Units
1. Introduction to Artificial Intelligence in Finance and Tax: Roles, Boundaries, Verification and Ethics 2. Tax Legislation Research: Law, Communiqué, Ruling and Judicial Decision 3. Declaration Preparation Support: VAT, Withholding Tax, Provisional and Corporate Tax 4. Tax Calculation and Base: Recalculation Discipline 5. Financial Statement Analysis: Balance Sheet, Income Statement and Ratio Interpretation 6. Tax Compliance Checklists and Process Automation 7. Notification, Objection and Correspondence Drafts: Tax Office and Reconciliation 8. Tax Review and Audit Preparation: Document, Evidence, Defense 9. Transfer Pricing, VAT Refund and Complex Issues 10. Customer Data and Privacy: KVKK, VUK Confidentiality and Secure Tools 11. Managing the Risk of Errors, Hallucinations and Misinterpretations 12. End-to-End Secure Workflow: Quality Control, Expert Approval and Automation
Unit 3 / 12

Declaration Preparation Support: VAT, Withholding Tax, Provisional and Corporate Tax

Gains:

  • Ability to understand the logic and items of declaration types and create a draft checklist with the support of artificial intelligence
  • Ability to use artificial intelligence in pre-declaration consistency and forgotten item control and verify final figures with the official system and legislation
  • Being able to apply the discipline of never turning the artificial intelligence output directly into a declaration, knowing that the legal responsibility of the declaration lies with the member of the profession who signs it.

A declaration is an official document in which a taxpayer informs the state of his taxable situation in a certain period. This is the heart of the professional's job and also the point of highest responsibility: the signed declaration is a legal commitment. In this unit, we will learn how to use artificial intelligence as a safe pre-controller and reminder in the declaration process. The critical principle is clear from the beginning: AI does not prepare the declaration, it only helps to check that you have prepared it. The numbers come from the official system, the verification is yours, the signature is yours.

Main declaration types and logic

Each declaration answers a different question. AI is powerful at quickly explaining this logic and extracting the item list.

  • VAT Declaration: It declares the difference between the VAT (Value Added Tax, indirect tax collected on goods and services) calculated in a period and the VAT to be deducted. Main logic: calculated − deductible = payable or rollover.
  • Withholding and Premium Service Declaration: Declares the withholding tax (tax deducted at source) and SSI premiums deducted by the employer from employee wages and some payments.
  • Provisional Tax Declaration: It is the advance payment of annual tax from period to period; It is calculated based on period earnings and offset in the annual declaration.
  • Corporate Tax Return: Declares the tax paid by a company on its annual earnings; The transition from commercial profit to financial profit (legally disallowed expenses, exceptions, deductions) is made here.
  • Annual Income Tax Declaration: Collects the annual earnings and revenues of the real person.
Attention: Artificial intelligence can explain the logic of these types correctly, but it may give incorrect current line numbers, codes and rates in the declaration form. The form structure and rates are always taken from the current official system (Digital Tax Office / Declaration Editing Program).

The safe role of artificial intelligence in the declaration process

You can safely use artificial intelligence at three points:

  1. Item reminder: "Which exemption and withholding items can be overlooked in the VAT declaration for this taxpayer type?" Creating a checklist such as:
  2. Consistency check: To question the logical consistency between the trial balance and accounts you have and the statement draft ("Is the calculated VAT and sales revenue proportionally compatible?").
  3. Explanation production: Drafting the explanation or explanatory text attached to a declaration.

What he cannot and should not do is produce the final figure and put it in the declaration. The figure comes from accounting records and the official system.

Safe flow step by step

The flow below shows the preparation of an AI-powered but liability-preserving declaration.

step

who does

Contribution of artificial intelligence

Trial balance and document compilation

professional

Missing document checklist

Item classification

professional

Draft question list for KKEG/exception distinction

calculation

Professional + system

Make steps visible, recalculate

Consistency check

professional

Proportional and logical cross-checking

Login to the system

professional

None (official system)

Signature and submission

professional

None (legal liability)

three mini cases

Case 1 — Forgotten withholding deduction. While preparing the VAT declaration of a construction company, a consultant asked the artificial intelligence, "What partial withholding items can there be in this sector?" he asks. The model recalls an overlooked type of withholding. The consultant confirms this from the legislation and corrects a missing withholding tax of 42,000 TL. Here, the advisor is responsible for the artificial intelligence reminder, confirmation and decision.

Case 2 — Error with wrong rate. An intern tells the AI ​​to “apply provisional tax rate”; The model uses an old rate and the draft shows 30,000 TL more tax. The consultant catches the error when he recalculates it with the current rate in the official system. Lesson: the rate is always taken from the current official source, not from the model's memory.

Case 3 — Corporate tax offset control. Before the corporate tax declaration, a CPA has the artificial intelligence write down the items for the transition from commercial profit to financial profit into a checklist (KKEG, exemptions, previous year's loss, provisional tax offset). It matches each item with its own records and notices a forgotten prior year loss offset. The statement is given correctly; AI has only served as memory.

Weak prompt / Strong prompt

Weak prompt:

Prepare my VAT return, find the tax to be paid.

The model makes up a number without accessing the data, without knowing the period and legislation; This is directly dangerous.

Powerful prompt:

Your role: VAT declaration pre-flight assistant.Task: CALCULATE THE DECLARATION based on the following period data; create a checklist instead. Check:- Which items might have been overlooked (exception, withholding, refund)?- Are the calculated and deducted VAT items logically consistent?- Which amounts should I confirm from the official system and legislation?Rule: you decide the final figure; produce only control questions.[SEMESTER SUMMARY (anonymised): ...]

The second prompt places the AI in its correct role (controller); keeps the responsibility in the hands of the professional.

Here is a general checklist prompt you can use before declaring:

Produce final checklist for the following declaration type:- Frequently forgotten items- Common classification errors- Consistency checks (rate, logic)- Current rate/limits to be confirmed (don't give me numbers, say "confirm")Declaration type: [VAT / Withholding / Provisional / Institutions]

To draft the accompanying description text:

Your role: assistant financial advisor.Task: Draft a short explanatory text in formal style to be attached to the declaration, explaining the following situation.Rule: if you add legal basis, mark "must be verified"; the amount is fabricated.[STATUS: ...]

And for consistency cross-check:

Compare the following two data and question whether there is any inconsistency:1) Calculated VAT in the declaration draft: [amount]2) Sales revenue for the period: [amount]Do these two seem proportionally compatible? If incompatible, list possible reasons. Don't give definitive results, generate control questions.

Common mistakes

  • Making the model produce numbers. The final amount comes from the accounting record and the official system; not from the model.
  • Getting the current rate from the model. Rates and limits must be confirmed from the official source.
  • Relying on the form structure model. Line codes and form layout are taken from the current official program.
  • Bypassing the consistency check. Declaration submitted without logical cross-checking carries the risk of error.
  • Pasting data without anonymizing. Taxpayer data should not be given directly to the tool; Confidentiality must be protected.

In summary

Declaration preparation is the most responsible step of the profession, and artificial intelligence is used here as an assistant that controls and reminds, rather than producing numbers. The model explains the logic of declaration types, reminds of forgotten items, generates consistency questions; But the final figure comes from the accounting records and the official system, the rates are confirmed from the current source, and the signature and responsibility remain with the professional. AI output is never directly converted into a declaration.

Application task

Select a declaration type you have prepared (with sample data). Get a pre-declaration checklist and consistency questions from AI with the powerful prompt template. Match each item on the list with your own records and identify at least one "overlooked item" and confirm it with the legislation. Do not use any ratios given by the model directly; Verify each one from the official source and make a note of it.

checklist

  • [ ] I used artificial intelligence in a controlling role, not in producing numbers.
  • [ ] I got the final amounts from the accounting record and the official system.
  • [ ] I have confirmed all rates and limits from the current official source.
  • [ ] I made a checklist for forgotten items and matched each item.
  • [ ] I cross-checked for consistency and logic.
  • [ ] I protected confidentiality by anonymizing the data.
  • [ ] I concluded the statement by confirming that I have the signature and responsibility.
Previous Next