Robo-Advisor Tax-Loss Harvesting: Real Alpha or Just Hype?

Their pitch is clean: automated tax-loss harvesting generates enough tax alpha to cover the fee and deliver net positive returns. Vanguard's own research puts gross annual tax alpha from harvesting at 0.47% to 1.27% over a 15-year window. Betterment reported that nearly 70% of customers who used its TLH feature covered their advisory fees through estimated tax savings in 2022–2023. On paper, the math works. In practice, the math depends on variables that the marketing material tends to bury in the footnotes.

We need to stress-test this.

The Math of Tax Alpha: Separating Marketing from Reality

Tax-loss harvesting is the practice of selling a security at a loss to realize a capital loss, then immediately buying a similar (but not "substantially identical") security to maintain market exposure. The realized loss offsets capital gains elsewhere in the portfolio, and up to $3,000 of net capital losses can offset ordinary income each year under current U.S. tax code. Any excess carries forward.

That's the mechanism. The robo-advisor version automates it daily, scanning for loss-harvesting opportunities across thousands of positions and executing trades without human intervention. The premise is that a human investor will be too lazy, too emotional, or too busy to do this themselves. The premise is also that the algorithm will catch more opportunities than a manual approach.

Gross tax alpha is not net tax alpha. The difference is what you actually keep after fees, friction, and decay.

The first reality check: the headline numbers are gross, not net. When Vanguard reports 0.47% to 1.27% in annual tax alpha, that comes off the top — before the 0.25% advisory fee, before any cash drag from harvesting activity, and before the erosion from wash-sale violations and portfolio turnover. Take the low end of that range, subtract the fee, and you have 0.22% of theoretical annual benefit. That's not nothing. But it's not the windfall the sales pages suggest.

The second reality check: 2018. The U.S. Securities and Exchange Commission charged a robo-advisor for making misleading claims about automated tax-loss harvesting and its impact on portfolio returns. The specifics matter less than the signal. When the federal regulator has to step in and call out marketing claims in this niche, the gap between advertised and actual performance is wide enough to be a legal problem.

The third reality check: vendor data is not independent data. Wealthfront reported an average annual harvesting yield of 1.59% across all client vintages and risk scores in classic portfolios over a one-year period ending in 2025, and 3.87% over a decade. These numbers are real, but they are yield, not net alpha. They don't include the advisory fee. They don't include the implicit cost of holding cash reserves to execute trades. And they don't apply uniformly across the customer base — the clients with the largest balances, the highest tax brackets, and the most diversified external accounts capture the disproportionate share of the benefit.

The numbers stack. They don't always stack in your favor.

The Decay Factor: Why Harvesting Benefits Front-Load in Early Years

Here is the part of the TLH pitch that almost no one talks about: harvesting benefits decay.

J.P. Morgan research found that 80% of tax-loss harvesting benefits are captured in the first five years of an account. After that, the yield drops sharply. The reason is mechanical. Every time you harvest a loss, you sell a position at a discount and reinvest in a similar security. The harvested loss reduces your cost basis on the new position. If the security recovers — and historically, broad equity indices do recover — the new position will eventually be sold at a gain, and that gain will be taxed.

You haven't eliminated the tax. You've deferred it. And you've reduced your cost basis, which means your future tax bill on the same dollar of appreciation is larger than it would have been without harvesting.

Tax-loss harvesting is tax deferral with a cost-basis haircut. The benefit is real but it is front-loaded, and it requires new capital to stay alive.

The mechanism that keeps harvesting productive over the long term is new cash deposits. Every time you add fresh money to the account, the algorithm can buy securities at a higher cost basis, creating new loss-harvesting opportunities when the market dips. If you stop contributing — or if your contributions slow down — the harvesting yield thins out. An account that was heavily harvested in year one might produce a fraction of that yield in year ten.

This is the dirty secret of long-horizon TLH. A buy-and-hold investor who makes a single $50,000 deposit and never adds another dollar will harvest aggressively for the first few years, then watch the algorithm grind through a portfolio with progressively lower cost bases. By year fifteen, the average tax alpha might be a tenth of what it was in year three.

The 2020 Financial Analysts Journal paper by Shomesh Chaudhuri, Terence Burnham, and Andrew Lo tested automated tax-loss harvesting against benchmark portfolios using historical U.S. stock market data from 1926 to 2018. The result: harvesting produced meaningful benefits, but the benefits were concentrated in the early years of the account and required active cash management to sustain.

Hidden Friction: Wash-Sale Risks and External Account Complexity

The wash-sale rule is the IRS's mechanism for preventing taxpayers from harvesting losses on paper while maintaining the same economic position in practice. If you sell a security at a loss and buy a "substantially identical" security within 30 days before or after the sale, the loss is disallowed. For an individual investor, this is annoying. For a robo-advisor managing multiple accounts across multiple platforms, it is a minefield.

The standard automated TLH strategy avoids wash sales by swapping one ETF for a similar but not identical ETF — for example, selling the Vanguard S&P 500 ETF (VOO) and buying the iShares Core S&P 500 ETF (IVV). These two ETFs track the same index but are issued by different providers, so the IRS treats them as distinct securities. The wash-sale rule does not apply.

But that only solves the in-platform problem. If you hold similar ETFs in an unlinked external brokerage account — a spouse's IRA, a 401(k) at work, an old account from a previous employer — the algorithm cannot see those positions. A trade that the platform believes is wash-sale-clean might actually trigger a violation if you happened to buy the same or a similar ETF in another account within the 30-day window.

The SEC does not require automated investing platforms to track wash-sale exposure across external accounts. Most of them disclose this limitation somewhere in their fine print. Almost none of them surface it in their marketing.

The result: an investor who thinks they are harvesting 0.5% in annual tax alpha might actually be harvesting 0.3%, with the rest disallowed at tax time. The algorithm does not know what it does not know. You do, or you should.

There is also a subtler form of friction. Every harvest is a trade. Every trade has a bid-ask spread, a small market impact, and a tax lot selection decision. Over a year, a high-turnover harvesting strategy might execute dozens of round-trip trades in a single account. The cumulative cost of those trades is small in percentage terms but real in dollar terms, and it shows up as a slight underperformance versus the theoretical model.

Beyond the 0.25% Fee: When Automation Fails to Pay for Itself

Let's run the actual numbers for a few common investor profiles.

The $25,000 starter account in the 12% federal bracket. Annual fee: $62.50. Potential gross harvesting benefit: 0.3% to 0.5% on a $25,000 balance is $75 to $125 per year — but only if the account actually generates harvestable losses, which is unlikely in a steady up market, and only if the investor's tax bracket is high enough for the offset to matter. Realistic net benefit in a typical year: often negative. The fee is fixed. The harvest is not.

The $250,000 account in the 32% federal bracket. Annual fee: $625. Potential gross harvesting benefit at 0.5% to 1.0%: $1,250 to $2,500 per year. The math starts to work. This is the profile that platforms design their marketing around, because this is the profile where the fee is most likely to be covered by the tax savings.

The $1 million account in the 35% federal bracket with active cash flow. Annual fee: $2,500. Potential gross harvesting benefit at 0.7% to 1.2%: $7,000 to $12,000 per year. The math works, and works well — but at this level, the investor is sophisticated enough to either hire a flat-fee CPA or run a direct indexing strategy themselves, often at lower total cost.

The 0.25% fee is a fixed percentage. The harvesting benefit scales with account size, tax bracket, and turnover. The crossover point where automation pays for itself is somewhere north of $100,000 in a moderate bracket, and the returns are concentrated in the first five years.

The do-not-claim list matters here. Harvesting does not provide permanent tax elimination. It defers taxes and reduces cost basis. The benefits do not stay high over decades without ongoing new cash deposits. And the platform cannot guarantee that harvesting will exceed the annual advisory fee for your specific situation. These are not edge cases — they are the default for a majority of the customer base.

For investors who fall outside the high-bracket, high-balance, high-cash-flow profile, the question is not whether TLH works. It works. The question is whether it works for you at a price you are willing to pay. For a useful comparison of how different investor profiles approach the same set of tools — including women investors, who often face different career and income trajectories that change the math — the resource at equityforher.com is worth reading before committing to any single platform's defaults.

The Direct Indexing Shift: Is It the Future of Tax Efficiency?

Direct indexing is the next step up from ETF-based automated harvesting. Instead of buying a small set of broad-market ETFs and swapping between them to avoid wash sales, the platform buys the individual stocks in the index — often 500 to 1,000+ positions — and harvests losses at the single-stock level.

The advantage is granularity. If the S&P 500 is up 12% for the year but 200 of its 500 constituents are down, an ETF-based strategy can only harvest the broad fund's gain or loss. A direct indexing strategy can harvest losses on those 200 individual stocks, generating tax alpha on positions that the ETF approach would never touch.

The threshold is typically $100,000 in assets. Below that, the platform cannot build a sufficiently diversified portfolio of individual stocks. Above it, the harvesting opportunities multiply.

The trade-off is complexity and cost. Direct indexing platforms charge more than the standard 0.25% — often 0.35% to 0.50% — and the underlying portfolio requires more maintenance, more rebalancing, and more sophisticated tax lot accounting. The gross harvesting yield is higher, sometimes substantially higher. The net yield after fees is a different question, and it depends on the same variables: account size, tax bracket, cash flow, and time horizon.

For an investor with $500,000 in a taxable account, a 32% federal bracket, and ongoing contributions, direct indexing can produce meaningful net tax alpha — enough to justify the higher fee and the additional complexity. For an investor with $75,000 in a 12% bracket, the math collapses. Direct indexing is not a universal upgrade. It is a specialized tool for a specific investor profile.

The broader trend in the industry is clear. As more platforms compete on harvesting yield, the marketing gets louder and the underlying economics get tighter. The 0.25% fee is no longer enough to fund a competitive direct indexing offering, which is why the major players have been pushing customers into higher-fee products with more sophisticated harvesting. Whether that is a good deal depends entirely on whether the additional harvesting yield exceeds the additional fee.

The Honest Answer

Is robo-advisor tax-loss harvesting worth the fee?

For some investors, yes. If you have a taxable account north of $100,000, a marginal federal tax bracket of 24% or higher, ongoing contributions that keep the cost basis fresh, and no wash-sale exposure from external accounts, automated harvesting will almost certainly cover the 0.25% fee and deliver some net alpha. The first five years will be the most productive. After that, expect decay.

For other investors, no. If your account is under $50,000, your tax bracket is low, your contributions are sporadic, or you hold similar ETFs across multiple external accounts, the fee is a cost you are paying for a feature that is not delivering. You are better off in a low-cost brokerage with a target-date or total-market fund, paying nothing in advisory fees, and harvesting manually once a year if your tax situation warrants it.

The marketing will not tell you which category you fall into. The algorithm will not tell you either. That is a job for a spreadsheet, a calculator, and a clear-eyed look at your own tax bracket and account structure.

The fee is the same for everyone. The benefit is not. Do the math before you pay it.