Credit Cards · August 2026
Bilt states card rates in text; applications open in a widget
An AI agent ran five ordinary credit cards tasks on the public pages of Bilt Rewards and scored what each page delivered.
Run 6 August 2026 · five tasks · public pages, logged out · Fintech
82/100
Mystery Agent score
Credit Cards · rank 1 of 25
Purchase APR
90Served as text
Card comparison
90Served as text
Pricing and terms
88Served as PDF
Welcome offer
90Served as text
Application page
50Partly served
Summary
- Bilt states the purchase APR, the intro rate, and each card's rewards multipliers as plain text on its card pages.
- An automated reader followed a direct link to an 8 page PDF covering the Schumer box for all three Bilt cards.
- Serving a short static description of the required fields at the application URL would let an automated reader describe that step accurately.
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 text90 out of 100
Reading the Bilt Blue Card page, an automated reader found the purchase APR served as plain text: an intro rate of 10% for the first 12 billing cycles, then a variable APR of 26.74% to 34.74%.
Task 02Card comparisonServed as text90 out of 100
Comparing the Blue and Obsidian cards, an automated reader read that Blue carries no yearly charge and earns 4X at dining partners, while Obsidian costs $95 a year and adds a $100 yearly travel credit.
Task 03Pricing and termsServed as PDF88 out of 100
Following a link from the Blue Card page, an automated reader reached an 8 page PDF listing the Schumer box for all three cards: purchase APR 26.74% to 34.74%, and fees of $0, $95, and $495 a year.
Task 04Welcome offerServed as text90 out of 100
The Obsidian card page states its welcome offer plainly: $200 in Bilt Cash upon approval, with no minimum spend required and the reward ready to use right away.
Task 05Application pagePartly served50 out of 100
A plain request to the application page returned the same marketing shell as the card overview page, with zero input fields, though its FAQ text discloses credit score, income, and existing debt requirements in full sentences.
The Bilt Obsidian Card welcome offer is $200 in Bilt Cash upon approval, with no minimum spend required.
The application page
What the account opening page states to a reader that runs no scripts.
344,744Bytes delivered
9,828Characters of visible text
0Form fields in the first response
StatedWhat an applicant needs
Strengths and opportunities
What is working
- Bilt puts precise numbers in delivered text where it counts most: the purchase APR range, the 12-cycle intro APR, per-card rewards multipliers, and welcome offer values all read as plain text inside the HTML response, not locked behind a script that only a browser can execute.
- The rates-and-fees PDF is a model of the format: one linked document covers the full Schumer box for all three cards, parses cleanly with pdftotext, and is reachable in a small number of hops from the card pages.
- robots.txt is short, current, and honest about what it disallows, and it points to a working sitemap.xml that an agent can use to find every card tier page directly.
Opportunities
- Publishing a real llms.txt would give agents a lightweight index instead of the current result, a 194KB copy of the full marketing 404 page delivered at that path.
- The core marketing and FAQ content lives inside a large Next.js JSON payload rather than semantic HTML elements, which works for a text-scraping agent but is heavier to parse than plain paragraph and list markup would be.
- The application path a plain GET actually reaches is the marketing page again, with no visible form fields, because the live application is handed off to an embedded third-party widget. Serving even a static description of the required fields at that URL would make the application page easier for an agent to describe accurately.
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
