Fee and APR Comparison Questions - Not Product Advice

Shopping for credit often starts with three PDFs and a foggy sense that “the lowest APR wins.” Then someone pastes the PDFs into a chatbot and asks which offer is best. The model may invent a winner while missing intro periods, penalty APRs, or fees buried in footnotes. You needed a table you control. You received unlicensed product advice.

This practitioner piece sits under the foundation AI is not your financial advisor. It is not household budget scenarios (income/expense what-ifs) and not subscription drift review (recurring charges). Here the unit of work is comparing credit or account offers using numbers you already read on the issuer’s documents.

What APR and fee labels are for

Annual percentage rate (APR) is a standardized disclosure used in consumer credit so shoppers can compare more than a headline interest rate. US truth-in-lending rules require APR and related disclosures in many credit ads and agreements - that is a shopping aid, not a robot’s pick (FTC Truth in Lending Act; FTC credit and finance; CFPB; Federal Reserve consumer community resources). Fees (annual, late, foreign transaction, origination) change the real cost even when APR looks similar.

Your job: copy the disclosed figures accurately. The model’s job: format and stress-test your transcription. Nobody in that loop should “award” a product.

Never ask which card, loan, overdraft, or deposit product to take. Never ask the model to invent typical market APRs for your country. If a figure is missing on the offer, mark unknown and verify on the issuer’s official page or with a licensed adviser.

Step 1: Build the grid from primary documents only

Open each offer PDF or official online disclosure. Retype into a table - do not upload the full PDF if it contains account numbers or application IDs (do not paste bank statements into AI).

Offer PDFs can embed application IDs, account stubs, and personal addresses. Retype only the fee and APR fields you need into your grid. Do not paste or upload the full disclosure file into a consumer chat.

Record these seven fields for each offer you are comparing:

1. Label — your own name for the offer
2. APR as stated — copied exactly, not rounded
3. Penalty or variable notes — what changes the rate, and when
4. Annual or monthly fee
5. Other fees noted — transfer, cash advance, late, inactivity
6. Intro period — length, and what the rate becomes afterwards
7. Source date — when you read the disclosure

If a cell is unclear, write unknown - verify rather than letting the model guess.

Step 2: Ask for consistency checks, not a ranking

Here is a comparison table I typed from primary offer documents.
Columns: label, APR as stated, fee as stated, intro period, source date.
Tasks:
1) Check arithmetic only if I include payment examples I typed.
2) List blank or inconsistent cells.
3) List questions I should verify on each issuer's official disclosure page.
Do NOT rank products. Do NOT recommend one. Do NOT invent missing APRs or fees.
Table:
[paste your typed rows]

Illustrative output shape (composite, not a chat log):

Rows checked: 3
Inconsistencies: Product B intro APR ends date blank; Product C foreign-tx fee "see footnote" not copied
Verify next: penalty APR text; how balance transfers are priced; whether annual fee is waived year 1 only
Recommendation: none (per instructions)

Step 3: Convert gaps into human questions

Turn each unknown into a question for the issuer, a nonprofit housing or credit counselor, or a licensed adviser - not into a model guess (HUD housing counseling; USA.gov money and credit; FCA consumers).

Examples of good questions (you ask humans):

• Is the APR fixed or variable, and what index does a variable rate follow?
• When does any introductory rate end, and what rate applies after?
• Which fees are avoidable with behavior, and which are mandatory?
• Are there prepayment penalties or balance-transfer costs not in my table?

Common failure modes

1. Uploading the whole offer pack including application barcodes and personal data.
2. Asking “which is cheapest” when terms differ (fee-heavy vs APR-heavy) - cheapness depends on how you will use the product; that judgment is yours or your adviser’s.
3. Trusting a model-invented “market average APR” as a benchmark.
4. Treating marketing emails as disclosures - verify on the firm’s official channel; UK readers can also check firm authorization (FCA Register; FCA protect yourself from scams).
5. Skipping marketing claim checks when the offer is wrapped in AI-polished hype - use financial product marketing claim check.

One offer at a time into the grid beats dumping five PDFs. Finish transcription, then prompt. Keep the prompt short so the model cannot “helpfully” add a winner paragraph.

Where this stops

If comparison reveals you need debt strategy, tax treatment of fees, or whether you should borrow at all, escalate: debt questions pack and when to stop and call a licensed adviser. EU readers comparing consumer credit marketing should remember unfair commercial practices constraints on misleading claims (Directive 2005/29/EC; Commission consumer complaints path).

Truth-in-advertising expectations still apply to how products are pitched (FTC advertising and marketing; FTC truth-in-advertising topics). Your grid is a defense against both human and AI-smoothed hype.

Worked comparison habits (still not advice)

When two offers look close, people ask the model to break the tie. Break ties with your usage assumptions written down first:

• Will you carry a balance or pay in full?
• Will you use foreign transactions?
• Is the annual fee waived only with spending you will not do?

Write those assumptions as your sentences, then ask only: “Given these assumptions I stated, show arithmetic on fees I typed - still do not pick a product.” If you cannot state assumptions, you are not ready for arithmetic or for an adviser meeting - you are still gathering facts.

Illustrative scenario, not a measured case: Offer A shows 0 annual fee and 22.9% APR. Offer B shows 95 annual fee and 18.9% APR. A model told to “pick the cheaper card” may ignore that the reader never revolves a balance. Writing the assumption down first (“pay in full monthly”) does not produce a winner; it tells you which column your own arithmetic has to run on and which one that assumption makes less relevant. Whether the assumption holds for the whole term is your judgment to make, not the model’s.

Relationship to other finance literacy pieces

• Boundary: AI is not your financial advisor
• Privacy while typing figures: do not paste bank statements into AI
• Ads wrapped around the offer: financial product marketing claim check
• If comparison reveals distress debt: debt questions pack
• If someone is pushing a “guaranteed” investment instead of a loan disclosure: investment claim verification

Deposit-product marketing that implies government insurance should be checked against primary deposit resources, not against a chatbot paraphrase (FDIC deposit insurance; NCUA consumers).

One exercise

Download the fee and APR comparison sheet. Fill two real offers by hand. Run only the consistency-check prompt. End the session without a ranked winner.