Betterment Robo Advisor: The Shift to Custom Tax Harvesting

The robo-advisor market tends to describe product updates as revolutions. In practice, most are changes to the machinery behind the interface: a different way to group securities, a new rebalancing rule, or a broader set of explanations layered on top of an automated portfolio.

Betterment's August 3, 2026, launch of its refreshed Custom Portfolios architecture belongs in that category. It is a meaningful change in how the platform handles portfolio construction, drift, and tax-loss harvesting, but it does not erase the basic trade-offs of automated investing. The useful question is not whether the new framework looks more personalized. It is whether it improves the two things that ultimately matter to an investor: after-tax returns and net results after fees.

That requires looking past the interface and into the operating rules. How much flexibility does Custom Portfolios actually provide? What does the 7% drift threshold do to a portfolio in a moving market? Where does Betterment tax-loss harvesting stop? And does the platform's AI layer add analytical value, or simply translate existing portfolio data into friendlier language?

The Evolution of Custom Portfolios: From Flexible to Tailored Architecture

The move from Flexible Portfolios to Custom Portfolios is best understood as a structural shift rather than a philosophical one. Betterment is still offering guided portfolio construction, not a blank canvas. You can adjust the allocation within a framework, but the platform continues to define the investment universe and the operating boundaries.

That distinction matters. "Custom" does not mean that every holding, security type, or trading decision is available for manual control. It means that the investor has more influence over the shape of an automated portfolio while Betterment continues to manage the implementation.

Under the refreshed architecture, the portfolio is organized around security groups that govern drift tracking and automated rebalancing. Each security group can represent a defined portion of the portfolio's intended exposure, allowing the system to compare the actual allocation with its target and decide when a correction is warranted. Tax-loss harvesting operates alongside that framework but applies specifically to eligible ETF positions rather than at the security-group level. The two systems share the same portfolio, but they are governed by different rules.

For the investor, the visible experience may still feel familiar. You are choosing from a menu of ETFs and setting an allocation rather than placing individual trades. The difference appears in the plumbing:

• The portfolio is assessed by security groups rather than only by its top-level allocation.
• Automated rebalancing is governed by a specified drift threshold.
• Tax-loss harvesting is applied to eligible ETF positions within the same account structure.
• Customization is available inside a managed system, not outside it.

That last point is the one most likely to be lost in the marketing language. Betterment is not trying to turn its robo-advisor into a self-directed brokerage account. Its value proposition remains the delegation of recurring decisions: how to monitor drift, when to trade, how to seek losses within the available ETF structure, and how to keep an investor from improvising during a bad market.

More control does not mean full control

There is a practical difference between choosing an allocation and controlling a portfolio. An investor may be able to tilt the portfolio toward or away from particular exposures, but the platform still determines how those exposures are represented and when the system acts.

That can be a benefit. Broad ETFs are easier to monitor, generally easier to substitute within a tax-loss harvesting system, and less operationally demanding than a portfolio built from dozens of individual positions. It can also be a limitation for an investor who wants to harvest losses in specific stocks, maintain a narrow list of holdings, or rebalance on a much tighter schedule.

The new architecture therefore makes Betterment automated investing more configurable, but not unlimited. It gives the investor more say over the destination while leaving much of the route under automated control.

Mechanics of the 7% Drift Threshold and Automated Rebalancing

The most important operational number in Custom Portfolios is the default 7% drift threshold at the security-group level. This threshold determines when the system considers an allocation sufficiently out of line to trigger automated rebalancing.

Suppose a security group has a target allocation and moves beyond the permitted band. Once the drift reaches the threshold, Betterment can sell assets that have become relatively overweight and direct the proceeds toward areas that have fallen below their intended allocation. If the drift remains inside the band, the system does not need to trade simply because prices have moved.

That design is deliberately conservative. It accepts some deviation in exchange for fewer unnecessary transactions.

A simple illustration makes the trade-off clearer. If a US large-cap ETF exposure moves just beyond its permitted 7% drift from target, the system may intervene. If it remains just below that level, it may leave the position alone. The difference between those two outcomes is not a judgment about whether the market is attractive. It is a mechanical response to the distance between the target and the current allocation.

The 7% threshold is not a feature in isolation. It is a trade-off: less trading friction and potentially fewer taxable events in exchange for tolerating a wider short-term departure from the target.

Why a wider band can help

Rebalancing has a cost even when commissions are not the issue. Selling an appreciated position can create a taxable gain. Frequent trading can also make a portfolio more reactive than the investor's long-term plan requires. A threshold prevents the system from chasing every minor movement in the market.

This matters particularly in a diversified ETF portfolio. Prices move constantly, but not every price movement changes the investor's strategic position in a meaningful way. A narrow threshold could result in more frequent interventions without producing a better long-term allocation. The 7% rule gives the portfolio room to move before the system decides that the deviation is material.

The potential benefits include:

• Fewer rebalancing transactions caused by routine market noise.
• Lower pressure to sell appreciated holdings after small price movements.
• Less operational activity in an account designed for long-term ownership.
• A clearer separation between strategic allocation and short-term market fluctuation.

The value of that restraint is strongest for investors who do not want their portfolio constantly adjusted. A buy-and-hold investor may prefer to accept temporary deviations rather than create a new taxable or trading event every time one segment of the market moves.

Why the same band can feel too wide

The limitation is equally straightforward: a portfolio can remain meaningfully different from its target before the system acts. In a volatile market, the allocation may spend time away from its intended mix. A portfolio designed around a 60/40 stock-and-bond structure could temporarily look closer to 67/33 or 53/47 before the drift rule forces a correction, depending on how the relevant groups move.

That does not mean the portfolio has failed. It means the automation is operating according to its tolerance band. But the distinction matters for investors who treat their target allocation as a tight risk-control mechanism rather than a long-term destination.

A 7% threshold may be perfectly reasonable for someone focused on minimizing intervention. It may be uncomfortable for someone who uses allocation bands to keep risk exposure closely aligned with a written investment policy. The right question is not whether 7% is objectively good or bad. It is whether the investor understands what kind of portfolio behavior that threshold permits.

The interaction with cash flows

New deposits and withdrawals can also affect how an automated portfolio is maintained. In an automated system, cash flows may provide an opportunity to direct money toward underweight areas rather than immediately selling something else. That can reduce the need for taxable sales in some situations, although the exact handling depends on the platform's rules and the account's circumstances.

This is one reason not to evaluate the drift threshold by itself. Rebalancing is not simply a switch that is either on or off. It is part of a broader operating system that includes deposits, withdrawals, target weights, security groups, and available tax-management opportunities.

The architecture is more useful when understood as a set of priorities. First, maintain the intended exposure within the permitted boundaries. Second, avoid unnecessary trading. Third, use the account's tax features where they are available. The platform's automation is designed to balance those goals, not to maximize any one of them at every moment.

Tax-Loss Harvesting Limitations: Why ETFs Differ from Individual Stocks

Betterment tax-loss harvesting is built around eligible ETFs inside Custom Portfolios. That restriction is not a minor technical note. It defines the role the platform can play in an investor's broader tax strategy.

Tax-loss harvesting works by selling an investment that has declined in value and replacing it with a suitable alternative, allowing the investor to realize a loss while maintaining similar market exposure. The replacement must be selected carefully because tax rules can limit the usefulness of a transaction that creates a superficial loss or otherwise fails to preserve the intended tax treatment.

In Betterment's system, that process operates within the ETF portion of the portfolio. As of the Custom Portfolios launch described here, automated TLH is supported for eligible ETFs. It is not supported for individual stocks or mutual funds.

That creates a clear boundary between the automated core and the rest of an investor's holdings.

The ETF advantage inside an automated system

ETFs are well suited to a rules-based harvesting process because they provide broad exposure and can often be paired with another fund that serves a similar portfolio function. The system does not need to determine whether a particular company's earnings outlook has changed or whether a stock-specific decline represents a permanent deterioration in the business. It can focus on maintaining the intended asset-class or market exposure through eligible fund positions.

This is a good match for a diversified core portfolio. If the investor's objective is broad market exposure, the automation can look for losses within the funds used to represent that exposure and manage the replacement process according to its rules.

The advantage is less about predicting markets than about reducing the amount of manual coordination required. The investor does not have to monitor every ETF, calculate its tax basis, decide whether a substitute is sufficiently similar, and remember to revisit the position later. The platform is taking responsibility for that operational layer.

The individual-stock wall

The limitation becomes obvious when the portfolio includes individual equities. If a stock position falls substantially, Betterment's automated system cannot simply apply the same process to that position through Custom Portfolios. The investor must manage that decision separately, in a different account or through a more manual strategy.

For an investor who holds a small number of individual stocks as satellite positions, this may be acceptable. The managed ETF core can still provide automated diversification and tax management, while the stock sleeve remains under personal control.

For an investor who views individual stocks as the main source of potential upside, however, the division is more consequential. The platform cannot harvest losses on those positions as part of its automated process. Any tax management attached to them requires the investor to track purchase lots, account for wash-sale concerns, consider replacement exposure, and decide whether selling still makes sense from an investment perspective.

The result is not that individual stocks are inherently inferior. It is that they do not fit the same automation model.

TLH is not the same as higher returns

Tax-loss harvesting can improve after-tax outcomes, but it is not a free return generator. It changes the timing and structure of taxable gains and losses. A harvested loss may be useful today while affecting the tax basis of a replacement investment. The benefit depends on the investor's tax situation, future transactions, income, and the eventual disposition of the investment.

That is why a platform's harvesting feature should not be evaluated by the number of trades it makes. More harvesting is not automatically better. The goal is to manage taxes without compromising the portfolio's intended exposure or turning tax management into an excuse for unnecessary activity.

Betterment's ETF focus keeps the feature narrow enough to automate. It also means the investor should not confuse an automated core with a complete tax strategy for every account and every security.

Tax-loss harvesting is a tax tool, not a return booster. The investor benefits only when the harvested loss is real, the replacement maintains the intended exposure, and the rest of the strategy is already in place.

Leveraging AI Insights for Portfolio Concentration Analysis

Betterment's AI Insights is the most visible attempt to make portfolio analysis more accessible. It uses a generative AI layer to analyze and describe concentrations in a portfolio. That makes it useful as an interpretive tool, but its role should be kept in proportion.

AI Insights is not a substitute for a fiduciary relationship, a personalized financial plan, or independent investment judgment. It is better understood as a portfolio X-ray: a way to identify what the holdings collectively represent, including concentrations that may not be obvious from the list of individual funds.

The tool analyzes and describes concentrations across categories confirmed by Betterment, including:

• Individual stocks.
• Sectors.
• Geographic regions.
• Asset classes.

That framing is important. A portfolio can appear diversified because it owns several ETFs while still carrying a large combined exposure to the same companies, sector, or region. Looking at each holding separately may conceal the overlap. A concentration analysis brings the portfolio back to the level that matters: the risks the investor owns in aggregate.

Why concentration analysis matters more in a custom portfolio

A standard model portfolio has a predefined structure. A custom portfolio introduces more room for personal preferences, and those preferences can accumulate.

An investor might increase an allocation to one sector in one part of the portfolio while holding another fund that already contains many of the same companies. A regional tilt can coexist with a broad international fund that provides additional exposure to that market. A collection of stock positions can overlap with the largest holdings of an ETF.

None of those choices is automatically wrong. The problem is that the investor may not realize how much of the portfolio is quietly concentrated in the same handful of names until a sector-specific event exposes the overlap. AI Insights is designed to surface that pattern before the exposure becomes a problem.

The tool is also useful after a change. Adding a new ETF, selling a position, or increasing a target weight all shift the portfolio's underlying concentration. A periodic look at the analysis can confirm whether the new mix still matches the investor's stated risk preferences.

Where AI Insights stops being useful

The same caveats that apply to any automated analysis apply here. The platform can describe what the holdings represent, but it cannot tell the investor whether a particular concentration is acceptable in the context of their broader financial plan. A heavy technology exposure may be appropriate for a young investor saving aggressively, and a problem for someone approaching retirement who already depends on a similar sector in another account.

AI Insights also does not replace a written investment policy. It can confirm a concentration, but it cannot decide whether the concentration reflects a deliberate decision or an unintended drift. That interpretation remains the investor's responsibility, supported where appropriate by a financial planner or tax professional.

The most honest way to use the feature is as a regular checkpoint rather than as a final word. A few minutes spent reviewing the analysis after a major change to the portfolio is usually enough to catch the obvious overlaps. The investor still has to decide what to do with what the tool finds.

Fee Structures and the Reality of the 0.25% Wrap Fee

Betterment's pricing model is straightforward on paper: a 0.25% annual advisory fee on assets under management for the standard Digital plan, with a higher tier that bundles access to human advisors. The 0.25% figure is competitive against traditional financial advisory fees, and it is the headline number most readers will see.

The fee, however, is not the only cost in the account. The underlying ETF expense ratios are charged by the fund companies themselves and are paid out of the fund assets. They are small, but they are not zero, and they exist on top of the advisory fee. A portfolio of low-cost index ETFs typically keeps that layer modest, but the investor should still see the total cost as the sum of both parts.

The same applies to the Cash Reserve account. Betterment cash reserve yield refers to the interest paid on uninvested cash held in the platform, and that interest comes from partner banks rather than from Betterment itself. The yield can be useful as a parking spot for short-term funds or an emergency cushion, but it is not a substitute for the portfolio. Treating cash as a permanent holding can quietly drag on long-term returns.

What the 0.25% really buys

The advisory fee pays for the operating system: drift monitoring, rebalancing logic, tax-loss harvesting, deposit and withdrawal handling, and the interface that ties it all together. It does not pay for individual security selection or for personalized stock advice. The investor is buying automation, not a bespoke manager.

For an investor who would otherwise hold a basic index portfolio and rebalance manually once or twice a year, the 0.25% is the price of letting a system handle the recurring work. Whether that price is worth it depends on the value the investor places on not having to monitor the account. Some investors will find the automation valuable. Others will find it redundant.

The higher tier adds human advisor access and more structured planning conversations. For investors with more complex situations — concentrated stock positions, equity compensation, multiple account types, or impending liquidity events — that upgrade may be the more meaningful difference. The base 0.25% is a fair price for the automation. The upgrade is essentially buying access to a planner, and that should be evaluated as such.

The compounding effect of small fees

A 0.25% annual fee does not feel significant on its own. The compounding of that fee over a multi-decade horizon is the more important calculation. Even modest differences in cost matter more the longer the money stays invested.

That is why the underlying ETF selection still matters. A portfolio that holds relatively expensive funds adds another layer of cost on top of the advisory fee, and the combined number is what actually leaves the account. The investor does not need to chase the cheapest fund in every category, but it is worth knowing how much is being paid in total and whether the underlying choices are consistent with the rest of the cost structure.

Navigating the Legacy of the 2023 SEC Settlement and Compliance

Any conversation about Betterment tax-loss harvesting has to acknowledge the regulatory history. In 2023, the U.S. Securities and Exchange Commission settled charges against Betterment regarding statements made about its tax-loss harvesting feature. The substance of the settlement concerned specific claims that the platform had made about how the feature worked and how reliably investors could expect to receive the described tax benefit.

The settlement is no longer a current event, but it remains a useful reference point. It reminded the industry that descriptive language about automated features has to match the actual operation of those features. Marketing copy that implies a tax-loss harvesting benefit in every situation, or that promises a benefit without specifying its conditions, creates a real exposure for the platform that makes the claim.

For an investor reading the platform's current material, the practical lesson is to evaluate tax-loss harvesting descriptions by what they actually do. Which positions are eligible? When does the feature act? What is the replacement process? Those questions matter more than any general assurance.

What changed after the settlement

Betterment revised the way it describes its TLH feature and the scope of the offering. The current product documentation is more explicit about which account types, position types, and conditions are covered. That is consistent with what most compliance-conscious platforms do after a regulatory engagement: tighten the language, narrow the implied scope, and bring the marketing closer to the mechanics.

The underlying automation did not disappear. The feature still operates on eligible ETFs in taxable accounts. What changed was the clarity around its boundaries, including the absence of support for individual stocks and mutual funds. An investor who reads the current documentation carefully will find fewer ambiguous claims and more specific operational detail.

A reasonable posture for an investor

The settlement is not a reason to avoid the platform, nor is it a reason to assume the current product is unreliable. It is a reason to read the documentation as an adult and to treat broad claims with appropriate skepticism. Automated tax features are useful when they are well understood and not oversold.

The same posture applies to the rest of the Betterment offering. The platform is a competent implementation of automated investing, with clear operating rules and an explicit set of limitations. Investors who understand those rules get the most out of the service. Investors who expect the platform to do more than its rules permit tend to be the ones who feel disappointed.

What Actually Matters in the New Architecture

The refreshed Custom Portfolios architecture is best read as a clearer version of what Betterment has always offered: more configuration inside a defined system, not a fundamentally different product. The new security-group framework makes drift and rebalancing more precise. The 0.25% advisory fee and the underlying ETF costs remain the main economic facts. The tax-loss harvesting feature continues to operate on eligible ETFs within a clear scope. The AI Insights layer adds a useful interpretive tool without changing the underlying portfolio.

For an investor evaluating the platform, the questions remain practical rather than rhetorical. Does the 7% drift threshold match the kind of control you want over the portfolio? Are the eligible ETFs sufficient for your asset allocation? Is the cost worth the automation? And are you clear about what the platform will not do for you?

Those are the same questions that applied before the August 2026 update. They are simply easier to answer now that the operating rules are more explicitly described. That, more than any headline, is the substance of the change.