After a missed visit: a warm rebooking message

    Translation quality deserves real attention. Tax vocabulary is specialized, and a casual translation can turn "dependent" or "direct deposit" into something confusing or misleading. Build a small glossary of key terms in each language with help from bilingual volunteers, give it to the AI along with the template, and have a fluent reviewer approve the result once. After that, the approved versions become reusable assets for every season. Our guide to translation quality review for AI-assisted content describes a practical review process.

    If you send reminders by text message, make sure your process respects consent and opt-out rules. Collect permission to text at booking, honor STOP requests, and keep messages informational. The rules that apply to nonprofit texting are covered in our article on text messaging and the TCPA. Also keep reminders free of anything sensitive. A message should confirm a time and list documents, never mention a refund amount, a credit, or anything from a client's prior return.

    A note on client-facing chatbots: some sites are tempted to put an AI assistant on their website to answer questions. If you do, restrict it tightly to logistics, such as hours, locations, how to book, what to bring, and how drop-off works, and have it route anything about a person's taxes to a phone number or appointment. A bot that tells someone they qualify for a credit, or that a situation is fine to bring to the site when it is actually out of scope, creates exactly the kind of harm the program exists to prevent.

    Scheduling Volunteer Shifts by Certification and Role

    Tax site scheduling is harder than most volunteer scheduling because roles are not interchangeable. A greeter cannot fill in for a preparer, and a preparer certified at the basic level cannot prepare a return that requires a higher certification. Every shift needs the right mix of screeners, preparers at the right certification levels, and quality reviewers, with a coordinator or designated lead present. Many sites manage this in a shared spreadsheet that grows more tangled every week.

    AI can help a coordinator reason through this puzzle using information that is about volunteers, not taxpayers. Given a list of volunteers by first name or ID, their roles and certification levels, and their availability, an AI tool can propose a draft schedule that meets minimum coverage for each role, flag shifts with no quality reviewer, and suggest where a returning volunteer might be paired with a new one. The coordinator still reviews and adjusts, since they know who works well together, who is fading, and who just mentioned a family emergency. Keep in mind that volunteer information is also personal data. Share only what the scheduling task needs, and use an approved tool.

    Coverage checks

    Flag every shift that lacks a certified quality reviewer, a coordinator or lead, or a preparer at the certification level needed for expected returns.

    Peak alignment

    Match staffing to the demand patterns from your queue analysis, adding reviewers on the days that backed up last season.

    Burnout watch

    Spot volunteers scheduled for many consecutive shifts and spread the load before your most experienced people quit in March.

    Retention across seasons is where the biggest savings lie. Every returning volunteer is one fewer person to recruit and certify from scratch, and experienced reviewers are especially hard to replace. AI can help draft thank-you notes that mention a volunteer's specific contribution, a short end-of-season survey, and a summer check-in message that invites people back before recruitment opens. Analyzing survey responses in aggregate can surface the operational frustrations that drive people away, such as chaotic first hours, unclear escalation paths, or too few breaks. Our article on predicting volunteer attrition explores how to spot early warning signs.

    Writing EITC and CTC Outreach That Reaches the Right Families

    Outreach is one of the clearest wins for AI at a free tax program. The IRS estimates that about one in five EITC-eligible taxpayers do not claim the credit, and the average EITC received for tax year 2024 was $2,916. Those unclaimed dollars represent families who could be helped by a free tax site if they knew it existed and believed it was for them. Common barriers include not knowing the credit exists, assuming income is too low to file, assuming the credit only applies to families with children, and distrust of anything that asks for personal documents.

    AI makes it practical to produce many variations of outreach for different audiences and channels. A coordinator can draft a flyer for a laundromat, a short script for a community radio spot, a social post for a school's parent group, a message for a workforce development partner to share with recent graduates, and a version for a faith community's bulletin, all from one approved set of facts. The key discipline is that the facts come from the IRS and from your program, not from the model. Write down the current eligibility messages you want to communicate, sourced from IRS outreach materials, and ask the AI to vary tone, length, and format while keeping those facts unchanged.

    An outreach workflow that keeps facts accurate

    Variety in voice, consistency in substance

    • Build a fact sheet first. Pull current-year talking points from IRS EITC outreach materials and your site details: locations, hours, languages, models offered, and how to book.
    • Generate variations from the sheet only. Instruct the AI not to add any figures, thresholds, or eligibility rules that are not on the sheet.
    • Check every number. A staff member compares every dollar amount and income figure in the draft to the fact sheet before anything is printed or posted.
    • Invite, do not promise. Outreach says families "may qualify" and should come in to find out. It never tells a reader they will receive a credit.
    • Time it to the calendar. Concentrate outreach around EITC Awareness Day in late January and again before the April deadline.

    Trust is the outreach problem AI cannot solve alone. Many families have been burned by paid preparers charging high fees or by scams that impersonate tax help. Messages that emphasize IRS-certified volunteers, free service, and a familiar local partner tend to land better than messages about refund size. AI can help draft those messages, but the credibility comes from the messenger. Distribute outreach through trusted partners, such as schools, clinics, food pantries, and congregations, rather than relying solely on your own channels.

    Site Coordinator Reports and Grant Reporting Narratives

    When the season ends, the reporting begins. IRS grant recipients have reporting obligations, United Way partners want outcomes for their community impact dashboards, corporate sponsors want a story for their own reports, and boards want to know whether the program is growing. Coordinators who have just worked sixty-hour weeks are rarely eager to write five versions of the same season summary.

    This is a natural task for AI, because the inputs are aggregate results and your own observations. Returns filed, refunds and credits returned to the community in total, volunteer hours, number of certified volunteers, sites and models operated, languages served, and lessons learned are all program metrics. An AI tool can turn a page of numbers and bullet points into a clear narrative, then adapt that narrative for each funder's format and word limit. It can also draft the board summary and the volunteer thank-you report from the same source material.

    What to feed the AI

    • Aggregate totals from your official reports
    • Season-over-season comparisons you have calculated
    • Coordinator notes on challenges and fixes
    • The funder's prompts, format, and word limits

    What to keep out

    • Any client-level data, even "just to calculate averages"
    • Client stories with identifying details, unless consented and de-identified
    • Raw exports from the tax software
    • Any number you have not verified against the source report

    Accuracy in grant reporting is not optional. If an AI-drafted narrative rounds a number, inflates a percentage, or invents a comparison, the error carries your organization's name into a federal grant file. Always calculate figures yourself from official reports, give the AI only the verified numbers, and check that every number in the final narrative matches. Ask the AI to list every figure it used at the end of the draft so the check takes minutes. For a deeper look at reporting workflows, see our guide to grant reporting and funder communication with AI.

    Strong reporting also feeds the next grant application. A coordinator who keeps a running document through the season, noting what worked, what broke, and what changed, gives the AI rich material for the narrative sections of next year's IRS application or United Way proposal. The program story becomes a cumulative asset instead of something rebuilt from memory each spring.

    Writing a Site AI Policy Before the Season Starts

    Volunteers will use AI whether or not you have a policy. Some will open a chatbot on their phone to look up a tax term during a slow moment, and some will be tempted to snap a photo of a confusing form and ask what it means. The safest response is not a blanket ban that people quietly ignore, but a short, specific policy that tells volunteers exactly what is allowed, what is prohibited, and whom to ask instead. Deliver it during training alongside the Volunteer Standards of Conduct and Publication 4299 privacy material, so it reads as part of the same professional standard.

    Keep the policy to one page. Coordinators and partner leadership should agree on which tools staff may use for communications and planning work, ideally an organizational account rather than personal consumer accounts. Volunteers who prepare and review returns generally do not need AI tools at all during their shifts, and many sites will simply say so. Our article on creating an AI policy for volunteers provides a template you can adapt.

    What a one-page tax site AI policy should say

    • Approved uses: outreach drafts, checklist and reminder templates, translation of templates with human review, scheduling from volunteer availability, analysis of aggregate counts, and report narratives from verified totals.
    • Prohibited uses: any taxpayer information, photos of documents, tax questions about a specific client, scope decisions, and anything involving certification tests.
    • Where to go instead: the Volunteer Resource Guide, the quality reviewer, and the site coordinator, in that order.
    • Approved tools: which accounts staff may use, and a reminder that personal chatbot accounts are not approved for site work.
    • Reporting: how to report an accidental disclosure immediately, following your existing Publication 4299 procedures.

    Timing matters because the season leaves no room for setup. A practical sequence starts in September: draft the policy and choose tools while recruitment begins. In October, run the queue analysis on last season's totals and draft the reminder sequences and checklists. In November, translate and review the client templates, finalize outreach materials, and build the certification reminder sequence for volunteers. In December, introduce the policy during training and test the reminder messages. By January, everything should be ready to run, so coordinators spend the season with clients and volunteers rather than drafting messages at midnight.

    A note for partner organizations

    Many VITA sites are hosted by organizations whose main mission is something else, such as a library, a housing agency, or a college. If your organization's general AI policy allows staff to use AI with client records in other programs, make sure the tax site is carved out explicitly. Taxpayer information is governed by the IRS volunteer program's privacy requirements, and staff who move between programs need to know that the rules at the tax table are stricter. Coordinate with whoever owns your organization's broader AI acceptable use policy so the two documents agree.

    Conclusion

    Free tax preparation programs do an enormous amount of good with very little slack. A small team of coordinators recruits and certifies volunteers, runs a high-pressure seasonal operation under detailed IRS requirements, reaches families who often do not know they are owed credits, and then reports to multiple funders. AI will not change the core of that work. Certified volunteers will still interview clients, prepare returns in IRS-provided software, and quality review every return, and the site coordinator will still make the calls on scope.

    What AI can change is the surrounding workload. Study aids help new volunteers reach certification. Analysis of last season's totals points to better appointment and staffing plans. Template-based checklists and reminders, translated and reviewed once, reduce no-shows and return visits in every language the site serves. Outreach drafted from a verified fact sheet reaches more families. Report narratives built from verified totals take hours instead of days. Each of these works on generic templates, volunteer schedules, or aggregate numbers, never on a client's return.

    The programs that benefit most will be the ones that set clear rules before the season begins: no taxpayer data in AI tools, no tax answers from AI, no shortcuts on certification. With those lines drawn, coordinators can use AI to win back the time they need for the people in front of them, which is where a free tax site has always made its difference.

    Get Your Tax Site Ready for the Season

    We help nonprofits that run free tax programs set up safe AI policies, multilingual client communications, volunteer scheduling workflows, and reporting templates before the filing season begins.