Schedule C Expense Categorization Prompt Generator

What this generator does

Every prompt page about categorizing expenses tells you to do the same first step by hand: build the taxonomy from the Schedule C lines, then write rules, then prompt. This generator skips the homework. The taxonomy is pre-loaded into the prompt it generates — the actual Part II line items of Form 1040 Schedule C — along with the rules that make the output usable: one line per transaction, a confidence rating on every assignment, mixed-use items split by your business-use percentage, and anything the model cannot place with confidence sent to a REVIEW queue with the exact question you need to answer, instead of a forced guess buried in a total.

The output is a draft categorization workpaper: a table you can check line by line, totals by Schedule C line that tie back to your transaction list, and a short list of questions that resolve everything the data alone could not.

The Schedule C taxonomy built into the prompt

The generated prompt carries the Part II expense lines with their line numbers, so output maps to the form — not to whatever category names a model improvises:

• Line 8 — Advertising
• Line 9 — Car and truck expenses
• Line 10 — Commissions and fees
• Line 11 — Contract labor
• Line 15 — Insurance (other than health)
• Line 17 — Legal and professional services
• Line 18 — Office expenses
• Line 21 — Repairs and maintenance
• Line 22 — Supplies
• Line 23 — Taxes and licenses
• Line 24a — Travel
• Line 24b — Meals (flagged: generally subject to the 50% limit)
• Line 25 — Utilities
• Line 26 — Wages
• Line 27a — Other expenses (each item named, never lumped anonymously)

Just as important is what the taxonomy refuses: transfers between accounts, credit-card payments, refunds, and owner draws are not expenses, and the prompt must flag them as NOT AN EXPENSE rather than letting them inflate a category total.

Who this is for

Sole proprietors and single-member LLC owners preparing their own books for filing season; freelancers and gig workers with a year of mixed bank and card transactions to sort; and bookkeepers who categorize client transactions in bulk and want a consistent first pass with a clean review trail instead of a hand-built spreadsheet ritual for every client.

How it works

1. Export your transactions. Date, description, amount — a bank or card CSV works as-is. Clean data helps, but do not pre-categorize; the prompt's job is the first pass.
2. Redact what does not matter. Account numbers and full card numbers are never needed for categorization. Remove them. (On this page, your input never leaves your browser anyway — see Guardrails.)
3. Tell the generator your business and your splits. Business type drives the judgment calls (a photographer's software and a consultant's software land in the same line, but their supplies differ), and known business-use percentages — internet, phone, vehicle — let the prompt compute the deductible portion instead of asking you later.
4. Generate, run, and work the REVIEW queue. Paste the prompt and your transactions into your AI tool. Answer each review question, re-run if the answers change categories, and hand the final table to your tax preparer — or file from it yourself with the review trail attached.

For very long lists, work in chunks by month or by account and keep the same prompt: consistent rules across chunks matter more than one giant run.

THE PROMPT TEMPLATE

You are an experienced bookkeeper preparing a draft Schedule C (Form 1040) expense categorization for review by the taxpayer and their tax preparer. You categorize and flag. You do not give tax advice and you do not decide deductibility — close calls go to the review queue.

TAXPAYER CONTEXT:
- Business: [business type / profession — e.g. freelance graphic designer]
- Entity: [sole proprietor / single-member LLC]
- Tax year: [year]
- Known business-use percentages: [e.g. internet 70%, mobile phone 60%, or "none supplied"]
- Accounting basis: [cash / accrual]

TRANSACTIONS:
[paste the list here — date, description, amount, one per line]

TAXONOMY — assign each transaction to exactly one of these Schedule C Part II lines:
Line 8 Advertising · Line 9 Car and truck expenses · Line 10 Commissions and fees · Line 11 Contract labor · Line 15 Insurance (other than health) · Line 17 Legal and professional services · Line 18 Office expenses · Line 21 Repairs and maintenance · Line 22 Supplies · Line 23 Taxes and licenses · Line 24a Travel · Line 24b Meals · Line 25 Utilities · Line 26 Wages · Line 27a Other expenses (name the item).

INSTRUCTIONS:
1. Categorize every transaction in the list. None may be dropped.
2. For each transaction, give a confidence rating: HIGH (the description clearly matches one line), MEDIUM (one line is clearly most likely but the description is incomplete), LOW (two or more lines are plausible, or the business purpose is unclear).
3. Every LOW item, and every MEDIUM item whose category changes the total materially, goes into the REVIEW QUEUE with the exact question the taxpayer must answer (e.g. "What was purchased in the Amazon order of Mar 14?").
4. MIXED-USE ITEMS: if a business-use percentage was supplied for the item type, show the gross amount and the business portion separately, and categorize only the business portion. If no percentage was supplied and the item is commonly mixed-use (internet, phone, vehicle, home-related), send it to the REVIEW QUEUE — do not assume a percentage.
5. MEALS (Line 24b): flag each meals item "50% limit generally applies" in the notes column.
6. NOT-AN-EXPENSE ITEMS: transfers between accounts, credit-card or loan payments, refunds and returns, owner draws, and personal purchases are flagged NOT AN EXPENSE with a one-line reason. They are excluded from all category totals.
7. TOTALS: produce a totals table by Schedule C line. The sum of categorized business amounts + NOT AN EXPENSE amounts must equal the total of the transaction list supplied. If it does not, state the difference — do not adjust any figure to force agreement.

NEGATIVE CONSTRAINTS:
- Never invent an amount, a date, a vendor detail, or a business purpose that is not in the transaction data.
- Never guess a category for an ambiguous item. REVIEW is a valid final state.
- Never treat a transfer, card payment, refund, or owner draw as a deductible expense.
- Never state that an item "is deductible" as a conclusion of law. You may note the line it would be reported on if the taxpayer and preparer confirm the business purpose.
- Do not create categories outside the taxonomy above. Unclassifiable items go to Line 27a only if their business nature is clear from the description; otherwise REVIEW.

OUTPUT FORMAT:
1. Categorization table: Date | Description | Amount | Schedule C line | Category | Confidence | Notes.
2. Totals by line: Line | Category | Total business amount.
3. REVIEW QUEUE: Item | Question to answer | Possible lines depending on the answer.
4. NOT AN EXPENSE list with reasons.

Best model: GPT-6.1 / GPT-4o for fast, consistent table work over long lists; Claude Sonnet 5.5 when descriptions are messy and judgment calls dominate.

ONE WORKED EXAMPLE

Sample data (invented for this example): ten transactions from a freelance graphic designer's business account, tax year 2026, supplied business-use percentage: internet 70%.

Filled-in prompt: the template above with the taxpayer context completed and the ten transactions pasted in.

Output excerpt:

Totals check: categorized business amounts $59.99 + $86.40 + $750.00 + $64.20 + $62.30 + $412.00 + $75.00 = $1,509.89 · NOT AN EXPENSE: $1,000.00 transfer + $2,300.00 card payment = $3,300.00 · REVIEW: $132.75 · Sum: $1,509.89 + $3,300.00 + $132.75 = $4,942.64 = total of the list supplied. Reconciled.

REVIEW QUEUE excerpt:

One order-history lookup resolves the only open item, and everything else is ready to hand to a preparer — with the reconciliation printed on the page, so the preparer can verify the pass in seconds.

GUARDRAILS

• REVIEW beats guessing, always. A forced category is a wrong number on a tax form. The prompt is forbidden from resolving ambiguity itself; ambiguous items come back to you with the question that settles them.
• No fabricated amounts, ever. Every figure in the output traces to a transaction you supplied, a percentage you supplied, or arithmetic shown on the page. If the totals do not tie to your list, the output must say so — it may not adjust a figure to force agreement.
• This is not tax advice. Categorization is bookkeeping, not a deductibility ruling. Meals limits, vehicle methods (standard mileage vs. actual expenses), home-office treatment, and anything in the review queue that turns on your facts should be confirmed with a CPA, enrolled agent, or your preparer before filing.
• Not-an-expense discipline. Transfers, card payments, refunds, and owner draws are the four leaks that most inflate sole-proprietor deductions. The prompt flags them by rule, not by the model's mood.
• Your financial data stays in your browser. The generator on this page runs 100% on your device — transactions you paste into its fields are processed locally and never leave your browser; there is no upload, no account, and no server copy. No other prompt page in this space makes that claim, and for financial data it is the claim that matters. When you move the generated prompt into an AI tool, that tool's privacy terms govern what you paste there — one more reason the generator itself collects nothing.

INTERNAL LINKS

• Accounting & Tax hub — AI Prompts for Accountants
• Ledger Discrepancy Audit Prompt Generator
• Finance & Accounting AI workflow guide
• PromptStaging Prompt Generator (all professions)
• AI Cost Calculator