What Changed in Wallets This Year
Four developments that actually altered the security position, and two that were marketed heavily and changed nothing.
An annual review of what moved, based on what we have tested rather than what was announced.
What genuinely improved
Transaction simulation became common. A year ago it was a differentiator; now several major wallets show expected balance changes before signing.
This is the most valuable single change, because it catches the category of attack that vigilance does not: a compromised front end on a correct domain, where nothing else is wrong.
Signature request decoding improved. More wallets now decode permit-style signature requests into readable terms rather than showing a blob.
This matters because signature phishing carries the same authority as a transaction with less friction, and an undecoded request is one you cannot evaluate.
Rollup support became standard. Adding a major layer 2 is now two clicks in most wallets rather than manual parameter entry.
Combined with fees falling substantially after dedicated data capacity was introduced, this changed what is economically sensible to do on-chain.
Withdrawal allowlists spread among venues, though they remain off by default.
What was marketed and changed nothing
AI features. Assistants and categorisation added to wallet interfaces. We have not found one that affects the outcomes we measure.
Social recovery marketing, ahead of the substance. Account abstraction genuinely enables it and the implementations we tested still require guardian sets that need maintenance, which is a dependency on people rather than a solved problem.
What has not improved
The network selector on send screens. Wrong-network transfers remain the most common expensive error and most interfaces still treat the network as a setting rather than as the critical field.
Passphrase handling. Still the most common cause of confusing restore failures, still inconsistently warned about.
Default provider disclosure. Most wallets still do not say which endpoint they query, which means most users do not know who sees their addresses.
What we would like next
A warning when the destination address format does not match the selected network. It would prevent more losses than any other single change and it is not technically difficult.
The position for a user
The tooling is better than it was and the arrangement still matters more than the software.
A separate wallet for connecting, a hardware wallet for holding, and a working balance at the exchange used for the withdrawal test sized to what it is for. That structure has not changed and it is still doing most of the work.
Products covered here are bought at full retail price and used for real transactions before anything is written. There are no affiliate links on this site, no sponsored placements and no review units. If that ever changes, it will be disclosed at the top of the article.