How Robo-Advisors Invest Your Money: From Questionnaire to Portfolio
Last updated August 2026
Short answer
A questionnaire produces a single risk score. That score selects one of roughly five to fifteen pre-built model portfolios, which differ mainly in their stock-to-bond split. The allocation is filled with broad, low-cost index ETFs. As markets move the mix off target, the platform rebalances, by threshold, by calendar, or by pointing new deposits at whatever is underweight. In a taxable account it may also harvest losses automatically. Nothing is designed for you individually: the personalisation is choosing which rung you land on. Walnut is not a robo-advisor and is not an investment adviser.
The interesting thing about robo-advisor machinery is how little of it is prediction. There is no forecasting step, no view on the market, and in most cases no AI in the sense the name implies. It is a scored questionnaire, a lookup table of allocations, and a rebalancing rule. Knowing that makes the category much easier to evaluate, because it tells you what you are and are not paying for.
The six stages, end to end
1. The questionnaire becomes a number
Six to fifteen questions produce a single risk score.
The questions cover your goal, your horizon, and how you say you would react to a fall. The answers are weighted into one number on a scale. Two things follow from this that are worth knowing: the horizon usually carries more weight than the emotional questions, and the score is the only output. Nothing else about your answers survives into the portfolio.
2. The score selects a model portfolio
You are matched to one of roughly five to fifteen pre-built allocations.
This is the step most people picture wrongly. No portfolio is designed for you. The platform maintains a ladder of allocations from conservative to aggressive, mostly varying the split between stocks and bonds, and your score picks the rung. Two people with different answers landing on the same score get the identical portfolio.
3. Funds fill the allocation
Broad, low-cost index ETFs, usually eight to twelve of them.
Total US market, US large and small cap, developed international, emerging markets, and a set of bond funds by duration and credit quality. Some platforms use their own funds, which moves revenue from the advisory fee into the fund expense ratio. This is why robo-advisor portfolios look alike: the same small universe of cheap index funds is available to everyone.
4. Money is invested, and some may stay in cash
Deposits are allocated on a schedule, and some platforms hold a required cash slice.
Most invest new deposits within a day or two. A minority require a percentage of the portfolio to stay in cash as part of the design, which is worth identifying because uninvested money is a cost even when it is not billed as a fee.
5. Drift triggers rebalancing
Either on a calendar or when an asset class moves past a threshold.
Markets move the mix away from its targets. Threshold rebalancing acts when a holding drifts more than a set amount from target, often around five percentage points; calendar rebalancing acts quarterly or annually regardless. Many platforms also rebalance using new deposits first, buying whatever is underweight, which avoids selling and therefore avoids realising gains.
6. In taxable accounts, losses may be harvested
Positions at a loss are sold and replaced with a similar fund.
The platform sells a holding that is down, books the loss to offset gains elsewhere, and immediately buys a fund that tracks something similar but not substantially identical, so you stay invested while the wash-sale rule is respected. After 30 days it may switch back. This does nothing in an IRA, because there are no taxable gains to offset.
The three ways rebalancing is triggered
| Method | What triggers it | What it means for you |
|---|---|---|
| Threshold | When a holding drifts past a set distance from target, often around 5 percentage points | Responsive to real drift; can act several times in a volatile year |
| Calendar | On a fixed schedule, quarterly or annually | Predictable; can leave a portfolio off target between dates |
| Cash-flow | New deposits buy whatever is underweight | Avoids selling, so avoids realising gains; only works while you are contributing |
Cash-flow rebalancing is the one worth understanding if your account is taxable, because buying the underweight asset with new money avoids selling anything, and selling is what creates a tax event. It only works while you are still contributing, so its usefulness fades once you stop.
Get a recommendation for your situation
Walnut is the AI that knows your portfolio: ask anything in plain English, research any fund, and get an honest second opinion. On the broker you already use, read-only, and you approve every trade. Walnut is not a registered investment adviser.
What is deliberately absent from the machinery
Reading the six stages back, the striking thing is what never appears. There is no step where anything forms a view about a company, a sector, or the market. There is no forecast, no valuation judgement, and in most cases nothing that would ordinarily be called AI, despite the name the category carries.
That absence is a design choice rather than an omission. A portfolio built from broad index funds and rebalanced on a rule is trying to capture market returns at low cost, and adding a prediction step would work against that. It is also why claims of market-beating performance should be treated with suspicion in this category: the machinery has no mechanism that would produce it.
The practical consequence for you is that platform differences are narrower than the marketing implies. Two robo-advisors at the same risk level are running similar allocations of similar funds on similar rules. What genuinely differs is cost, account-type coverage, tax features in a taxable account, and how much of your financial picture the platform can see.
How to read a platform's methodology document
Every platform publishes one, almost nobody reads it, and four things in it are worth finding. How many model portfolios there are, which tells you how coarse the personalisation is. Which funds fill them, and whether those are the firm's own, which tells you where the revenue sits.
What triggers a rebalance, threshold or calendar, and at what distance from target. Whether any cash allocation is required, which is the cost that never appears as a fee. Four answers, usually findable in ten minutes, and together they tell you more than any comparison table including this one, because they describe the actual product rather than its positioning.
What this tells you about the fee
Laid out this way, the advisory fee is buying execution and discipline, not insight. The allocation ladder is not proprietary, the funds are available to anyone, and the rebalancing rule is a threshold. That is not a criticism: doing those things reliably for years is genuinely valuable, and most people who intend to do them by hand do not.
It does mean the fee should be judged against whether you would perform the maintenance, rather than against any expectation of better returns. The arithmetic is in robo-advisor fees explained, and the trade is worked through in robo-advisor vs doing it yourself.
Checking the same things on a portfolio you already own
Every measurement in this page has an equivalent you can run against your own account without moving anything: how far each position has drifted from where you wanted it, how the mix compares to a target you set, and how the whole thing has done against the S&P 500. Those are data questions, and they do not require custody.
What you do not get that way is the automation: nothing rebalances on its own. See how to rebalance a portfolio with AI and how to check concentration.
FAQ
How many model portfolios does a robo-advisor actually have?
Typically somewhere between five and fifteen, arranged as a ladder from conservative to aggressive. That is the real extent of the personalisation: your questionnaire score picks a rung. It is worth knowing because it explains why two people with quite different circumstances can end up holding exactly the same thing.
What happens to my money on the day I deposit it?
Most platforms invest new deposits within a day or two, buying across the model allocation or directing the money at whatever is currently underweight. Some hold it briefly to batch trades. The gap matters little on a recurring contribution and matters more on a single large deposit, which is worth asking about if you are transferring a lump sum.
Do robo-advisors trade often?
No, and that is deliberate. Beyond investing new deposits, trading happens when drift crosses a threshold or on a rebalancing schedule, which in a calm year can be very few transactions. Low turnover is a feature: it keeps costs down and, in a taxable account, generates fewer taxable events.
Can two people with the same risk score get different portfolios?
Generally no on the same platform, which is the point of the model-portfolio approach. Where they can differ is by account type, since a taxable account may get a tax-managed variant, and by any restriction you have applied, such as choosing a socially responsible version.
How do robo-advisors invest your money?
A short questionnaire produces a single risk score. The score selects one of a handful of pre-built model portfolios that mainly differ in their stock-to-bond split. That allocation is filled with broad, low-cost index ETFs. New deposits are invested on a schedule, and as markets move the mix off target the platform rebalances, either when drift passes a threshold, on a calendar, or by directing new money to whatever is underweight.
Does a robo-advisor build a portfolio just for me?
No, and this is the most common misunderstanding. Platforms maintain a ladder of pre-built allocations and your questionnaire score picks which rung you land on. Two people with quite different answers that produce the same score receive the identical portfolio. The personalisation is in selecting the rung, not in designing the allocation.
What funds do robo-advisors buy?
Broad index ETFs, usually eight to twelve: total US market, US large and small cap, developed international, emerging markets, and bond funds split by duration and credit quality. Some platforms use their own funds instead, which shifts cost from the advisory fee into the fund expense ratio.
How often does a robo-advisor rebalance?
It depends on the method. Threshold rebalancing acts whenever a holding drifts more than a set distance from target, often around five percentage points, so it can act several times in a volatile year and not at all in a quiet one. Calendar rebalancing acts quarterly or annually regardless. Many platforms also use incoming deposits to buy whatever is underweight, which rebalances without selling.
How does automated tax-loss harvesting work?
The platform sells a holding that is showing a loss, books that loss to offset capital gains elsewhere, and immediately buys a different fund tracking something similar but not substantially identical, so your market exposure is unbroken and the wash-sale rule is not triggered. After the 30-day window it may swap back. It only applies in taxable accounts.
Does a robo-advisor keep some of my money in cash?
Most invest essentially all of it. A minority require a percentage of the portfolio to sit in cash as part of the design, and those are often the platforms advertising no advisory fee, because interest on that cash is part of how they earn. Uninvested money is a real cost on a long-horizon portfolio even though it never appears as a fee.
Can I see or change the portfolio a robo-advisor picks?
You can almost always see it. Changing it is limited by design: some platforms let you shift to a more or less aggressive rung, exclude a category, or add a socially responsible variant, but choosing individual holdings defeats the product. If you want to select what you own, a self-directed account is the right tool.
What happens to my existing holdings if I transfer them in?
Most platforms sell them and buy their own allocation, because the whole model depends on holding the model portfolio. In a retirement account that is harmless. In a taxable account it realises gains on the platform's timing rather than yours, so ask before transferring whether they can hold or transition positions gradually.
Do robo-advisors use AI to pick investments?
Generally no, despite the name. The machinery is rules-based: a scored questionnaire, a fixed ladder of allocations, and threshold or calendar rebalancing. That is automation rather than prediction, and it is a feature: the portfolio is designed to track markets cheaply, not to forecast them.
Related articles
Walnut is informational and is not an investment adviser. This describes the standard machinery of the category; individual platforms differ in the number of model portfolios, the funds used, and the rebalancing rule. Verify the specifics on the provider's own disclosures.