Gains:
- Being able to understand the classification logic (general interpretation rules) by using artificial intelligence for feature extraction and candidate code suggestion when determining the HS/HS Code of a product.
- Ability to implement the obligation to verify the tariff position suggested by artificial intelligence with the official tariff schedule, binding tariff information and customs consultant approval
- Ability to understand the legal-financial risk posed by wrong GTIP in terms of taxes, quotas and bans, and to adopt the discipline of never using the artificial intelligence output as the final declaration.
One of the most critical and most mistaken points in foreign trade is placing the product in the correct tariff code. Because this single code determines the customs duty to be paid, value added tax, additional financial obligations, surveillance and quota practices, and even whether the import of the product is prohibited. Incorrect code means either paying unjustly too much tax or facing penalties, additional accruals and legal investigations due to under-declaration. In this unit, we will cover how to use artificial intelligence as a classification assistant and what we can never leave to it.
Let's start with the terms. HS code (Harmonized System): It is a six-digit code developed by the World Customs Organization that classifies all products with a common international system; The first six digits are the same around the world. GTIP (Customs Tariff Statistics Position): It is the code (usually twelve digits) used in Türkiye, consisting of digits added at the national level to the first six digits of HS. Tariff schedule: a list of all codes and their corresponding tax rates. Binding Tariff Information (BTB): It is the official and binding code information given by the customs administration for a specific product upon the application of a company; It is the safest way of confirmation. General Interpretation Rules (GYK): These are the logical rules of the tariff system that determine which position a product will fall into.
How classification works: understanding the logic
HS/GTIP classification is not random; It is based on a hierarchy and logic. The code places the product into increasingly narrow groups: chapter (first two digits, broad product family—e.g. 84 "machines"), heading (four digits), subheading (six digits), and national subsections. In order for a product to fit into the right place, it is necessary to clarify a few questions: what does the product do (function), what is it made of (material), in what form (raw/processed/set), does it come alone or as a set/accessory.
The General Rules of Comment guide these decisions. For example, one rule says that a product should be classified according to the material or component that "gives it its essential character"; Another wants the most specific definition to be preferred for products that can fall into more than one position. This is where artificial intelligence is valuable: it can structure the product's features and extract possible candidates and their justifications, such as "this product can be in this position because of this, but it can also be in that position because of this feature." This is a preliminary analysis that makes the human expert's job easier; It's not the final decision.
Attention: AI can fluently give you an "exact" GTIP code and tax rate. Writing this code and rate in the declaration without the official tariff schedule, Binding Tariff Information when necessary and the approval of the customs consultant means submitting an unaudited financial declaration. Can make up model code from memory; The rate may be out of date.
Step by step: AI-assisted classification research
Step 1 — Describe the product fully. Classification starts from knowing the product correctly. Function, material, technical specifications, area of use, packaging and set condition. Incomplete recipe leads to incorrect code. Give the AI the product with these dimensions.
Step 2 — Generate candidate positions. Ask the AI for 2-4 possible positions and “why this position, which feature points in this direction” justification for each. Not the exact code, but the candidate and justification.
Step 3 — Identify distinctive questions. Which product feature determines the difference between candidates? AI can list these distinctive questions; You answer these questions with the real features of the product.
Step 4 — Confirm with official source. Search for the candidate code in the official tariff search system, read the disclosure notes, verify the tax rate from the official schedule.
Step 5 — Expert and BTB if necessary. The final decision is made by the customs consultant. For hesitant, high volume or recurring products, applying for Binding Tariff Information is the safest way.
Candidate codes rubric
Evaluation question
Why is it important?
Contribution of AI
final confirmation
What is the main function of the product?
Determines the chapter/position
Clarifies function
expert
What is the main material/component?
According to the comment rule
Justifies candidates
Expert + note
Set/set or single?
Changes classification
separates the possibilities
expert
What is the tax rate?
Determines the cost
Candidate rate (to be verified)
Official tariff
Is there surveillance/quota/ban?
legal risk
Control head pops up
Official notification + expert
Is preferential tariff possible?
Depends on origin
Lists questions
Agreement text + expert
Four copyable templates
1) Product feature extraction:
Your role: customs tariff research assistant. Create the FEATURE LIST that will be the basis for classification for the following product: main function, main material, technical specifications, area of use, single or set, processing status. If there is missing information, write "MUST ASK".Product: [product description]
2) Candidate position and justification:
Your role: tariff classification assistant (research only).Suggest 2-4 POSSIBLE HSCP positions for the following product. For each:- why this position (which feature it points to),- which general rule of interpretation might be relevant,- the counter-argument that leads away from this position. GIVING the exact code and tax rate; Mark "to be verified from official tariff".Product features: [features]
3) Distinguishing questions:
To decide between candidate positions, list distinctive questions about the product that I MUST ANSWER. For example: "does the device work on its own or is it part of a machine?" Briefly state which position the answer to each question leads to. Candidates: [candidate codes]
4) Control/risk headings:
The headings that need to be checked after classification for this product and the target country appear as a CHECKLIST: surveillance application, quota, import permit / certificate of conformity, possibility of additional financial liability. These are reminders; Each title will be confirmed by official notification and counsel.Product: [...] | Country: [...]
Weak prompt / Strong prompt
Weak prompt:
What is the GTIP code and tax of the Bluetooth headset?
This prompt invites only one “definitive” answer; The AI concocts a fluid but unverified code and rate. Since the features of the product (is it wireless, does it have a microphone, with which device is it used) are not asked, the classification logic is also skipped.
Powerful prompt:
Your role: tariff research assistant, you only produce drafts. Product: wireless earbuds, with microphone, pairs with phone, charging box. Give 2-3 possible HSCP positions and write justification + counter-argument for each. Finalizing the code and tax; Say "verify from the official tariff and consultant." Also list the distinguishing questions that will determine the decision among these candidates.
This prompt turns classification into a process of reasoning; It does not impose a definitive code, it produces the obligation of justification and verification.
three mini cases
Case 1 — Getting rid of excess taxes. An importer had been hesitantly declaring a machine part without AI for years from a high-tax position. With AI, candidate positions and their justifications were extracted; He saw that a lower position might be more suitable. He confirmed the official tariff with his consultant and received Binding Tariff Information. Result: in the right position, the tax burden decreased from 8% to 2.5%, with legal guarantee. AI showed direction, the decision and application were made by experts.
Case 2 — Made-up code penalty. An employee typed the code the AI gave (“he seemed very confident”) directly into the declaration. The code belonged to another chapter and was subject to product surveillance. Customs detected this; Additional accruals and penalties arose, and the shipment waited for days. The damage could have been prevented with a few hours of verification. Lesson: AI code = candidate; statement = expert confirmation.
Case 3 — Set/odd distinction. A company was importing an "installation kit" containing different parts. YZ marked the critical distinction as to whether this product should be classified "according to the part that gives its essential character as a set or piece by piece" and raised the distinguishing questions. The consultant confirmed that the set should be declared in a single code according to its main part; This ensured both accurate tax and simpler filing.
Common mistakes
- Assuming that the code given by the AI is accurate. The code is a candidate; Official tariff and expert confirmation are required.
- Defining the product incompletely. There is no correct classification until the function and material are clarified.
- Getting exact tax rate from AI. Rates vary; It should be read from the official chart.
- Bypassing the supervision/quota/ban check. Even correct code can create additional liabilities.
- Not buying BTB on recurring high volume product. Binding Tariff Information is the strongest legal guarantee.
Tip: When asking the AI for a code, set clear limits such as "don't give the exact code, give candidates and justification, mark the fields to be verified". This single sentence shifts the model from a dangerous tone of “certainty” to a safe tone of “research.”
In summary
GTIP/HS classification is a critical decision that determines tax, restrictions and legal status. In this process, artificial intelligence is a powerful assistant in structuring the product, extracting candidate positions and their justifications, listing distinguishing questions and control headings. However, the exact code and tax rate cannot be written on the declaration without passing the official tariff schedule, Binding Tariff Information when necessary, and the approval of the customs consultant. The cost of wrong code is high; AI output is never the final statement.
Application task
Choose a product that you declare in your business. Extract candidate positions, rationales, and distinguishing questions from AI with the templates above. Then search for these candidates in the official tariff search system, read their descriptions and compare them with the code you currently use: is there a difference, which one is more accurate? Share your findings with your customs broker and ask for confirmation. If in doubt, consider the Binding Tariff Information application.
checklist
- [ ] I have fully described the product with its function, material, technical features and set condition.
- [ ] I asked the AI for a candidate position and justification, not the exact code.
- [ ] I answered the distinctive questions with the actual features of the product.
- [ ] I verified the candidate codes from the official tariff schedule and explanation notes.
- [ ] I confirmed the tax rate from official source, not from AI.
- [ ] I received approval from the customs consultant for the final code; I evaluated BTB if necessary.