How to Choose an Online Broker in 2026: The Checklist, the Documents, the Traps

Nobody audits their broker. The position gets reviewed, the funds get compared, the allocation gets rebalanced, and the intermediary that touches every single one of those decisions gets chosen once, on an app-store rating, and kept forever. This page is the checklist for choosing it deliberately: what actually varies between brokers in 2026, which documents prove it, which traps cost the most, and how to price the whole relationship on your own plan before it exists. The economics that hold at any broker once the account exists, sweep mechanics, order-flow economics, protection limits, margin, are the subject of the account-level economics at any broker; this page compares the intermediaries themselves.

1. The broker is a recurring cost, not a storefront

A broker is chosen the way a bank branch used to be, on proximity and familiarity, and priced like neither of those things: it takes a share of every deposit, every trade, every idle dollar, for as long as the account lives. That reframing is the page’s working thesis. The decision belongs early in the sequence mapped by the beginner’s investing roadmap, and choosing on the wrong criteria sits high in beginner mistakes with the biggest price tag: not because any given broker is bad, but because the differences that matter are exactly the ones the onboarding flow never shows. The flow shows the app. The account agreement shows the business.

Your broker is the one holding you never rebalance. Positions get reviewed yearly; the intermediary holding them almost never is, which gives brokers something no fund enjoys: a customer base whose switching rate approaches zero, and pricing power follows retention wherever retention is inattention. The economics follow directly from that retention. Since headline commissions hit zero in 2019, revenue moved to the lines inertia shields most effectively, the cash sweep, order routing, margin, lending, and the dispersion on those lines is now wider than commissions ever were: on a $100,000 cash balance, the 2026 sweep gap alone exceeds $3,000 a year, more than a decade of the commissions the industry stopped charging. Sorting deliberately, once, is worth more than any amount of fund selection layered on top of a costly pipe.

The dominant selection method deserves naming, because it is the trap. Brokers are chosen on sign-up bonuses, interface polish and familiarity, three criteria that share one property: they price the first month, not the next twenty years. A $200 transfer bonus against a sweep line that underpays by $1,000 a year on a normal cash balance is not a close call, yet the bonus converts and the sweep does not, because one is visible at the decision moment and the other surfaces slowly, on statements nobody annualizes. The industry’s acquisition economics assume exactly this asymmetry; the checklist exists to invert it.

2. The evaluation grid: six criteria, all verifiable

The grid replaces the ranking question with the answerable one: on which lines does this broker earn, and what do the documents say each line costs. Six criteria, each with a public document attached, none requiring the broker’s cooperation to check:

The grid deliberately omits sign-up bonuses, app polish and star ratings: the first is a one-time transfer priced against years of recurring lines, the second is free to copy, the third measures onboarding friction rather than cost, three different ways of grading the lobby instead of reading the lease. What the six lines share is permanence. A broker can redesign its app in a quarter; its sweep economics and routing arrangements are business-model choices that persist, which is what makes them worth choosing on.

The wrapper layer runs upstream of all six and is settled before any broker matters: the wrapper question before the broker question decides where a dollar lives, and the match-first funding order decides which account gets it first, both independently of the intermediary. The six lines do not weigh equally for a given investor, and the weighting is the half of the work no comparison site can do. An index household activating two trades a quarter lives on the sweep line and the fund menu; an options trader lives on routing and per-contract fees; anyone borrowing lives on margin terms; everyone lives on protection, which is why it is pass-fail rather than weighted. Filling the grid takes one evening per broker; weighting it takes knowing your own order sizes, cash balances and leverage, numbers only you have and the next section’s simulator turns into a single figure. One structural note for readers comparing across the Atlantic: the account-type line is where national systems diverge hardest, and France’s PEA versus a standard brokerage shows how a tax wrapper can be broker-dependent in one country and universal in another. The comparison is a useful mirror: it isolates which parts of a broker’s grid are business choices, portable across borders, and which are artifacts of national regulation that no amount of shopping changes.

3. Costs, visible and invisible

Cost has two layers and the visible one is nearly extinct: stock and ETF commissions at zero across the large houses, options at about $0.65 per contract, transaction-fee mutual funds at $50 to $75 per purchase (published schedules, 2026). What survived is invisible by design. The sweep spread on idle cash is the largest single line for most holders: defaults observed in 2026 run from 0.01% at brokers sweeping to affiliated banks to roughly 3.3% to 3.6% at those sweeping to money market funds, a gap exceeding $3,000 a year per $100,000 of cash, which drew SEC attention and class actions in 2024. Routing revenue is the second invisible line: payment for order flow, disclosed per broker in quarterly 606 filings, prices in hundredths of a cent per equity share and tens of cents per options contract, which is why the line weighs little for an index household and heavily for an options trader. Margin interest, securities-lending splits and FX conversion on foreign securities fill out the stack.

The FX line deserves its own sentence for anyone holding foreign listings: currency conversion is priced as a spread on the exchange rate, rarely itemized, occasionally layered on top of an explicit fee, and it applies twice, at purchase and at sale, plus on every dividend repatriated. For a plan weighted toward foreign-listed securities it can quietly outweigh every other line on the schedule. The trendline behind the snapshot is worth one paragraph, because it explains where the traps migrated. Explicit trading costs have fallen for two decades on both sides of the Atlantic, from fixed commissions to discount pricing to zero in 2019; each compression pushed revenue one layer deeper into the account. The French regulator’s 2026 survey shows the same gradient in explicit form, a one-to-four ratio between traditional banks and new entrants on an identical order, and the American market shows it in implicit form, identical zero stickers over sweep rates two orders of magnitude apart. The lesson transfers: wherever the visible price converges, the comparison moves to the invisible lines, and the invisible lines are documented for anyone who asks.

The comparison method matters more than any single line: cost is a product of price times personal usage, and only the holder knows the second factor. A 3% sweep gap on an account that holds no cash costs nothing; a $0.65 contract fee on four hundred contracts a year is real money. The checklist question is never “which broker is cheap” but “which lines does my behavior activate, and where are those lines thin”. Two investors at the same broker can pay costs an order of magnitude apart; two brokers can cost the same investor identical amounts through different lines. Where the idle-cash line itself can sit, inside or outside the brokerage, is the shelf mapped in where uninvested cash can sit.

4. What fees do to an investment plan

The simulator runs one automatic investment plan under two generic pricing profiles: profile A charges a flat fee per order; profile B charges a percentage with a minimum, plus a conversion cost on the foreign-listed half of the plan. Same deposits, same declared 5% gross assumption on both branches; the order size you choose sets how often the plan trades, which is precisely the lever that decides which grid bites harder. Neither profile is a recommendation; both are patterns documented in published 2026 fee surveys. Watch the minimum-fee mechanics in particular: a percentage grid with a floor punishes small orders disproportionately, which quietly pressures the plan toward larger, rarer executions, a behavioral side effect the flat grid does not produce. The tool prices both without endorsing either.

5. Protection and solidity: what the documents actually say

Protection is where broker marketing and broker documents diverge least, and where reader misunderstanding runs deepest. Every US brokerage worth the name is a SIPC member: custody of securities and cash is protected up to $500,000 per customer including $250,000 in cash if the firm fails, with private excess-SIPC coverage layered above at most large houses. The line is pass-fail rather than a ranking, and it protects custody, never market value. What differentiates brokers on this criterion is not the ceiling but the plumbing: whether fully paid securities are enrolled in lending programs by default, how cash sweeps interact with FDIC versus SIPC coverage, and how account-takeover security is built, the failure mode retail investors actually experience. Each answer sits in a disclosure document, not in the marketing. The lending question repays two minutes of reading: programs that lend fully paid shares split the revenue anywhere from half to nothing, and lent shares receive cash-in-lieu rather than qualified dividends, a small recurring tax cost, invisible until the 1099 arrives, carried by holders who never knowingly enrolled. Whether enrollment is opt-in or default is a one-line difference between brokers with a compounding price. Related discussion: our analysis “Passive management”.

Solidity, in the balance-sheet sense, is the quietest criterion because the regime works: segregation rules kept customer property whole through the 2008 Lehman brokerage liquidation, the one full-scale test in living memory. One coverage subtlety separates otherwise identical brokers: the sweep destination changes the insurance regime. Cash in a money market sweep is an investment under SIPC custody rules; cash in a program-bank sweep is a deposit under FDIC rules, spread across partner banks with per-bank ceilings that large balances can exceed without anyone flagging it. Neither regime is superior in the abstract; they are different documents answering different failure scenarios, and the account agreement names which one applies in the same paragraph that discloses the rate. The practical reading for a chooser is therefore inverted: spend the diligence minutes on the sweep paragraph and the 606 report, where brokers differ by orders of magnitude, not on failure scenarios where they barely differ at all.

6. Practical frictions: the costs of moving and the costs of staying

Two frictions bracket the relationship, and only one of them is advertised. Entering costs little, by design; leaving is where the terms hide. An ACATS transfer moves whole positions between brokers in about a week, in kind and without a tax event, but the delivering side typically charges an outgoing fee, and anything proprietary, funds that exist only on that broker’s shelf, cannot travel: it is sold or abandoned; the receiving broker often reimburses the outgoing fee on request, a detail worth knowing since the delivering side rarely advertises that its charge is refundable elsewhere. That detail about proprietary shelves turns some product lineups into soft lock-in, worth spotting before the account exists rather than after a decade of accumulation in a fund that cannot move.

The second friction is operational and chronic: how dividends are handled and reinvested, whether fractional shares make small automatic purchases possible, which order types exist, what market data costs, how tax documents arrive and when. None of these lines is individually large; together they set the monthly texture of the plan. Tools deserve the same document treatment as fees: real-time data that costs extra, order types missing from the mobile app, alerts that exist on desktop only, tax lots that cannot be selected at sale. A platform trial with a funded test order reveals in an hour what the feature comparison tables cannot, because the tables are written by the marketing department and the order ticket is not. An hour of trial per finalist is the cheapest diligence on this page. The product route chosen upstream, funds, ETFs or direct holdings, determines which of them bind, a dependency mapped in three routes into the market, and the instrument-level screen that follows account opening has its own grid in screening ETFs once the account is open, with the vocabulary base in ETF basics for first-time investors. The frictions are small precisely because they are chronic: a plan that runs monthly for twenty years executes two hundred forty times, and anything that costs attention per execution costs it two hundred forty times.

7. The questions to ask before opening

Documents beat demos. Seven questions compress the grid into an evening’s work, each answerable from documents: What does the default sweep pay today, in writing? What does the current 606 report show about routing concentration and the payment received per share and per contract? Which fees would my actual behavior trigger, at my order sizes and frequency? What does an outgoing ACATS transfer cost, and which holdings could not travel? Are my shares lent by default, and on what split? Which account types and order features do I need on day one, from the menu in monthly investing amounts for beginners to fractional purchases? And what would this broker earn from me in a year, the question that reconciles all the others, answered by the simulator above with your own numbers. Any broker whose documents make these seven answers hard to find has, in a sense, already answered an eighth, and answered it in the only language a fee schedule speaks.

The regime backdrop gives the exercise its weight. In the current transition reading, mixed signals without a clear direction (Eco3min classifier, June 2026, via the macro regime right now), expected real returns on diversified portfolios are modest, and every basis point of friction is a larger fraction of a smaller pie: friction that consumed a twentieth of a strong regime’s real return consumes a fifth of a weak one’s; performance history across regimes shows how thin the real-return margins have run in comparable configurations. Friction is the one line the investor controls completely, and control has no expiry date: the grid chosen today prices every deposit of the next decade. Rates, inflation and drawdowns are not offered for negotiation; the broker’s grid is, once, at account opening, by choosing it on documents.

8. FAQ

What does execution quality mean for a retail order?

How close the fill lands to the most favorable price available at the moment of execution, measured against the consolidated market benchmark. It varies by broker because routing choices vary: venue-level Rule 605 statistics quantify price improvement, and each broker’s 606 report shows where it sends orders. For small orders in liquid stocks the differences are cents per share; the point of checking is confirming they stay that small at your broker, which one quarterly report per candidate answers.

How does PFOF affect the fill price?

Payment for order flow means a market maker pays the broker for the right to execute its orders, earning the spread. Fills still land at or inside the national benchmark; what varies is how much price improvement is passed to the customer versus retained in the chain. The effect per equity share is fractions of a cent, but options flow pays the broker roughly a hundred times more per contract, so the answer differs by what you trade, and the 606 report is where the difference stops being theoretical.

What does SIPC cover if a broker fails?

Restoration of custody: up to $500,000 per customer per capacity, including $250,000 in cash, when a brokerage collapses or client assets go missing. Market losses are never covered, and cash swept to a program bank falls under FDIC rules instead. Large brokers add private excess-SIPC insurance above the statutory limits; the disclosure naming the carrier and ceiling is on each broker’s site.

Which fees survive in commission-free brokerages?

Options per-contract fees, transaction-fee mutual funds, outgoing transfer and wire charges, margin interest, FX conversion on foreign listings, market-data subscriptions, and the sweep spread on idle cash, the largest of them for most accounts. Every one appears in the published fee schedule or the account agreement; commission-free describes one line of a long table, and the table is the document worth printing before opening anything.

How do sweep rates differ from money market yields?

A sweep rate is what the broker chooses to pay on idle cash; a money market yield is what the market pays on short government paper. When the default sweep is a money market fund, the two nearly coincide, near 3.3% to 3.6% in mid-2026; when the sweep is an affiliated bank deposit, the broker keeps the difference, and published defaults run as low as 0.01%. The gap between the two numbers is the broker’s margin on your cash, and it is disclosed, not hidden; the disclosure is simply written where reading is optional.

9. After the broker

The intermediary chosen, the work moves back to what it holds and how the account behaves: the account-level economics that operate identically at any broker, sweep, routing, protection, margin, compounding costs, are the grid of the companion page linked in the introduction, and the full learning sequence around both sits in the full beginner curriculum. The figures on this page carry July 2026 dates and will be replaced each January; the checklist will not, because it maps where the business model lives, and business models outlast rate cycles. A cutting cycle would compress the sweep line mechanically and thicken the margin line; the criteria would not move. A broker chosen on documents can still disappoint; it cannot surprise. That, on a shelf where every storefront looks free, is the entire edge available to a retail chooser.

Last updated — 12 July 2026