Wallet Onboarding in 2026: Where the Funnel Still Leaks

The wallet got invisible, and the drop-off did not

In the last week of October 2026, Samsung will switch on a stablecoin feature inside Samsung Wallet for roughly 82 million eligible Galaxy devices in the United States, letting owners hold and send USDC without downloading a separate app or writing down a recovery phrase, according to Samsung. A few weeks earlier, OKX launched OKX Money, a standalone stablecoin app aimed at Latin America, Africa, South Asia and the Middle East, and said that roughly 70 percent of the people it wants to reach have never used a crypto app at all, per Cryptobriefing. Both products share a design goal that would have sounded like heresy in 2021: keep the blockchain out of sight.

This is the clearest sign yet that the industry has won the fight it was having with its own front door. For a decade the first thing a new wallet asked you to do was copy twelve random words onto paper and promise never to lose them. That screen is finally optional. Passkeys, embedded wallets and social login have made account creation feel like signing up for any other app, and the numbers at the top of the funnel show it.

And yet the adoption math has barely moved for the hard part. Industry teardowns still put the share of new users who quit before their first on-chain transaction at between 60 and 90 percent, with fewer than one in five staying a week and onboarding churn above 70 percent, according to a review of wallet UX by The Block. The leak did not close. It moved.

This piece walks the onboarding funnel the way a product team would, one step at a time, and asks the same question at each stage: where does a normal person give up, and what did 2026 actually fix? The short version is that the door is solved and the rooms behind it are not. Funding, the first fee-paying transaction, recovery and the tax exit are where people still fall through the floor.

What a funnel teardown actually measures

A funnel is just the ordered list of things a person has to do before a product delivers its promise, with a measured drop-off at each step. For a neobank the promise is spending money and the funnel is short: install, verify identity, add a card, tap to pay. Studies of consumer fintech routinely show the vast majority of new accounts transacting inside the first month. For a self-custody crypto wallet the promise is owning and moving an asset with no intermediary, and the funnel is long, unfamiliar and unforgiving of mistakes.

The gap between those two retention curves is the single most important number in consumer crypto, and closing it is what every onboarding product of the last two years has been chasing. The honest way to read the funnel is step by step, because the fixes that landed in 2026 are concentrated at the top while the losses are spread across the middle and the end. The table below maps the six stages, what a user actually hits at each one, where they tend to abandon, and the tooling that is meant to rescue the step.

Read down the last column and a pattern appears. Step one has a clean, shipped answer. Steps two through six have partial answers, workarounds, or an answer that quietly changes what kind of wallet you are using. That is the story of onboarding in 2026.

Step one: install and create, where the seed phrase finally left

The biggest win of the cycle is the one users will never notice, because the whole point is that nothing happens. Creating a wallet in 2026 can be as simple as tapping a Sign in button backed by a passkey, a cryptographic credential stored in the phone’s secure hardware and unlocked with a face or a fingerprint. The FIDO Alliance counted roughly five billion passkeys in use worldwide by its World Passkey Day 2026 tally, with 90 percent of people aware of them and three quarters having enabled at least one. The technology new users already trust to log into their email now stands in for the seed phrase.

Under the hood, a precompile called RIP-7212 lets a smart-contract wallet verify those device-generated signatures on-chain for roughly 3,450 gas instead of around 300,000, a change documented by Alchemy that turned passkeys from a nice idea into something economically usable on networks such as Arbitrum and Polygon. The other half of the win is the embedded wallet: a key quietly generated and held inside a secure enclave by an infrastructure provider, so a user signs in with an email or a Google account and never sees a private key at all.

That category consolidated fast. Stripe bought the embedded-wallet startup Privy in June 2025; Fireblocks acquired a similar firm, Dynamic, in October 2025; and Coinbase made its CDP Embedded Wallets generally available to apps of any size, offering email, SMS and social sign-in with no seed phrase and an on-ramp built in, per the Coinbase Developer Platform. For a builder, adding a self-custody wallet is now a few lines of code rather than a UX research project.

The marketing language tells you where this is heading. Samsung’s Woncheol Chai, who runs the company’s Digital Wallet team, framed the stablecoin launch around removing exactly this step, saying eligible Galaxy users should be able to “get started without installing another app or managing private keys themselves.” Phantom chief executive Brandon Millman has described the ambition even more plainly, telling investors in the wallet’s Series C announcement that the goal is to become the “world’s biggest and most-trusted consumer finance platform.” Nobody sells seed phrases anymore. They sell money.

The hidden cost of a frictionless door

There is a catch in step one that most onboarding copy skips over. A wallet can become frictionless in two very different ways, and they are not the same product even when they look identical on screen. In the first, the key is still yours; it is just generated and stored inside secure hardware by an infrastructure provider, split into shares or sealed in an enclave the provider itself cannot read. Coinbase’s embedded wallets and passkey-based smart wallets work this way: self-custody, with the backup burden moved off the user and onto the device and the cloud. In the second, there is no key for you to hold at all, because a licensed company holds the asset and you hold a claim on it.

Samsung’s stablecoin feature is the second kind. The USDC balance sits with Bastion, a licensed custodian, with Coinbase Prime acting as sub-custodian, which is why the experience can feel as smooth as any banking app: there is nothing for the user to lose. OKX Money is the same shape, an exchange balance dressed as a savings app. This is not a criticism, it is a definition. The smoothest onboarding in 2026 is frequently custodial onboarding, and the thing that got removed to make the door easy was self-custody itself.

For a remittance user sending dollars to 60 countries, that trade is probably worth it. For someone who came to crypto specifically to hold their own assets, it is the whole game, and the funnel they actually want is the harder one. A useful onboarding product is honest about which door it is, because the answer determines everything downstream: who can freeze the account, who can recover it, and who shows up when something goes wrong. We return to that question when we get to recovery and regulation.

Step two: funding, where KYC and the on-ramp still bite

Assume the user has an account. Now it has to hold value, and this is the first place the modern funnel bleeds badly. Getting money in means an on-ramp, and an on-ramp means identity verification, a linked card or bank account, a purchase that a card issuer may decline as suspicious, and often a wait. The flow runs through a familiar sequence: email verification, know-your-customer checks, payment linking, a deposit, asset selection and finally the buy. Every one of those steps is a place to quit, and plenty of users do.

The 2026 answer is to collapse the on-ramp into the wallet and denominate everything in a dollar stablecoin, so the user thinks in money rather than in tokens. OKX Money lets a user deposit local currency and converts it automatically into a dollar-backed stablecoin, holding USDG, USDC or USDT, with a card that spends at no foreign-exchange markup and up to 10 percent advertised yield on eligible balances, per Cryptobriefing. Samsung routes USDC transfers to more than 60 countries from inside the wallet already on the phone. Funding stops being a scavenger hunt and becomes a top-up.

That yield deserves a footnote, because it is where funding quietly turns into a financial product. When an app pays you to hold a stablecoin, someone is earning on the float and deciding how much to pass along, and the rules governing who may do that, and who keeps the difference, are still being written. We mapped that fight in our look at who keeps the interest under 2026 stablecoin yield rules. For onboarding, the point is narrower: the smoothest way to fund a wallet now ships with an interest rate attached, and a new user has no easy way to tell a sustainable yield from a subsidy.

Step three: the first transaction and the gas wall

Here is the step that breaks the illusion. A user who funded with USDC, sees a dollar balance and feels like they are using a bank app suddenly tries to send it and learns they need a second, different token to pay the network fee. On Ethereum that token is ETH; on other chains it is something else; and the amount changes minute to minute. To a newcomer this is baffling: they have money, the screen says they have money, and the app will not let them move it until they go buy a sliver of a coin they have never heard of. Gas-fee confusion is one of the most cited reasons setup fails, and it arrives precisely when the user thought they were done.

The structural fix is the paymaster, a contract that sponsors the fee so the user never touches a gas token. In the sponsored model a dapp or wallet eats the cost to get you through your first action. In the pay-with-what-you-have model the fee comes out of the same stablecoin you are already holding: Circle’s Paymaster lets a user pay gas in USDC on Arbitrum and Base, works with both ERC-4337 smart accounts and EIP-7702 upgraded EOAs, and charges a 10 percent surcharge on the gas for the convenience, per Circle. Either way the goal is the same: the word gas never appears.

Paymasters are also why so much of a wallet’s economics now sits in the step after the first transaction, the one the funnel calls keep using. A hot wallet earns when you trade inside it, and by one estimate close to 80 percent of a hot wallet’s revenue comes from in-wallet exchange activity rather than from the holding itself, which is why every consumer wallet is racing to add swaps, cards and even perpetual futures. The wallet is turning into the exchange, a convergence we traced in our explainer on how perp DEXs work. For onboarding, it means the friction you remove at the gas step pays for itself in trading fees later.

The approval screen you cannot read

Even with gas abstracted away, there is one screen left that scares people for a good reason: the signature request. The first time a wallet asks a user to approve a transaction, it often shows a wall of hexadecimal and a button that says Confirm. Experienced users call signing something you cannot read blind signing, and it is not a beginner problem only; it was the mechanism behind the largest exchange hack on record, where signers approved a transaction that did not do what the interface claimed. A new user staring at an unreadable approval and a warning about irreversibility will very reasonably close the app.

The onboarding answer has three layers. Clear signing translates the raw payload into a human sentence describing what the transaction actually does, an effort now standardized as ERC-7730 and stewarded by the Ethereum Foundation. Transaction simulation runs the action against a fork of the chain first and shows the user the assets that would leave and arrive, so a drain looks like a drain before it is signed. And protective routing tries to keep the transaction from being exploited in the mempool, where searchers reorder and sandwich trades for profit, a market we described in the MEV oligopoly. None of this makes a first approval risk-free, but it turns an act of faith into a decision a person can actually make.

Recovery: the step everyone postpones

Recovery is the step that is easiest to skip during onboarding and most expensive to get wrong afterward, because in pure self-custody there is no recovery at all: lose the key and the money is gone, with no support line and no reset. Users do not understand this, and the data is blunt about it. In a peer-reviewed CHI 2025 study of 643 crypto users, only 43.4 percent could correctly identify an image of a seed phrase, and a majority of those who could still believed they were able to choose or reset it like a password, according to the ACM paper. People are being handed the single most important secret of their financial lives and misfiling it as a login they can recover.

The modern fixes trade the paper phrase for mechanisms humans handle better. Passkey wallets back the key up through the device ecosystem, syncing an encrypted copy via iCloud Keychain or Google, which is wonderful until the user loses the Apple ID or falls to a SIM-swap and discovers the recovery path runs through a platform account they also neglected. Social recovery spreads the ability to restore an account across a set of guardians, friends, other devices or an institution, so no single loss is fatal. Multi-party computation splits the key into shares so there is no single seed phrase to lose in the first place. Each one moves the burden somewhere the user is less likely to fail.

This is the exact problem Vitalik Buterin flagged years ago when he argued that usability and self-custody are not opposites. The weakness of the seed phrase, he wrote, is that it fights human nature: “The human brain is quite poorly suited for remembering passwords and tracking paper wallets, but it’s an ASIC for keeping track of relationships with other people,” which is the whole case for social recovery wallets. The lesson for onboarding is that a wallet which cannot be recovered has not finished onboarding the user; it has only postponed the moment of failure.

The exit nobody designs

The last step of the funnel is the one no onboarding team puts on a slide, because it happens months later and feels like someone else’s problem: getting value back out, and accounting for what happened in between. Off-ramping to a bank account runs into the same KYC and limits as the on-ramp, in reverse. And in the United States the tax treatment is the quiet trap, because selling, swapping or spending a crypto asset is a disposal that can trigger a capital gain or loss. The stablecoin app that made funding feel like a top-up also turned every card tap that converts the balance into a reportable event, and the user has no idea.

That reporting is now arriving in the mail. US custodial brokers have begun issuing the new Form 1099-DA for digital-asset sales, a reporting regime the Internal Revenue Service phased in starting with 2025 transactions, which means a lot of first-time users will get a tax form for activity they thought was invisible. Custodial onboarding, the frictionless kind, is exactly where those forms come from. A good product would surface the tax consequence at the moment of the taxable action rather than in February, but almost none do, and the exit remains the least-designed part of the entire journey.

Four onboarding models, scored

Because the fix for a leaky step so often changes what kind of wallet you are holding, it helps to line up the main onboarding models side by side. The four below cover almost everything a mainstream user will meet in 2026, from the original write-down-your-words wallet to the fully custodial balance that hides on a phone. The column that matters most is the last one, because it decides who is responsible when the account is lost, frozen or hacked.

No row is strictly best. The friction drops as you move down the table, and so does the amount of control the user actually has. The design question for 2026 is not how to eliminate the trade-off, which is impossible, but how to let a user pick their row knowingly instead of being defaulted into one by whichever product had the smoothest signup.

Where the friction actually moved

Step back from the individual stages and the shape of the problem is clear. In 2021 the funnel lost most people at the very first screen, the seed phrase. In 2026 that screen is gone for anyone who wants it gone, and the losses have redistributed: to the on-ramp, where KYC and payments fail; to the first transaction, where gas and an unreadable signature frighten people off; to recovery, which users skip until it is too late; and to the tax exit, which nobody warned them about. The leaky bucket did not get a new bottom. The holes just moved up the sides.

This is why the headline retention numbers have been so stubborn even as signup got dramatically easier. A product can win the install and still lose the user three steps later, and the aggregate drop-off of between 60 and 90 percent before a first transaction reflects a funnel that is excellent at the top and mediocre in the middle. The teams closing the gap fastest are the ones that treat funding, the first fee and recovery as part of onboarding rather than as the user’s problem once the account exists.

There is also a new frontier opening just past the funnel, and it is worth naming because it changes what onboarding even means. Once a wallet can create itself, fund itself and recover itself with almost no user effort, the next question is what it is allowed to do on its own, through session keys, subscriptions and AI agents that spend on your behalf. We examined that shift in a companion piece on when your wallet spends without you. Onboarding used to end when the user could act; increasingly it ends when the user has decided what the wallet may do without asking.

The new trust surface behind the invisible wallet

Making a wallet invisible does not remove the parts the user stopped seeing; it hands them to someone else. Every frictionless onboarding flow rests on a stack of companies the user will never hear named, and the trust that used to live in a seed phrase on paper now lives in their infrastructure. When Stripe owns Privy and Fireblocks owns Dynamic, a large share of new consumer wallets are being minted by a short list of providers, and the signup screen that feels like magic is really a dependency on a key held in someone’s enclave, a recovery flow written by a vendor, and a custodian you did not choose.

The table below names the layers a typical embedded or custodial wallet depends on, and what you are actually trusting each one to do. None of these are inherently bad; secure enclaves and audited custodians are genuinely safer than a sticky note in a drawer for most people. But concentration is its own risk: a bug or a breach in a widely used embedded-wallet SDK or a shared custodian has a blast radius measured in millions of accounts, which is why the question of who audits the infrastructure matters more as the stack consolidates. We looked at that in our review of which audit firms to trust after a brutal year for security.

Phishing is the reminder that the user remains the weakest layer no matter how good the plumbing is. Wallet-drainer losses fell sharply in 2025, down 83 percent to roughly 84 million dollars across just over 106,000 victims, according to Scam Sniffer data reported by Cointelegraph, but the same report found that malicious token approvals, the Permit and Permit2 signatures, accounted for 38 percent of the losses above a million dollars. The attacks that survive are the ones that trick a person into signing, which is exactly the screen onboarding tries hardest to make effortless.

What the SEC makes of a wallet you cannot see

The regulatory line in the United States runs straight through the middle of the onboarding models, and it falls exactly where custody does. In April 2026, staff at the Securities and Exchange Commission signaled that software which merely lets a user transact from a self-hosted wallet is not acting as a broker, a position reported by CoinDesk. Pure self-custody tools, including the passkey and embedded wallets that keep the key with the user, largely sit outside the broker framework, which is also why nobody is coming to refund a user who signs their funds away.

Custodial onboarding is a different animal. The moment a company holds the asset, as Bastion does for Samsung or an exchange does for its own balances, it steps into custody, money-transmission and consumer-protection obligations, with licensing, audits and the duty to give the money back. For a new user this is the paradox worth understanding before picking a wallet: the smoothest, most protected-feeling products are protected because someone else is holding the bag and is answerable for it, while the self-custody products that feel riskier are the ones where the law expects the user to protect themselves. The SEC can police the broker at the edge; it cannot un-sign a transaction. Onboarding that hides this distinction is doing the user a quiet disservice.

What good onboarding looks like in 2026

Pulling the funnel together, the wallets that onboard real people best in 2026 are not the ones with the flashiest signup; they are the ones that refuse to let any single step become a cliff. A short, honest checklist captures what that looks like in practice.

• No seed phrase at the door, but a clear, up-front statement of who can recover the account and who holds the keys.
• Funding that works in local currency and dollars without sending the user to a separate exchange just to get started.
• A first transaction with gas abstracted away, either sponsored or payable in the stablecoin the user already holds.
• Signatures shown in plain language, with simulation that previews what will actually leave the wallet.
• A recovery method set up during onboarding, not offered as an afterthought the user will skip.
• An honest word about tax before the first disposal, not a surprise form in the new year.

Notice that almost none of this is about cryptography. The hard problems left in wallet onboarding are product and disclosure problems: making the trade-offs legible, meeting the user in the money they already understand, and treating the whole journey from install to tax as one experience rather than a signup flow followed by a shrug. With Bitcoin trading near 83,000 dollars and Ether around 2,500 in mid-October 2026, per CoinGecko, the market these wallets are onboarding into is large and mainstream enough that the funnel is no longer a niche concern; it is the growth story.

The seed phrase leaving the front door was the headline of the last two years, and it was a real achievement. But a door is only the first step into a building. The wallets that win the next phase will be the ones that figured out the rooms: funding, the first fee, recovery and the exit, the quiet places where normal people still turn around and leave. That, and not the login screen, is where wallet UX is actually decided in 2026.

Frequently Asked Questions

Do I still need a seed phrase to use a crypto wallet in 2026?

No. Many wallets now let you sign in with a passkey, an embedded key tied to your email or social login, or a custodial balance with no key at all. The trade-off is recovery: before you pick one, make sure you know who can restore the account if you lose your phone, and whether anyone can freeze it.

Why do most people quit a crypto wallet before their first transaction?

Because the friction moved past signup. Industry teardowns put the drop-off between 60 and 90 percent before a first on-chain transaction, and the losses now cluster at funding, where KYC and on-ramps fail, and at the first transaction, where gas fees and an unreadable signature scare people off.

Is an embedded or invisible wallet still self-custody?

It depends on who holds the keys. Coinbase embedded wallets and passkey smart wallets keep the key with you, sealed in secure hardware, so they are self-custody. Samsung Wallet and OKX Money hold the stablecoin for you through a licensed custodian, so they are custodial. Check which one you are using before you trust it with real money.

How do gasless transactions actually work?

A paymaster, a smart contract that pays the network fee on your behalf, either sponsored by the app or funded from your own balance. Circle Paymaster, for example, lets you pay gas in USDC on Arbitrum and Base and adds a 10 percent surcharge on the fee, so you never need to hold a separate gas token.

Does using a crypto wallet create taxes in the United States?

Usually yes. Selling, swapping or spending a crypto asset is a taxable disposal that can produce a capital gain or loss, and US brokers have started issuing the new Form 1099-DA for 2025 sales. Self-custody wallet software itself is not treated as a broker, but your transactions are still reportable.

By Yuki Tanaka, senior wallets and exchanges correspondent at HOGE Wire.