How to Read Your Broker After the Payment for Order Flow Ban

There is a date worth circling on the calendar if any of our money sits with a European broker: 30 June 2026. That is the day the EU’s ban on payment for order flow takes full effect under Article 39a of its revised markets rulebook, after a temporary carve-out that only Germany formally claimed. The headline reads like a clean win for the small investor. The mechanism underneath reads like something closer to a relocation.

Payment for order flow (the fee a market maker pays a broker for the right to fill a trade against its own inventory, rather than the order going to a public exchange) is what let a generation of neobrokers offer trading at one euro or zero. The regulators in the Netherlands and Spain looked at the fills it produced and did not like them. So the EU is removing the payment. The question I keep coming back to is narrower and more useful than “is this good”: after 30 June, where does an order actually go, and how would we know if it got a worse price?

What we know about the post-ban routing

Start with the revenue at stake, because revenue is what shapes a routing decision. Payment for order flow accounted for less than 30% of Trade Republic’s revenue by the firm’s own admission, a meaningful slice that does not vanish just because the rebate is banned. DEGIRO, by contrast, drew only about 3.2% of revenue from it as of late 2021, which is why DEGIRO treats the ban as a tailwind and Trade Republic treats it as a problem to engineer around.

The engineering is the part to watch. Scalable Capital already routes its orders through its own exchange, the European Investor Exchange, where the broker operates the venue and assigns one specific market maker per instrument. Trade Republic’s subsidiary obtained a license from the German regulator in January 2026 to run its own trading venue, which lets it match orders internally and act as a market maker itself. eToro states plainly that it is the sole execution venue for its trades. The common shape: the broker stops being paid by an outside venue and becomes the venue.

The independent execution studies on this structure are not reassuring. The Dutch and Spanish regulators measured what happens on single-market-maker venues, and the numbers travel.

The Dutch regulator found price deterioration of 1.44 to 3.46 euros in 68% to 83% of cases on 3,000-euro trades. After the United Kingdom banned the practice in 2012, one study found best execution in 90% of cases afterward against 65% before. The German regulator found the worst effects on non-index securities above 500 euros and index names above 2,000 euros, which is to say the trades large enough to matter.

The ban targets the payment, not the structure

Here is the part that surprised me when I traced it through the rulebook. The ban is written against the payment, not against the structure. Article 39a prohibits a firm from receiving any fee, commission, or benefit from a third party for routing client orders to a particular venue. A third party. When the broker owns the venue, there is no third party and no payment changes hands, so there is nothing to receive.

What was an explicit, disclosed, per-order rebate from an outside market maker becomes an implicit margin captured inside the broker’s own exchange, as the bid-ask spread its single assigned market maker quotes. The broker no longer collects a few cents from someone else. It keeps the whole spread itself. And the incentive to route an order to its own venue is now stronger than the old incentive to route to whichever outside venue paid the most, because before, the broker shared the economics with a counterparty, and now it owns all of them.

That is the real shift. A rebate is a number a regulator can demand and publish. A spread captured on a broker-owned venue is embedded in the fill price, and no retail statement breaks out how much of that spread was the genuine cost of liquidity versus the broker’s margin. The conflict got harder to see at the exact moment it got bigger.

The ban targets the payment. The broker kept the venue. So the conflict did not die, it just stopped sending an invoice.

This is not only a European footnote. On 17 June 2026, Kraken shipped an open-source connector that lets AI agents place live orders through tools like Claude and ChatGPT, the latest in a wave of broker-and-agent plumbing that now reaches into real accounts. An agent that places an order does not read the venue’s rulebook or compare the spread to a competitive quote. It submits, and inherits whatever execution the broker’s own venue gives it. The more of our order flow gets initiated by something that cannot see where it lands, the more the only defense left is the structure itself, and the structure just got more conflicted.

Find out where your broker routes after the ban

The point is not to flee any broker that runs its own venue. Plenty of perfectly good fills happen on them, and for a 200-euro savings-plan purchase the spread cost is genuinely small. The point is to know what to look at, and to size the question to the trade.

1. Find out where the broker routes after 30 June

   Check whether the broker uses an outside exchange or its own affiliated venue with a single assigned market maker. The venue name sits in the order confirmation and the terms. If it is the broker's own exchange, that is the structure the regulators flagged.

2. Compare the fill to a reference price on larger trades

   For any order above roughly 2,000 euros, note the bid and ask on a neutral source at the moment of the trade, then check the fill against it. One comparison a month tells us whether the spread being paid is competitive or padded.

3. Use limit orders where the spread is the cost

   A market order accepts whatever the single market maker quotes. A limit order caps the accepted price, which is the cleanest defense against a wide spread on a broker-owned venue. The cost is the occasional unfilled order, which on a non-urgent trade is no cost at all.

4. Match the broker to the trade size

   The studies show the spread damage concentrates in larger and less liquid orders. A single-venue neobroker can be fine for small recurring buys and a poor choice for a 20,000-euro position in a thin name. Splitting brokers by trade type is a defensible move.

5. Keep an agent on a short leash near execution

   An AI tool that places orders should be held to limit orders and a per-trade size cap, and its fills read for venue the same way we read our own. The agent will not check the spread. We have to.

What stays with me is the shape of the trade-off. We spent years being told the free trade was the catch, that the rebate was the hidden price. Now the rebate is going away and the catch is moving into the one number no statement breaks out. The most honest thing I can say about 30 June is that the cost of trading did not fall. It just got quieter, which is a harder thing to watch than a fee.