Competitive Compass Competitive Compass Signal for Financial Leaders
Mystery Agent  ›  Deposits  ›  Bank of America
Deposits · August 2026

Bank of America's fee PDF works; its rate page shows placeholders

An AI agent ran five ordinary deposits 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 · Bank
61/100
Mystery Agent score
Deposits · rank 10 of 25
Savings rate
25Automated reader declined
Checking comparison
85Served as PDF
Fee schedule
88Served as PDF
Routing number
78Served as text
Application page
30Partly served
Summary
Bank of America also appears in the Credit Cards study, where it scores 55 and ranks 19 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 01Savings rateAutomated reader declined25 out of 100

Both the savings hub and the dedicated rate sheet page arrive at the correct destination, yet the APY figure sits inside an unresolved template placeholder, with the actual number populated only after client-side script runs.

Task 02Checking comparisonServed as PDF85 out of 100

The checking hub's fee schedule PDF shows Advantage Plus Banking at a $12 monthly fee waived by a $250 deposit, and Advantage Relationship Banking at $25 waived by a $20,000 balance, with Relationship paying interest.

Task 03Fee scheduleServed as PDF88 out of 100

A single retry with a plain user agent turned a 406 error into a clean 200 response, delivering a 23-page Personal Schedule of Fees PDF effective February 20, 2026, covering checking, savings, CD, and IRA accounts.

Task 04Routing numberServed as text78 out of 100

The help center's own routing-number link leads to a sign-in screen, but a wire-transfer FAQ page states the routing number as 026009593 in plain text, along with SWIFT codes BOFAUS3N and BOFAUS6S.

Task 05Application pagePartly served30 out of 100

The savings account-opening page carries identity, Social Security Number, and funding requirements inside an unrendered JSON configuration block, while the equivalent checking application link redirects a plain fetch to the mortgage homepage instead.

A single retry with a plain user agent delivered a complete 23-page Personal Schedule of Fees PDF.

The application page

What the account opening page states to a reader that runs no scripts.

166,052Bytes delivered
55,674Characters of visible text
1Form fields in the first response
StatedWhat an applicant needs

Strengths and opportunities

What is working

  • The complete Personal Schedule of Fees is one retry away as a clean, well-organized PDF that fully answers both the fee-schedule job and the checking comparison job with real dollar figures and waiver conditions.
  • The wire-transfer FAQ page publishes the bank's routing number and both SWIFT codes as plain readable text, a genuine agent-readable answer with no login wall.
  • Product hub pages for checking and savings clearly name every product and describe waiver logic and features in readable prose, giving an agent real structural understanding even where exact dollar amounts are withheld.

Opportunities

  • Publish the current savings APY as plain text on the savings and rate-sheet pages instead of leaving it as an unresolved template token that only resolves after client-side script runs.
  • Fill the monthly-maintenance-fee and minimum-balance placeholders on the checking and savings product pages with the same numbers already published in the PDF fee schedule, so the two sources agree without a second fetch.
  • Give the checking account 'Open now' path a GET-able landing page. Today the backend workflow endpoint silently redirects a plain fetch to the mortgage homepage instead of the checking application, which would strand an agent trying to read requirements before applying.
  • Publish llms.txt so agent traffic has a documented, intentional entry point instead of a 404.

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 deposits leaderboard  ·  Read the report  ·  How the study works