The agent keeps watch
An autonomous AI research agent revisits official terms, help centres, announcements and verified public channels on a recurring schedule.
Scoring methodology
How we turn card details into a useful rating
An autonomous research agent keeps the evidence under observation. Reviewed terms become comparable facts, those facts become seven category ratings, and a ranking lens decides their emphasis.
43Current cards
258Sources
3,556Reviewed facts
33Weighted components
7Category ratings
01 · A living evidence system
How the catalogue stays under continuous observation
This is not a static card directory. The catalogue runs on a synchronized evidence system that connects source monitoring, change detection, review and publication without letting an automated observation silently become a public financial claim.
Every signal meets the same data model
New observations are matched to a shared structured layer of sources, facts and card profiles. Complete exact matches can confirm freshness; meaningful differences preserve what changed.
People approve public changes
The agent can discover, compare and prepare an update, but reviewed facts remain the only input allowed to change public claims, scores and rankings.
One useful stream
Approved news and offers meet in one chronological feed
Product launches, availability changes and special offers can begin as signals found by the agent. After review, useful items enter one feed with News and Offers filters; offers also remain available in the dedicated offer explorer. Publication in either place never raises a card's score.
02 · Comparable facts
How provider claims become comparable data
Provider claims are kept, then broken into the values that can be compared across cards. A missing value stays unknown; it never becomes zero cost, full availability or a supported feature.
We separate the $0 card price from mandatory delivery and score the total cost to obtain it.
The entry rate, cap, tier cost and conditions are separated.
Deposit and outbound routes, networks, fees and limits are checked independently.
Internal transfers, bank payouts and on-chain sends are checked separately; only the verified route is treated as instant.
03 · The card itself
Seven ratings describe the product before any scenario
01Cost efficiency
The recurring and transaction costs that determine what the card is likely to cost in practice.
Cost efficiency
The recurring and transaction costs that determine what the card is likely to cost in practice.
Included in this rating
- FX, conversion and top-up
- Plan, card and delivery
- ATM percentage and fixed fees
- Spend, transfer and edge-case fees
02Rewards value
Practical entry-tier purchase rewards, with only limited risk-adjusted credit for separate Earn / Yield utility.
Rewards value
Practical entry-tier purchase rewards, with only limited risk-adjusted credit for separate Earn / Yield utility.
Included in this rating
- Entry rate and monthly cap
- Gated-tier value
- Reward asset and conditions
- Staking, lockup, promotions and yield risk
03Everyday usability
How broadly and conveniently the card can be used for everyday payments and cash access.
Everyday usability
How broadly and conveniently the card can be used for everyday payments and cash access.
Included in this rating
- Card network
- Apple Pay and Google Pay
- Virtual and physical cards
- ATM access and instant issuance
04Funding & transfers
How money can enter and leave the account through approved routes, including the practical quality of each route.
Funding & transfers
How money can enter and leave the account through approved routes, including the practical quality of each route.
Included in this rating
- Receive and send direction
- Crypto and bank payout rails
- Stablecoins plus bounded BTC, ETH and long-tail coverage
- Fees, speed, limits and applicability
05Availability & eligibility
Where the product is offered and how clearly signup and KYC requirements are defined.
Availability & eligibility
Where the product is offered and how clearly signup and KYC requirements are defined.
Included in this rating
- Global status and reach
- Selected-country context
- Country-list clarity
- KYC and signup requirements
06Custody & control
The structured custody boundary, CEX dependency and whether the user controls funds until payment.
Custody & control
The structured custody boundary, CEX dependency and whether the user controls funds until payment.
Included in this rating
- Custody type
- CEX dependency
- Control before payment
07Company & infrastructure
The legal and operational stack behind the card, including whether each named role applies.
Company & infrastructure
The legal and operational stack behind the card, including whether each named role applies.
Included in this rating
- Legal entity and jurisdiction
- Issuer and programme applicability
- Banking and payment responsibilities
Worked example
How “no monthly fee” becomes a Cost efficiency input
A true statement, but only one part of the cost.
- Plan
- $0 / month
- Required delivery
- $40
- FX cost
- 2%
- Conversion cost
- 1.5%
- Transfer fee
- Unconfirmed
Ownership, transaction, cash-access and transfer costs contribute to Cost efficiency.
An unconfirmed transfer fee is not treated as zero.
04 · Rating & scenarios
The ratings stay; the emphasis changes
Overall / Top cards
The balanced baseline used for the main card rating and the default Top cards order.
Cost efficiency25%
Rewards value16%
Everyday usability16%
Funding & transfers10%
Availability & eligibility9%
Custody & control12%
Company & infrastructure12%
05 · Filtering & ranking
Filters narrow the list. Scenarios change what matters most.
- Country context
- Adjusts Availability & eligibility. A card is removed only when the selected country is explicitly unsupported or falls outside a complete supported-country list.
- Other filters
- Selected fee, reward, asset, KYC, custody and feature filters narrow the list to matching cards.
- Scenarios
- Reweight the same seven category ratings, changing the order of the remaining cards.
- Missing or ambiguous country data keeps the card visible.
- Unknown is never a benefit. It cannot improve a score or pass a matching filter.
- Partner links never affect ranking.
06 · A common question
Why do so many top-ranked cards have special offers?
The direction of cause is the reverse. We rank cards from reviewed product facts first. Only then do we prioritize outreach and special-offer work for cards that already rank near the top. That is why offers are concentrated among higher-ranked cards: the ranking guides where we seek an offer; an offer does not lift the ranking.
Bleap Card is a useful counterexample. It ranks near the top, but has no public offer on this site. Bleap's consumer referral links are personal and cannot be used for commercial referral activity without express permission, so we do not publish one. This does not change Bleap's score or position.
Use the model