Credit Cards · August 2026
Bank of America embeds APR and publishes full terms in PDFs
An AI agent ran five ordinary credit cards tasks on the public pages of Bank of America and scored what each page delivered.
Run 6 August 2026 · five tasks · public pages, logged out · Issuer
55/100
Mystery Agent score
Credit Cards · rank 19 of 25
Purchase APR
70Served as text
Card comparison
58Partly served
Pricing and terms
86Served as PDF
Welcome offer
48Partly served
Application page
15Arrives after scripts
Summary
- Bank of America embeds schema.org data directly in every product page, giving an agent a structured APR figure without running any script.
- An agent extracted a complete six-page pricing table from the Customized Cash Rewards agreement PDF, covering purchase, balance transfer, and penalty APR.
- Moving the annual fee and welcome-offer spend requirement into the same data block that already carries the APR would close the gap.
Bank of America also appears in the Deposits study, where it scores 61 and ranks 10 of 25.
See that scorecard.
Task by task
Each task is scored out of 100 on completion, answer quality and how directly the answer was reached. Each note describes what the page delivered, with a link to the page the answer was found on.
Task 01Purchase APRServed as text70 out of 100
The Customized Cash Rewards product page carries structured data stating an annual percentage rate of 18.24% to 28.24%, the same figure repeated on the Travel Rewards page and confirmed in a six-page agreement PDF.
Task 02Card comparisonPartly served58 out of 100
Comparing Customized Cash Rewards and Travel Rewards, an agent read a 6% category-choice cash back structure against an unlimited 1.5 points per dollar rate, both listed under the site's own no-annual-fee card category.
Task 03Pricing and termsServed as PDF86 out of 100
The Credit Card Agreements page links 21 named PDF disclosures with no login required; the Customized Cash Rewards agreement extracts to a full six-page pricing table with APR ranges, fees, and a Military Lending Act notice.
Task 04Welcome offerPartly served48 out of 100
The product page states the welcome bonus value directly: a $200 cash rewards bonus alongside 6% cash back in a chosen category for the first year, though the spend requirement and time window load through a client-side call.
Task 05Application pageArrives after scripts15 out of 100
The application workflow link returns a legacy page titled Timeout, whose entire visible text asks readers to enable JavaScript, with the application form and its requirements arriving only after scripts run.
Earn 6% cash back in the category of your choice for the first year, plus a $200 cash rewards bonus.
The application page
What the account opening page states to a reader that runs no scripts.
8,977Bytes delivered
102Characters of visible text
0Form fields in the first response
After scripts runWhat an applicant needs
Strengths and opportunities
What is working
- Bank of America embeds schema.org CreditCard JSON-LD directly in the delivered HTML of every product page, giving an agent a structured, machine-readable APR figure while staying clear of execute any script.
- The Credit Card Agreements page publishes a clean, unauthenticated index of 21 named PDF disclosures, and the PDFs themselves are well-formed, text-extractable, and genuinely complete Schumer-box-style documents.
- Meta description tags double as a second reliable source of plain-English product facts (rewards rate, welcome offer value), so key numbers survive even when the visible page body is a client-rendered placeholder.
- robots.txt is transparent and specific about what is off-limits, and the credit-cards section is deliberately left open to crawlers and agents.
Opportunities
- Publish an llms.txt file. Right now an agent has no manifest pointing to the JSON-LD, the PDF agreements index, or the product pages, and has to discover all three by trial and error.
- Move the annual fee, rewards tiers, and welcome-offer spend requirement into the same JSON-LD or meta description block that already carries the APR. The data clearly exists since the marketing copy on the rendered page references it; publishing it as static text would close most of the gap in one step.
- Render the 'Terms and Conditions' HTML view server-side, or at minimum leave the linked PDF fully populated. Right now that page is an empty div with a non-functional 'View as PDF' button in the static response.
- Serve a lightweight, no-JavaScript landing step for the application flow that states the basic requirements, SSN, income, and eligibility, before handing off to the interactive form. A one-paragraph disclosure would turn the current 'Timeout' shell into a genuinely informative application page.
Access
Every page the agent requested was delivered on the first attempt. Publishing an llms.txt file would give agent traffic a documented entry point.
Back to the credit cards leaderboard · Read the report · How the study works
