CHI Company Investigation · Finance · Multi-product consumer-finance ecosystem
Robinhood Markets, Inc.Brokerage · Options · Crypto · Margin · Cash · Banking · Cards · Retirement · Menlo Park, California
Ten research streams · Verdict: weakly supports · Band: Concerning · No numerical score assigned
Robinhood made investing free and then spent six years learning how to say what free cost.The app got safer; the model got wider.
This investigation was commissioned to test a single proposition: whether a company that removed the most visible price in retail investing replaced it with costs and conflicts the customer could not see. The answer is neither of the two available slogans. It is both, sequenced. The severe case against Robinhood is real, quantified by regulators, and belongs to 2015–2021. Most of it was fixed, with dates. What survives into 2026 is narrower and different in kind — and it is not deception.
Then 2015–2021
Serious failures, established by regulators and quantified in dollars.
A deleted disclosure and a training instruction to avoid the subject. Execution that internal analysis said lagged competitors “by a wide margin” while the marketing said the opposite. Buying-power displays that showed some customers double what they owed. Options approval run by bots. A platform that went dark for a full trading day with no telephone line to call.
$34.1m of customer harm found net of the commissions customers saved. $12,598,445.16 of restitution ordered in a single settlement.
Now 2025–Sep 2026
Safer, broader, more bundled — and harder to leave.
Execution is independently reported and competitive. Margin is priced at 5.00% where a major incumbent charges more than double. There is a broker-funded retirement match no competitor offers. No regulatory order sanctions conduct that began after 2022, and federal enforcement has been silent for eighteen months.
What is left is structural: yield behind a subscription, a subscription that has become a hub, $100 to leave, five-year strings on the bonuses, and revenue that grows with transaction intensity.
Read both columns before reading anything else on this page. CHI does not allow later remediation to erase established conduct, and it does not allow established conduct to be reported as though it were still happening. Everything below is dated for that reason.
The opening frame
Robinhood is no longer a stock-trading app, and it has not been one for years.
Assessing it as a brokerage would miss most of what it now does with a customer’s money. By its own description it is a “financial superapp”, and by revenue it is a transaction business sitting on top of a balance sheet funded by customer cash. The contemporary ecosystem a single customer can hold:
The longitudinal question: did Robinhood simply democratise financial access — or did the apparently free surface create customer costs and conflicts somewhere else?
The answer this investigation reaches is both — but at different times, and through different mechanisms. The mechanisms that produced the severe findings of 2015–2021 are not the mechanisms operating in 2026. Reporting them as one continuous case would be inaccurate. Reporting only the current one would be incomplete.
The anchor exhibit · 2015–June 2019
The strongest evidence on this page is not GameStop. It is a deleted FAQ.
In December 2020 the SEC entered an order against Robinhood Financial concerning how it described the way it made money, and how well it executed customer orders. It is the single most important document in this investigation because it is the only regulatory finding that goes directly to the proposition CHI was asked to test: that the price of “free” was moved somewhere the customer could not see it.
SEC Order 33-10906 · 17 December 2020
Between late 2016 and September 2018, Robinhood’s “How Robinhood Makes Money” FAQ omitted payment for order flow — then its largest revenue source. The order records that the firm deleted the payment for order flow FAQ
and kept no records showing when… why… or who was responsible for approving its removal.
Customer-service training materials explicitly instructed [representatives] to ‘avoid’ talking about payment for order flow
and stated that it was incorrect
to name it in answer to the question of how the company made money.
A principal trading firm told Robinhood that large retail brokers typically receive four times as much price improvement
relative to the payments they accept. Robinhood explicitly offered to accept less price improvement for its customers in exchange for higher payments to itself.
Internal analysis in March 2019: no matter how we cut the data, our % orders receiving price improvement lags behind that of other retail brokerages by a wide margin.
An October 2018 benchmarking exercise had already found that most of Robinhood’s metrics were worse.
From October 2018 to June 2019 — while those analyses sat internally — Robinhood publicly claimed its execution quality matched or beat that of its competitors.
Harm, as found: customers were deprived of approximately $34.1 million in price improvement between October 2016 and June 2019 — after netting off the commissions they would have paid elsewhere. The shortfall concentrated in orders above 500 shares. Civil penalty: $65 million.
Keep the characterisation precise. This was settled without Robinhood admitting or denying the findings, and the charges were negligence-based — Securities Act § 17(a)(2) and 17(a)(3). No regulator has ever entered a scienter or fraud finding against Robinhood. The conduct in the order does not need to be upgraded to be serious: a company deleted the disclosure of its largest revenue source, told staff not to raise it, and marketed an execution quality its own analysts had told it was not true.
Then and now, on the same measure
CHI credits a correction in full when describing a customer’s present experience. On the specific thing the SEC penalised — execution quality — the evidence says the problem was real and so was the remediation.
Execution measurably behind peers, while marketing said otherwise
The private class action on the same facts was denied class certification and settled for $2m in June 2025 — a proof problem in litigation, not a finding that the harm was smaller.
Independently compiled, competitive, and no equivalent current finding
Statistics compiled by S3 Matching Technologies, stated to be unaffiliated with Robinhood. In a peer-reviewed five-broker field test using simultaneous orders placed December 2021–June 2022, Robinhood ranked third of five, ahead of the no-payment-for-order-flow broker, which placed last; the authors concluded that variation in payment for order flow does not explain the large variation in price execution between brokers.
A boundary this page will not cross. Robinhood’s current order-flow economics are not hidden. Its Rule 606 disclosures state the terms to the cent: a fixed 12.35% of the spread during regular market hours, 9% in extended hours, per-contract options payments of $0.30–$1.20 on single-leg orders by spread tier, and no payments at all on index options. Payments from market makers were 55% of total net revenue in FY2025, with Citadel Securities at 13% and Wintermute at 6%. The current criticism in this investigation is about the words used in the plain-language layer — see framing — not about concealment.
Interface as harm · 2016–2021
The second finding is about what the screen told people.
In June 2021 FINRA issued what was then its largest-ever financial penalty against a member firm. Robinhood consented to the findings without admitting or denying them; the underlying facts are not disputed. Where the SEC order was about a disclosure that was removed, this one is about information that was displayed and wrong.
FINRA AWC No. 2020066971201 · 30 June 2021
Margin, misrepresented. Approximately 818,000 options-approved customers were told margin could be “disabled” on their account, while options spreads could and often did automatically trigger the use of margin anyway.
Balances, wrong. More than 135,000 customers were shown inaccurate cash balances between December 2019 and June 2020. Separately, roughly 4.2 million customers were shown negative buying power at double the true figure between September 2016 and September 2020.
Options approval, automated. Approval ran through “option account approval bots” with principals reviewing fewer than 0.1% of applications. A single internal review found around 3,200 accounts held by people under 21 who claimed three or more years of options experience, and 5,350 customers approved for Level 3 when the firm’s own criteria suited them only to Level 2. Rejected applicants could re-apply mere minutes
later with changed answers and be approved.
Outages. On 2–3 March 2020 the platform went down for all of Robinhood’s customers (12.5 million accounts at the time)
. During it, the firm had no live customer service telephone line.
Complaints, unreported. Between January 2018 and December 2020 tens of thousands
of written customer complaints went unreported to FINRA under a firm policy that exempted margin-call complaints, complaints the firm judged to lack merit, and crypto complaints.
Sanction. A $57 million fine plus $12,598,445.16 in restitution: $5,731,520.67 for options-spread losses, $1,653,366.51 for erroneous margin calls, and $5,213,557.98 for outage losses.
Alex Kearns
The display defect described above had one consequence that cannot be described in aggregate. In June 2020 a twenty-year-old Robinhood customer, Alex Kearns, saw a balance of −$730,165.72 on his account. The regulator later established that the true figure was −$365,530.60 — the screen was showing him double what he actually owed. He sent three requests to support overnight and received automated replies; there was no telephone line to call. He died by suicide, leaving a note that Robinhood itself described as citing confusion with its product. The family’s wrongful-death claim was settled in 2021 on undisclosed terms.
CHI records this once, and draws from it only what the evidence supports. One death does not establish a statistical pattern, and this page does not use it as one. The systematic finding is the one the regulator made: a defective interface and control system that displayed incorrect information to more than 135,000 customers, told 818,000 people margin was off when it was not, and offered no human being to reach when it mattered. Kearns is the point at which the cost of that system stopped being measurable in dollars.
What changed, with dates. Balance and buying-power displays were corrected during 2020. Options eligibility criteria were revised in September 2020 and principal review was raised from roughly 20 to about 500 applications a week by May 2021. Level 3 options now require a margin account. Live telephone callback support launched on 5 October 2021. No comparable finding on margin disclosure or options approval appears in any regulatory order issued in 2024 or 2025. Whether immediate re-application with changed answers is still possible could not be verified from public sources and is recorded as unverified.
January 2021 · Three different questions
The meme-stock week is not one story, and CHI does not score it as one.
Most accounts of late January 2021 collapse a clearing-system event, a capital-adequacy failure and a communications decision into a single accusation. They are three separate questions with three different answers, and only one of them is a customer-hostility finding. This episode is important. It is not the thesis of this page.
A · The restriction itself
A clearing call the firm did not set and could not refuse
At approximately 5:11 a.m. ET on 28 January 2021, the National Securities Clearing Corporation notified Robinhood Securities of a deposit deficit of approximately $3 billion — a $1.3bn value-at-risk charge plus a $2.2bn Excess Capital Premium, against $696m already on deposit.
Robinhood restricted buying in a set of volatile symbols to position-closing-only. After 9 a.m. the clearing house waived the Excess Capital Premium; the net requirement fell to about $1.4bn and Robinhood deposited a further $737m. That day the DTCC waived $9.7bn of requirements across six member firms — this was an industry event, not a Robinhood-only one.
SEC staff subsequently attributed broker restrictions in that week to margin and capital charges imposed by the clearing house. Customers could always sell.
Regulatory / infrastructure driven — not scored as customer hostility
B · Capital and risk preparedness
Being in a position to receive that call was a company choice
The Congressional review of the episode found that Robinhood exhibited troubling business practices, inadequate risk management, and a culture that prioritized growth above stability
during the event.
Its collateral models did not account for the prospect of Excess Capital Premium charges at all — the single largest component of the bill it received. And the reprieve was not its own: Without this waiver, which Robinhood had no control over, the company would have defaulted on its regulatory collateral obligations.
A later regulatory settlement separately addressed supervision of the clearing system that struggled through the same period.
Proven operational failure — historical, remediated
C · Communication
The only part of this week that is a CHI finding
On 27 January, as the restrictions were being designed, an internal exchange addressed what customers would be told. A manager wrote: The real reason is firm risk and us needing to control the velocity of trading.
And then: But we shouldn’t expose that.
The Chief Operating Officer agreed that customers should not be given the real reason.
Publicly, in the same period, the company’s messaging was that there was no liquidity problem and that the restrictions had been imposed preemptively. Within four days Robinhood had raised $3.4 billion.
This is the residue: a documented decision to manage what customers knew, at the moment it mattered most to them.
Isolated — documented opacity under stress
Evidentiary boundaries — all of these are part of the finding
- Isolated. No comparable episode has been identified in the five volatile years since.
- Never adjudicated as fraud. No court or regulator has found the January 2021 statements fraudulent.
- No scienter finding has ever been entered against Robinhood, in this matter or any other.
- Self-corrected within days. A second statement the same day did cite clearing-house deposit requirements, and the approximate size of the call was disclosed publicly within seventy-two hours.
- Customers lost in court. Every fully litigated customer claim — negligence, breach of contract, fiduciary duty — was dismissed and affirmed on appeal; the customer agreement permitted the restriction.
- The conspiracy theory was rejected. Claims that Citadel Securities, hedge funds or other market participants directed the restrictions were dismissed at the pleading stage twice and affirmed on appeal. CHI does not publish that allegation, and neither the Congressional review nor SEC staff found it.
The three eras
The same company, assessed three times.
CHI’s verdict is not an average. It is a sequence. Each era below was tested against the same proposition and returned a different answer, and the direction of travel is the most important single fact about Robinhood.
2015 – 2021
Supports
- Payment-for-order-flow disclosure deleted; staff instructed to avoid the subject; execution marketed as matching competitors while internal analysis said it lagged “by a wide margin”.
- $34.1m of harm found net of commissions saved; $65m penalty.
- False buying-power and balance displays to 135,000+ customers; margin represented as off when options spreads triggered it for 818,000.
- Options approved by bots with under 0.1% principal review.
- A full-day outage across 12.5m accounts with no telephone line.
- January 2021: capital and modelling failure, and a decision not to give customers the real reason.
Severe, regulator-quantified, and not in dispute on the facts.
2022 – 2024
Weakly supports
- Remediation lands. Gamified features removed by January 2022; suspicious-activity backlog cleared; identity-theft programme rewritten; improper transfer-rejection categories ended July 2022 — each of these before the order that sanctioned it.
- Compliance debt persists. Short-sale marking failures ran to December 2023; regulatory blue-sheet errors to April 2024; unsupervised paid-influencer marketing to March 2023.
- Exit is repriced: the outgoing transfer fee rises from $75 to $100 in July 2022.
- The subscription begins to become a hub: retirement match, then the Gold Card.
Growth funded ahead of controls — being paid down.
2025 – September 2026
Weakly supports
- No regulatory order sanctions conduct that began after 2022. Federal enforcement has been silent for eighteen months; a crypto investigation closed with no action.
- Yield, banking, the card and the retirement match sit behind a $5 subscription; the unsubscribed cash rate is not published.
- $100 to transfer out — on partial as well as full transfers.
- Five-year strings on retirement bonuses; a foreign-transaction fee added to new cards without announcement.
- Revenue concentrated in options, event contracts and crypto.
Narrower, structural, disclosed — and growing.
Read the three columns as a trajectory rather than a total. The 2015–2021 column contains everything that would justify a severe verdict; almost none of it describes the product a customer opens today. The 2025–2026 column contains nothing that would justify a severe verdict on its own; all of it is disclosed and lawful. Weakly supports is the only reading that respects both columns.
The cleanest current finding
The broken exit process was fixed. The expensive exit remains.
Of everything still live in 2026, this is the finding that rests most squarely on a primary document and needs the least interpretation. It is also the clearest longitudinal exhibit on the page, because the same mechanism — leaving — produced a historical failure that was corrected and a current cost that was increased.
Charged to transfer assets out — on partial transfers as well as full ones
Robinhood’s fee schedule and help centre both state it plainly: a $100 fee for partial and full outgoing account transfers. The fee applies to retirement accounts too.
Charging the full amount on a partial transfer is the part that stands out. It means a customer cannot move part of a portfolio elsewhere — to test another broker, to diversify custodians, to consolidate a spouse’s accounts — at any lower cost than leaving entirely. Among large brokers this is a minority practice. The one comparator verified directly for this investigation, Merrill Edge, charges $49.95 on a full transfer and nothing on a partial one.
The fee rose from $75 to $100 with effect from 1 July 2022. No cost change was disclosed as its basis.
116,000 outbound transfer requests improperly rejected
Between November 2018 and July 2022, roughly 116,000 requests to transfer assets out were rejected improperly — for unsettled trades, for crypto held at the affiliate, for options expiring within five days — and the rejected transfers were never reviewed. This is the only exit friction in the record that a regulator has actually found. It ended in July 2022 and no restitution was ordered.
The price of leaving, applied symmetrically to partial and full exits
Disclosed, lawful, and at the top of the retail range. Its weight falls hardest on small accounts: the last median account balance Robinhood disclosed was approximately $240, in 2021, and no later median has been published.
Inbound transfers are subsidised
Robinhood reimburses up to $75 of the other broker’s exit fee on transfers in of $7,500 or more. The asymmetry is ordinary competitive design rather than deception — but it is an asymmetry: the company pays to attract and charges to release.
Mechanics that apply at every broker
Fractional shares are sold rather than transferred, and crypto cannot travel by the industry transfer system at all. These are constraints of the clearing infrastructure, not Robinhood choices, and CHI does not score them. On-chain crypto withdrawal has been available since 2022, so the liquidation is avoidable.
Pattern: Exit Resistance. The CHI test for this pattern is asymmetry between joining and leaving — and here the asymmetry is priced rather than procedural. Nothing about the cancellation process is obstructive; the friction is the invoice.
From subscription to hub
Gold costs $5 a month. That is not what makes it interesting.
Robinhood Gold is $5 per month or $50 per year, with a 30-day trial, and roughly 4.84 million subscribers — 17.0% of 28.4 million funded customers. The price has not risen since 2019 while the benefit list has grown to around a dozen items. The CHI finding is not that Gold is bad value. For a customer who uses it, it is frequently excellent value. The finding is what has been attached to it, and what happens at the moment of cancellation.
or $50 / year One subscription now sits upstream of a bank account, a credit card and a retirement bonus.
What the architecture does
- A customer using three or four of these is no longer deciding whether $5 a month is worth it. They are deciding whether to unwind a bank account, a credit card and a retirement bonus.
- Card rewards redeem at full value only into Robinhood accounts, and are forfeited on closure.
- The subscription agreement reserves the right to “add or remove features” at Robinhood’s sole discretion, with 30 days’ notice required only for price increases.
- The company describes the loop in its own words: “cash back flows back into brokerage.”
What the architecture is not
- Not a trap, and not concealed. Every term above is published; the cancellation flow itself is two taps where no Gold Card is attached.
- The price has been unchanged since 2019 while roughly ten benefits were added, and there is a 30-day trial.
- The retirement match is real money: more than $500 million has been paid to customers, and no other major US broker funds one at all.
- Gold subscribers hold materially more assets than the average customer — consistent with a bundle that attracts engaged customers, not only one that holds them.
Patterns: Dependency Stack and Exit Resistance. The finding is the dependency architecture — the cascade a cancellation now triggers — not the value of the subscription.
Lexicon candidate
Yield Gating
This investigation identified one mechanism that the existing CHI lexicon does not describe well. It is presented here as a candidate, not as an adopted entry.
Yield Gating
A financial platform conditions attractive yield on paid membership while the unsubscribed customer’s return is materially lower, zero, or not prominently published — allowing the subscriber yield to function simultaneously as a financial return and as a subscription-acquisition mechanism.
Why Robinhood is the current type specimen
- Gold cash yield of 3.35% APY on eligible brokerage cash.
- No non-Gold programme rate is published anywhere on Robinhood’s own pages; the help centre describes the programme as one for Gold subscribers.
- Robinhood Banking: “Both checking and savings earn 0%” without Gold, and the headline 3.50% additionally requires monthly direct deposit.
- The rate is marketed as “more than eight times the national savings average” on a page that does not state what a customer earns without the subscription.
- The subscription is therefore required for the headline economics of a product whose input is the customer’s own money.
What the candidate does not claim
- It does not claim an exact unsubscribed rate. Robinhood publishes none, and CHI will not print a third-party figure as if it were the company’s own.
- It does not claim the Gold rate is uncompetitive. Through the rate-cutting cycle it tracked within roughly 15 to 25 basis points of the bottom of the federal funds target at each dated point.
- It does not claim tiered pricing is inherently hostile. Tiering a feature is ordinary. The question this candidate isolates is what changes when the tiered good is the return on money the customer already holds.
- It does not yet claim to be a pattern. One company is a specimen, not a pattern.
Status: candidate. Yield Gating is not an adopted CHI lexicon entry and does not appear in the Lexicon. Adoption depends on whether the same mechanism is found operating independently at a second company under a different regulatory regime. The forthcoming Revolut investigation — a tiered consumer-finance platform where savings yield rises with plan price — is the intended test. If the mechanism does not travel, the candidate is withdrawn.
Current transparency
Framing, not deception.
This distinction is the most important editorial judgement on the page, and CHI holds it firmly. The evidence does not support calling Robinhood deceptive today. Its fee schedule is published and dated. Its order-routing economics are disclosed to the cent in regulatory filings. Its retirement clawback appears in the FAQ rather than only the footer. What the evidence does support is a recurring house style in which the attractive headline is technically accurate and the economically important condition is relegated.
The plain-language help page describes the mechanism accurately but does not use the industry term — payment for order flow — anywhere. The Rule 606 filings disclose the rates in full. This is the same plain-language layer the 2020 order penalised; the euphemism survived the remediation.
Wording · not concealmentAccurate at the card level. To earn the headline 3% rate the Gold Card requires an active annual Gold subscription, and the annual plan is the only accepted payment option for cardholders. The card has no fee; holding it is not free.
Most aggressive construction foundTrue of the 3.35% Gold rate. The same page does not state what a customer earns without the subscription — and Robinhood publishes no such rate anywhere.
Condition relegatedPresented beside a table listing a $25 outgoing wire. Technically defensible — a wire is a transaction charge, not an account fee — and exactly the construction consumer regulators have criticised in bank marketing generally.
Category framingSaid of a February 2026 change that moved over $6 billion of swept cash on-balance-sheet to help fund the margin book. The rate paid was unchanged, which is what the statement asserts. The coverage regime on the first $10,000 changed with it, which the statement does not mention. The change is permitted broker economics practised across the industry.
Accurate on price, silent on structureA live help page still quotes a 3.5% margin rate against a fee schedule that has said 5.00% since December 2025. The Gold page quotes an instant-withdrawal fee of 1.5% where the fee schedule governs at up to 1.75%. A retirement page headline says $210 where its own body says $225. Note the direction: the stale margin figure understates the price. This is disclosure hygiene, not deception.
HygieneEditorial label: Headline Maximization / Condition Relegation. Every statement above is lawful and most are literally true. Read individually, none would support an adverse finding. Read together, across six years and five product lines, they describe a consistent choice about where the qualifying condition goes — and that consistency is what CHI records. It is a Trust and Transparency observation about the current company, and it is materially less serious than what the 2020 order found.
Direction of travel
One clean example of a term quietly getting worse.
The Gold Card foreign-transaction fee, 1 July 2026
The Gold Card was announced in March 2024 with “no annual fee and no foreign transaction fees” as headline positioning. Accounts opened on or after 1 July 2026 carry a 3% foreign transaction fee. Cardholders who opened before that date were grandfathered at the time this was examined. The change was not the subject of a public announcement; it surfaced through support and legal materials.
Kept in proportion: this is not a large customer-harm event. It is disclosed at application in the standard credit-card terms box, existing customers were protected, and repricing a card benefit is lawful and ordinary. CHI includes it for one reason only — it is the cleanest available instance of a marketed benefit being withdrawn quietly, and it is therefore a useful specimen of Quiet Rescission / Quiet Term Degradation rather than a finding of weight in its own right.
Two adjacent, milder instances point the same way: the Cash Card round-ups feature was discontinued in December 2025, and Robinhood’s public status page has been retired in favour of a social-media account — a transparency step backwards no comparable broker has taken.
Structural incentive
What Robinhood earns from, and why that is a question rather than an accusation.
Robinhood’s revenue remains heavily transaction-dependent, and the transaction line is concentrated in three product classes. This is a matter of public record in its own results, and it is the single most durable structural observation in this investigation.
Transaction-based revenue was $776m of the quarter’s $1.308bn. Across the 2025 financial year the same shape held: transaction revenue of $2,628m against total net revenue of $4,473m, with options at $1,123m and crypto at $901m — and equities, the product class where the independent execution-quality evidence is strongest, the smallest of the three at $302m. Subscriptions were $179m, about 4% of revenue.
The correct CHI formulation, and the limit of it: Robinhood economically benefits when customers transact more heavily in product classes where independent literature identifies weaker average retail outcomes. That is an incentive-alignment question, and it is a real one — options contract volume grew roughly 50% year over year against funded-customer growth of about 7%, so intensity per customer is rising, and the product roadmap consistently lowers friction for those classes.
It is not a finding that Robinhood intentionally causes customer losses, and this page does not say so. The loss literature is asset-class evidence, not Robinhood-specific; no post-2021 study measures Robinhood customers directly; and the same tooling that lowers friction — profit-and-loss charts, probability estimates, simulated returns — is precisely what critics said the platform lacked in 2020. Equity options carry no commission and a four-cent regulatory pass-through, against sixty-five cents a contract at major incumbents. Structural incentive is the label. Intent is not inferred.
Behavioural design · 2018–2020, and after
The confetti era was real. It also ended, before anyone ordered it to.
The most familiar criticism of Robinhood is that it turned investing into a game. For a defined period, the documentary record supports that criticism. For the current product, it does not — and the honest answer about today is that the evidence is inconclusive, not that it is exonerating.
Documented mechanics, and peer-reviewed evidence of harm to the customers who followed them
Confetti on completed trades. A scratch-off card to reveal a free stock. A waitlist you could tap up to a thousand times a day to move up. Push notifications steering customers into curated lists — including a “Top Movers” list that, unusually, blended gainers and losers into one ranking.
Peer-reviewed, published in the Journal of Finance, using data through August 2020.
Every named feature removed — and removed ahead of the order
The removals preceded the settlement by two to three years. Robinhood neither admitted nor denied the engagement findings, and no regulator has ever found that its engagement design violated a rule. The state’s highest court upheld the validity of the fiduciary regulation at issue — not the charge against Robinhood.
Why current behavioural harm is inconclusive rather than cleared. The peer-reviewed evidence of herding losses ends in August 2020, and no equivalent study of the post-2022 product exists. Published work from the same period also found that the aggregate Robinhood portfolio did not underperform and did not panic-sell in the March 2020 drawdown — the average holding was fine; the extreme herding episodes were not. Today’s tools are materially different in kind: information-dense rather than game-like, and carrying loss warnings on every page. CHI will not use 2018–2020 gamification as proof that the 2026 product is behaviourally manipulative, and it will not declare the question closed in Robinhood’s favour either. One residual is noted: the animation that replaced trade confetti now fires on signing up for Gold — a paid-product nudge rather than a trading one.
Restrictions and support
The supported finding is about recourse, not about freezing.
Account restrictions generate the largest volume of customer complaint about Robinhood, and CHI has deliberately kept this section subordinate — because the evidence supports a narrow claim and not the broad one.
What is supported. Robinhood names the categories that trigger a restriction, but publishes no expected duration, no service-level commitment and no formal appeal route. A restriction on the brokerage account also blocks crypto purchases at the affiliate. And a customer locked out of the app must authenticate in order to open a support ticket — the recurring shape of the complaint record is not that money disappears, but that the person cannot find out why, or when it ends. Recurring themes in the public complaint databases are consistent on this point across 2024–2026.
What is not supported, and CHI does not assert. No regulator has found that Robinhood freezes accounts improperly, or more often than the law requires. The contractual powers involved are matched almost clause for clause in the customer agreements of the largest incumbent brokers, and a court has held Robinhood’s agreement not unconscionable. Most importantly, the causation runs the other way from the intuitive reading: in 2025 both the SEC and FINRA sanctioned Robinhood for under-verifying customer identity and missing account takeovers. Much of the friction customers meet in 2026 is the mandated correction of that failure, and it has a legitimate regulatory explanation.
Two figures, and what they can and cannot show. Complaints recorded against Robinhood in the federal consumer database rose from 314 in 2023 to 950 in 2024 and 1,917 in 2025, with 1,153 recorded so far in 2026 — against a customer base of 28.4 million and a materially expanded product range. Over the same period the company’s timely-response rate rose from 49% in 2020–2022 to 96.7% in 2026 to date. No per-account normalisation exists for any broker, so CHI does not claim Robinhood is an outlier in either direction, and readers should not infer one from raw counts.
Watchlist — regulatory-contested
Prediction markets: the fastest-growing line, on unresolved legal ground.
What CHI is and is not saying
Event contracts produced $156 million in the second quarter of 2026 on 13.6 billion contracts traded — a line that now out-earns both equities and crypto. Commission is charged at 10% of p × (1 − p) per contract, halved for Gold members, plus a one-cent exchange fee. Robinhood has also acquired its own exchange and clearing house, which will place it on more sides of the same customer order than any comparable US broker.
CHI does not call these contracts illegal gambling and does not resolve the legal dispute. The question is genuinely contested: one federal appeals court has ruled for the exchange model and another against it, a federal regulator has proposed rules that would permit most broad-outcome sports contracts, and Robinhood is a defendant in several state actions while having sued pre-emptively in others. That fight is about jurisdiction, not about deception.
The narrower CHI-relevant question is this: Robinhood is scaling a speculative product line rapidly, and the protective scaffolding around it appears thinner than the safeguards conventional gambling operators are required to provide. The per-match landing pages carry no age or state-eligibility statement, no problem-gambling resources, and no deposit or loss limits or self-exclusion tools were located. The company’s own educational article names compulsive behaviour as a signal to stop, but links to no external help.
Classification: regulatory-contested, current, watchlist. This is an emerging risk item, not a finding of weight, and it does not move the verdict.
The case for Robinhood
Why this verdict is only weakly supports.
CHI requires every investigation to build the strongest available case in both directions. On this page the defence is not a courtesy paragraph — it is the reason the verdict sits where it does. A great deal of what Robinhood has done to its customers has been unambiguously good for them, and some of it has been good for customers of every other broker as well.
It helped force commissions to zero across the industry
In the first week of October 2019 the largest incumbent brokers eliminated online stock commissions within days of one another, each citing competitive pressure from low-cost mobile entrants. Taken from their own revenue guidance at the time, the change returned on the order of $1.5–1.7 billion a year to retail investors from three firms alone — an inference from those disclosures, but a well-founded one. That benefit went to customers who never opened a Robinhood account.
It brought in customers the industry had not served
In its listing prospectus Robinhood reported that over half of customers who funded an account between 2015 and early 2021 said it was their first brokerage account. Independent research on the 2020 cohort of new investors found them younger, lower-income and more racially diverse than experienced investors, with a third holding balances under $500 and the most-cited motivation being the ability to invest small amounts. Over the same period, national survey data recorded direct stock ownership rising from 15% to 21% of families — the largest change on record.
Its current visible prices are frequently the best available
- Margin at 5.00% on the first tier, uniform for subscribers and non-subscribers alike — against 11.825% at a major incumbent for balances under $25,000.
- Equity options at no commission plus a four-cent regulatory pass-through, against 65 cents a contract at large incumbents.
- Regulatory fees waived on equity sales of $500 or less and on sales of 50 shares or fewer — a concession aimed squarely at the small accounts, which peers do not make.
- No inactivity, maintenance, minimum-balance, returned-payment or incoming-wire fees.
The retirement match has no competitor
A 1% match on retirement contributions for every customer and 3% for Gold subscribers, with more than $500 million paid out. No other major US broker funds a contribution match at all. The five-year holding condition is a real tether — and it is disclosed with worked examples, the fee never exceeds the match, and gains can be withdrawn without triggering it.
Cash yield tracked the Fed honestly
Through the rate-cutting cycle the Gold sweep rate stayed within roughly 15 to 25 basis points of the bottom of the federal funds target at each dated point — 5.0% when the floor was 5.25%, 3.35% when it was 3.50%. There is no evidence Robinhood widened its own take as rates fell. Deposits carry pass-through federal insurance, and Robinhood Banking pays a competitive rate on checking.
It offers an opt-out from the thing it is criticised for
Customers can route orders to exchanges rather than to wholesale market makers, per order. No legacy payment-for-order-flow broker offers that choice. Index options carry no order-flow payment at all, and every order on the desktop platform goes to an exchange. In the United Kingdom, where the practice is prohibited, Robinhood runs the same zero-commission model without it — evidence that the model does not depend on it.
The remediation was real, and often preceded the order
Gamified features were removed company-wide up to three years before the settlement that recorded them. The suspicious-activity backlog was cleared and the identity-theft programme rewritten before the order that sanctioned those failures. Improper transfer-rejection categories ended in July 2022, nearly three years before the settlement citing them. Restitution was actually paid — $12.6m in 2021, $3.75m in 2025, plus private settlements.
The legal record has gone Robinhood’s way
No scienter finding has ever been entered against the company; every SEC charge has been negligence-based or a books-and-records matter. Every fully litigated customer claim arising from January 2021 was dismissed and affirmed on appeal. A crypto investigation closed with no enforcement action. No consumer-protection regulator has brought an action. No regulatory order sanctions conduct beginning after 2022, and federal enforcement has been silent for eighteen months.
Customers keep choosing it, with money
28.4 million funded customers, $369 billion of platform assets, and $75.7 billion of net deposits over twelve months — from customers who can move to a zero-commission incumbent at any time. Every finding on this page was already public while that money arrived. Growth is not proof of welfare, and CHI does not treat it as such. It is relevant revealed-preference evidence, and it is substantial.
Remediation, with dates
Options guardrails announced after the Kearns death: additional Level 3 criteria, an options education specialist, changes to how multi-leg positions and buying power are displayed.
Options eligibility criteria revised; principal review raised from roughly 20 to about 500 applications a week by May 2021; Level 3 moved behind a margin account.
Trade-tied confetti removed company-wide; the scratch-off free-stock reveal followed on 5 April.
Live telephone callback support launched for all customers, ending the email-only era that the March 2020 outage exposed.
Remote-access tooling blocked and multi-factor authentication made mandatory for new devices after the customer-list breach.
Curated-list push notifications removed — the specific channel the peer-reviewed herding research had identified.
On-chain crypto withdrawals enabled, ending the period in which crypto bought on the platform could not leave it; the improper transfer-rejection categories end in July.
Suspicious-activity backlog cleared, identity-theft programme rewritten, and the short-sale position-calculation defect fixed.
Margin pricing unified for subscribers and non-subscribers and cut to among the lowest posted rates in retail brokerage; cut again to 5.00% in December 2025.
What survives against Robinhood
- A deleted disclosure, a training instruction to avoid the subject, and execution marketed as better than the company’s own analysts said it was.
- $34.1m of harm net of commissions, and $12.6m of restitution for wrong displays, bot-approved options and outages.
- A documented decision, in writing, not to give customers the real reason for a restriction that cost them money.
- A pattern of growth funded ahead of controls across five regulators and roughly $262m of penalties.
- Currently: $100 to leave on partial as well as full transfers, yield behind a subscription, five-year tethers, a hub whose cancellation cascades, and a consistent habit of relegating the condition beneath the headline.
What answers it
- Almost all of the severe findings are historical and remediated, several fixed before any order compelled it.
- No scienter finding, ever; no order covering conduct that began after 2022; eighteen months of federal silence.
- Execution today is independently reported and competitive, and the academic evidence does not link order-flow payment to worse execution.
- Current visible pricing — margin, options, small-trade fee waivers, the retirement match — is frequently the best in the market.
- The company helped force industry-wide zero commissions and brought first-time investors into direct ownership at scale.
Neither column cancels the other, and that is the verdict. A page that printed only the left column would describe a company that no longer exists. A page that printed only the right column would erase conduct that regulators quantified in dollars. Weakly supports is what is left when both are held at once.
The regulatory record
Twenty-three matters, classified by what was actually found.
CHI distinguishes an allegation from a settlement entered without admission, and both from an adjudicated finding. The principal matters, with the resolution type stated:
| Authority | Date | Conduct period | Core finding | Resolution | Sanction |
|---|---|---|---|---|---|
| FINRA | 19 Dec 2019 | 2016–17 | No regular and rigorous review of execution quality | Settled, no admission | $1.25m |
| SEC | 17 Dec 2020 | 2015–Jun 2019 | Order-flow disclosure deleted; false execution-parity claim; $34.1m harm net of commissions | Settled, no admission · negligence-based | $65m |
| FINRA | 30 Jun 2021 | 2016–Feb 2021 | False margin and balance displays; options approval bots; March 2020 outage; unreported complaints | Settled, no admission | $57m + $12.6m |
| NY DFS | 2 Aug 2022 | 2019–21 | Anti-money-laundering and cybersecurity programme deficiencies at the crypto entity | Consent, findings acknowledged | $30m |
| State regulators | 6 Apr 2023 | 2018–Mar 2021 | Outages, options and margin approvals, customer-support systems | Settled, no admission · no willful conduct found | up to $10.2m |
| Massachusetts | 17 Jan 2024 | 2018–21 | Engagement design and supervision; separately, the 2021 breach | No admission on engagement · breach facts admitted | $7.5m |
| California AG | Sep 2024 | 2018–22 | Customers unable to withdraw crypto | Settled, no admission | $3.9m |
| SEC | 13 Jan 2025 | 2018–Apr 2024 | Short-sale marking, regulatory blue sheets, suspicious-activity reporting, identity-theft programme, 2021 breach, off-channel communications | Partly admitted; remainder settled without admission | $45m |
| FINRA | 7 Mar 2025 | 2014–Sep 2023 | Undisclosed repricing of order collars (8.7m cancelled orders); 116,000 improperly rejected outbound transfers; identity-programme and influencer-supervision failures | Settled, no admission | $26m + $3.75m |
| SEC | 21 Feb 2025 | 2023–24 | Crypto listings and custody investigation | Closed — no action | — |
| Pending, not findings: a state attorney-general inquiry into execution quality and order collaring; state examinations of an August 2024 overnight-venue disruption; a federal deposit regulator’s electronic-transfer inquiry; a cash-sweep civil claim that survived dismissal in part and has reached settlement in principle; and the multi-state prediction-market jurisdiction dispute. | |||||
↔ Scroll for the full table on smaller screens
Cumulative regulatory penalties from 2019 to September 2026 come to approximately $262 million — about 1.9% of cumulative revenue over the same period, peaking at 6.8% of revenue in 2020 and running between zero and two per cent since. Revenue figures for 2019–2022 used in that ratio are drawn from annual filings and were not independently re-verified for this page. Every order above sanctions conduct that began in or before 2022. The listing-era securities litigation is investor-facing rather than customer-facing and is deliberately excluded from this assessment.
Product by product
Robinhood is not uniformly hostile, and the variation is the point.
A single verdict flattens a company with nineteen consumer product lines. Assessed individually, several are among the most customer-favourable in this index, several are unremarkable, and the adverse findings cluster in a small number of specific places. This grid is where the “weakly” in the verdict comes from.
Brokerage & equities
Historical adverse record on disclosure and execution. Current execution independently reported and competitive. A framing concern remains in the plain-language layer.
Historical → neutralOptions
Historical control failures, regulator-established and remediated. Very competitive visible pricing today. The current concern is structural incentive, not conduct.
Concerning · structuralCrypto
Cost is embedded in the execution price rather than itemised. Historical withdrawal and control failures were meaningfully remediated, including on-chain withdrawal from 2022.
Concerning · framingMargin
5.00% on the first tier, uniform for all customers, less than half a major incumbent’s rate for small balances. Historical display failures fixed.
Customer-favourableCash sweep
Gold economics are genuinely good and track the Fed honestly. The two-tier architecture — and the unpublished unsubscribed rate — is the concern.
Concerning · tieringRobinhood Banking
Competitive economics with Gold, no monthly fee and no overdraft product. Gated behind an annual subscription, with yield falling to zero without it.
Concerning · gatingCash Card & spending
No debit fees, transfer-error rights restated accurately, industry-standard terms throughout. Round-ups were discontinued in December 2025.
NeutralGold
Potentially excellent economic value at $5 a month, unchanged since 2019. The dependency architecture around cancellation is the finding.
Concerning · dependencyGold Card
Best-in-market headline economics at 3% on everything with no card-level annual fee. Concerns: subscription dependency and the July 2026 foreign-transaction change.
Favourable · watch termsIRA & retirement
The only broker-funded contribution match in the US market, with over $500m paid. Carries an unusually long five-year retention tether.
Favourable · long tetherRecurring investing
Dollar-based, fractional, automated. A pro-saving default of exactly the kind behavioural research recommends.
Customer-favourableTransfers & ACATS
The strongest current adverse finding on the page: $100 to transfer out, charged on partial as well as full transfers, against a $75 inbound subsidy.
Adverse · currentRestrictions
No published duration, service level or appeal route, and locked-out customers must authenticate to ask why. The underlying controls are substantially regulator-driven.
Concerning · recourseSecurity & fraud
Serious historical failures — account takeovers, a customer-list breach, a boilerplate identity-theft programme — substantially remediated before the orders that named them.
Historical → neutralDisclosures & marketing
Regulatory-compliant throughout, with accurate footers and a dated fee schedule. The finding is a consistent habit of headline maximisation.
Concerning · framingSecurities lending
Opt-in, with the trade-offs stated above the fold. The customer’s share of gross lending revenue is up to 15%; comparable peer splits could not be verified.
Neutral · disclosedStrategies (managed)
Cheaper than the standalone robo-advisers above $100,000 for subscribers. Managed retirement accounts are excluded from the match.
NeutralPrediction markets
Fastest-growing revenue line, on contested legal ground CHI does not resolve. Protective scaffolding appears thinner than gambling-industry norms.
Watchlist · contested24 Hour Market
Overnight trading through a third-party venue. An August 2024 venue failure cancelled executed trades; the failure was the venue’s and is under examination.
Neutral · unresolvedLegend, Cortex & agents
Information-dense professional tooling with unusually blunt disclaimers. Third-party AI agents trading without Robinhood supervision is a novel risk surface, not a finding.
Neutral · flaggedLexicon
Which CHI patterns this investigation actually uses.
CHI does not coin terminology to create novelty. Three existing entries describe most of what is live at Robinhood; one new candidate is proposed and one mechanism is deliberately not given a new name.
Applied from the existing lexicon
- Dependency Stack — the Gold hub, where one subscription sits upstream of a bank account, a credit card, a cash yield and a retirement bonus.
- Exit Resistance — the $100 outgoing transfer fee on partial as well as full transfers, and the five-year retention tethers on retirement bonuses.
- Two-Sided Spread Opacity — first named in CHI’s Uber investigation. Order-flow payment and the embedded crypto spread are the brokerage instance of the same mechanism: compensation paid by a third party, set inside the execution price, invisible at the point of transaction. CHI reuses the existing mechanism rather than coining a second name for it.
Proposed, and declined
- Yield Gating — proposed as a candidate only, pending validation against a second company. See the candidate entry above. It is not in the Lexicon.
- Quiet Rescission / Quiet Term Degradation — the July 2026 card change is offered as a second specimen of a mechanism first raised in CHI’s United investigation, not as a new entry.
- Declined: no new term was coined for the hub architecture or the bonus tethers. Dependency Stack and Exit Resistance already describe them, and inventing “Hub Subscription Cascade” or “Tethered Generosity” would add vocabulary without adding understanding.
Reserved
Robinhood ↔ Revolut
This investigation captured eighteen comparator fields against Revolut — account licence, cards, cash yield, brokerage pricing, crypto, transfers, foreign exchange, subscription tiers, rewards, support, restrictions, fee transparency, switching friction, cross-selling, monetisation model and regulatory architecture. The comparison is not published here. The Revolut investigation has since been conducted to the same standard and is published separately; the head-to-head remains a separate piece of work and has not been written.
Four constraints are recorded now so that the eventual comparison is built correctly rather than conveniently:
Binding non-comparability rules, carried forward
- Foreign exchange is not a clean comparison. It is Revolut’s founding product and essentially absent from Robinhood’s US offering. Robinhood must be scored not-applicable, never given a clean sheet.
- Order-flow payment differs by jurisdiction, not by virtue. It is prohibited in the UK and EU. Revolut’s explicit commissions are partly a regulatory consequence; Robinhood’s model is partly a US permission.
- Freeze volumes cannot be compared raw. One is a payments bank carrying mandatory fraud-reimbursement liability; the other is a brokerage. Compare the quality of notice, explanation and recourse — not the counts.
- Crypto pricing must distinguish an explicit percentage fee from embedded compensation, and deposit protection differs across three separate regimes. A fee-percentage comparison flatters one firm; a disclosure comparison flatters the other.
Robinhood made investing free and then spent six years learning how to say what free cost.
The early case against it is unusually strong for a company of its size: documented decisions about what to disclose, execution its own analysts knew lagged, interfaces that told customers things that were not true, controls that had not kept pace with growth, and harm a regulator was able to quantify in dollars.
The current case is different in kind. Robinhood fixed most of what made its early growth dangerous — frequently before anyone made it — and today competes on prices that are often the best available anywhere. At the same time it stopped being one free brokerage. It became an integrated financial ecosystem whose economics increasingly rest on subscription gating, retention tethers, embedded compensation, transaction intensity and an expensive exit.
That is enough to keep Robinhood at weakly supports, in the Concerning band.
It is not enough to call today’s Robinhood deceptive, or broadly hostile.
Reading this beside the rest of the index
Every company here is assessed against the same question. Robinhood is the one where the answer changed over time.
Read it against the investigation where the same embedded-compensation mechanism was first named, against the two lexicon entries that carry most of its current weight, and against the methodology that explains why no number appears on this page.