Robo-advisors, chat assistants that read your accounts, AI stock scores and funds picked by a model all get called AI investing. What each one does, how much of it is really AI, what changes when you hold for years, what the SEC has said and done, and seven questions to ask before you trust one with money.
November 16, 2026
18
min to read
The short answer
AI investing is using software that learns from data or reads language to help you choose, run or research investments you hold for months or years. It covers robo-advisors that keep a mix of index funds on target, assistants that read filings and accounts, and AI stock scores and funds. In robo-advisors, much of the “AI” is plain rules.
AI investing is using software that learns from data or reads language to help you decide what to own, keep a portfolio on track, or research an investment you plan to hold for months or years. The phrase covers four kinds of product, and only some of them use AI in the strict sense: machine-learning models that learn patterns from past data, and language models, the AI behind ChatGPT and Claude, that read and write.
| Kind | What it does | Who decides | Examples |
|---|---|---|---|
| Robo-advisors (automated investing) | Builds a portfolio of index funds from your answers to a questionnaire, then rebalances it and reinvests dividends | The algorithm, within the risk level you picked | Betterment, Wealthfront, Vanguard Digital Advisor |
| Research and portfolio assistants | Reads filings, news and your linked accounts, and answers questions in plain English | You | ChatGPT with its Finances feature; Claude or ChatGPT with web search |
| AI stock scores | Rates stocks with a model trained on past data | You, with the score as one input | Danelfin’s AI Score |
| AI-run funds | A fund whose holdings a model picks | The fund’s model and its managers | Amplify AI Powered Equity ETF (AIEQ) |
The word “AI” does different work in each row. The SEC describes robo-advisers, as it spells them, as “registered investment advisers that use computer algorithms to provide investment advisory services online with often limited human interaction.” An algorithm can be a fixed set of rules that people wrote; nothing in that definition requires machine learning. Vanguard’s own explainer says a robo-advisor “uses an algorithm to automatically build and rebalance a strategic portfolio” (Vanguard Digital Advisor). So the first question about any AI investing product is which part is learned from data, which part is rules, and which part is you.
| Term | What it means |
|---|---|
| Automated investing | Money invested on a schedule and managed by software: rebalancing, reinvesting dividends, harvesting tax losses |
| Rebalancing | Selling some of what grew past its target weight and buying what fell behind, so the mix stays where you set it |
| Tax-loss harvesting | Selling an investment at a loss to offset taxes on gains, and buying a similar one so the money stays invested |
| Fiduciary | An adviser who must act in your best interest and tell you about its conflicts |
| AI washing | Claiming to use AI that a firm doesn’t use, or doesn’t use the way it says; the SEC has fined firms for it |
The robo-advisors on this page run the same loop. Betterment sums it up as “Get a globally diversified portfolio and we’ll handle the trading, rebalancing, and tax management for you.”
Your goal, your timeline, how you would react to a fall. The SEC’s staff notes that many robo-advisers rely solely on questionnaires, so the answers you give are the advice you get. Answer the way you would really behave.
Usually a mix of stock and bond index funds. Vanguard’s all-index option holds low-cost Vanguard ETFs across US and international stock and bond markets.
Prices move and the weights drift. Vanguard rebalances when a portfolio drifts more than 5% from its recommended allocation; Wealthfront says its automation keeps a portfolio “balanced to your risk level” and reinvests dividends.
In a taxable account, software can harvest losses every day the market gives it one.
A recurring deposit invests on a schedule, which is dollar-cost averaging.
Software suits harvesting because it can watch every holding every day. Wealthfront says its software “automatically captures daily losses from market movement.” The catch is the wash sale rule. Under IRS Publication 550, a loss is not deductible if you buy substantially identical stock or securities within 30 days before or after the sale; the disallowed loss is added to the cost basis of the new shares instead. One of the SEC’s first two cases against a robo-advisor, covered below, was about exactly that.
FINRA defines dollar-cost averaging as investing your money “in equal portions, at regular intervals, regardless of current market conditions.” Its May 2026 guide says a fixed schedule “can remove some of the emotion from investing,” and you buy more shares when prices are low and fewer when they are high. It is also frank about the trade-off: spreading money out “has lower risk but often produces lower returns than lump sum investing, especially over longer periods of time.” Our DCA backtest tests that same trade-off on crypto, weekly buys against a lump sum over three years.
Mostly it is automation. The SEC’s 2017 staff guidance gives modern portfolio theory, the math of spreading risk across assets, as an example of what a robo-adviser’s algorithm may be based on, and asks firms to describe “the assumptions and limitations of the algorithm.” A robo-advisor that never forecasts anything can still be a good tool, because most of what it does is keep you invested, diversified and on schedule.
This is where AI in the strict sense does the work: models that read language, or that learn patterns from market data.
ChatGPT’s Finances feature links your financial accounts through Plaid and “can securely access balance, transaction, and investment data,” so you can ask how your money is spread or what changed this month. OpenAI is clear about the limits: ChatGPT “cannot see full account numbers or make any changes to your accounts,” and it “is not a replacement for professional financial advice.” A general assistant with web search can also read an annual report, sum up an earnings call or compare two funds’ holdings in minutes. Check every number it gives you against the document itself; our guide to what ChatGPT can and can’t do with a brokerage account covers where it stops.
An AI score rates stocks with a model trained on the past. Danelfin’s AI Score, for example, “estimates the probability of beating the market over the next three months,” built from fundamental, technical and sentiment indicators, and Danelfin calls it “one input to a decision, not a recommendation.” That is the right way to use any score: as a probability learned from history, on a horizon of a few months, which is short for money you mean to hold for years. Screeners, chart tools and the AI assistants that place stock orders are compared in our guide to AI stock trading.
The Amplify AI Powered Equity ETF (AIEQ) has traded since October 17, 2017. It tracks the AI Powered Equity Index, whose stocks are chosen by a model from EquBot that uses IBM Watson to “analyze data points across news, social media, industry and analyst reports, and financial statements on thousands of U.S. companies.” Over the five years to October 31, 2026, the fund returned 3.13% a year at net asset value, against 11.93% a year for SPY, a fund that tracks the S&P 500, by each fund’s own figures. One fund proves little either way. It is still a useful reminder that reading more data is not the same as picking better stocks.
Three things change, and each one rewards patience and spread over cleverness.
Hendrik Bessembinder studied every common stock in the CRSP database, a standard research record of US stock prices, since 1926 (Journal of Financial Economics, 2018). Most had lifetime buy-and-hold returns below one-month Treasury bills. The best-performing 4% of listed companies explained the net gain of the whole US stock market, while the rest, taken together, matched Treasury bills. He notes that this helps explain why poorly diversified active strategies most often trail the market. For an AI picker that holds a few dozen names, the danger isn’t only owning losers. It is missing the handful of companies that end up carrying everything.
Laurent Barras, Olivier Scaillet and Russ Wermers separated luck from skill across actively managed mutual funds (Journal of Finance, 2010). They found that 75% of funds had zero alpha: their stock picking earned back their expenses and nothing more. Skilled funds made up a significant share before 1996 and had almost vanished by 2006. Eugene Fama and Kenneth French reached a similar verdict the same year: few funds earned enough above their benchmark to cover their costs. An AI picker faces the same test as a human one, against the same index fund.
Over decades, the expensive mistakes tend to be human ones: selling after a fall, chasing what just went up, trading too often. In Brad Barber and Terrance Odean’s study of 66,465 US households (Journal of Finance, 2000), the average household turned over 75% of its stock portfolio each year, and from 1991 to 1996 those who traded most earned 11.4% a year while the market returned 17.9%. Automation helps here more than prediction does. A schedule invests whether the news is good or bad, and rebalancing trims what rose and adds to what fell, the opposite of what fear and excitement suggest.
Whether AI can forecast prices well enough to change any of this is a separate question, with sobering answers; our guide to whether AI can predict the stock market goes through the research.
The same kinds of models can sit behind both, but the jobs are different. Investing asks what to own for years. Trading asks when to get in and out, often within days or hours.
| AI investing | AI trading | |
|---|---|---|
| Horizon | Months to decades | Minutes to weeks |
| What gets automated | Allocation, rebalancing, deposits, tax-loss harvesting | Entries, exits, stops and position size |
| What the AI is asked to do | Research, rate and keep you on plan | Read charts and news, test rules, place orders |
| Leverage | Usually none: a portfolio of index funds owns what it holds | Common on futures and perpetual futures: at 10x, a 10% move against you erases the margin |
| How it goes wrong | Slowly: a poor mix lags for years | Fast: a bug or too much leverage can lose the stake in a day |
| Who answers for it | Usually a registered investment adviser with a fiduciary duty | Often you: the tool runs your rules in your account |
| What to measure it against | An index fund, over years | Buy and hold, and its own worst drawdown |
They fail differently, too. An investing algorithm can rebalance at a bad moment, which is why the SEC’s staff asks robo-advisers to warn that the algorithm “may not address prolonged changes in market conditions.” A trading rule fails faster: leverage turns an ordinary move into a liquidation, as our guide to perpetual futures shows with real numbers. If what you want is software that reads a chart and acts on a rule, start with what an AI trading agent is and algorithmic trading for beginners; for AI agents that trade inside brokerage accounts, see agentic trading. For AI trading as a whole, see what AI trading is.
US regulators treat a robo-advisor as an investment adviser that happens to use software, with the same duties as a human one. The SEC’s staff put it plainly in 2017: robo-advisers “are subject to the substantive and fiduciary obligations of the Advisers Act,” the law that governs investment advisers. Since then, the SEC’s cases have come back to one theme: say what the software does, and do it.
| Date | What happened | What the SEC said |
|---|---|---|
| February 23, 2017 | SEC staff guidance for robo-advisers, with an investor bulletin | Firms should explain the algorithm, its assumptions and limits, and how closely people oversee it. The bulletin asked investors to weigh human contact, the information used, the investing approach and the costs |
| December 21, 2018 | The SEC’s first cases against robo-advisers: Wealthfront ($250,000 penalty) and Hedgeable ($80,000) | Wealthfront said it would watch accounts for wash sales and didn’t; they occurred in at least 31% of accounts in its harvesting program. Hedgeable compared its returns with rivals’ using under 4% of its accounts, ones with above-average returns |
| February 10, 2022 | Wahed Invest ($300,000 penalty) | It advertised its own funds before any existed, and promised to rebalance accounts periodically without doing it |
| June 13, 2022 | Three Schwab subsidiaries agree to pay $187 million to harmed clients | Schwab Intelligent Portfolios held cash that Schwab’s own data showed would earn clients less in most market conditions, and didn’t tell them, while Schwab earned the spread on that cash by lending it through its bank. Schwab also agreed to an independent review of the robo-adviser’s disclosures |
| July 26, 2023 | Proposed rules on “predictive data analytics” | Firms would have had to find and neutralize conflicts in technology, AI included, that steers investors. The SEC withdrew the proposal on June 12, 2025 |
| March 18, 2024 | “AI washing” cases: Delphia ($225,000) and Global Predictions ($175,000) | Both made false and misleading statements about their purported use of AI |
| October 23, 2026 | Exam priorities for fiscal year 2027 | Firms offering automated investment advice are a priority, with checks on whether what they say matches what their algorithms do |
Every settled order above was agreed without the firm admitting or denying the SEC’s findings. The AI washing orders show what the SEC looks for. According to the SEC, Delphia claimed it put “collective data to work to make our artificial intelligence smarter so it can predict which companies and trends are about to make it big and invest in them before everyone else,” when it did not have the AI and machine learning it described. Global Predictions, the SEC found, falsely claimed to be the “first regulated AI financial advisor” and misrepresented that its platform provided “[e]xpert AI-driven forecasts.” The SEC’s enforcement director at the time, Gurbir Grewal, drew the line: “if you claim to use AI in your investment processes, you need to ensure that your representations are not false or misleading.”
The vocabulary keeps moving; the rules don’t. In its fiscal 2027 priorities, the SEC’s exam staff writes of firms “incorporating super intelligence (SI) into their businesses,” including generative SI tools for “advisory and portfolio management functions.” It says it will check whether firms “continue to comply with existing, technology-neutral obligations.” For outright fraud that uses the word AI, see is AI trading legit?
Use the regulators’ own list as your checklist. A tool worth your money answers each of these questions in plain English, in writing.
I’m considering [name of the service]. Here are its Form CRS and its adviser brochure. In plain English, tell me: what its algorithm decides and what people decide; what information about me it uses, and how I update it; every conflict of interest it discloses, including any cash it holds in my portfolio and any of its own funds; whether any performance it shows is hypothetical or backtested, and over which accounts; and what it says it does in stressed markets. Quote the passage behind each answer, and say plainly when the documents don’t answer a question.
We build Ethy, so read this part knowing that. Ethy is an AI trading agent you ask before you trade. It is not a robo-advisor: it doesn’t build or rebalance a long-term portfolio, and it doesn’t hold shares for you at a broker. What it does for an investor sits around the decision:
Its trading side is Hyperliquid perpetual futures: crypto, plus stocks, indices, commodities and FX listed on trade.xyz. Those are leveraged contracts on a price, with no shares, votes or dividends, so they are tools for trading, not for a retirement account. What tokens tied to a company’s shares give their holders is in our guide to tokenized stocks. Ethy is software for research and analysis, and it doesn’t give investment advice.
It is as safe as the firm and the portfolio behind it. A registered robo-advisor owes you a fiduciary duty, but your money still rises and falls with the markets, and software can fail the way the SEC’s cases show. Look the firm up on the SEC’s IAPD site or FINRA’s BrokerCheck, read its Form CRS, and be wary of any product that says its AI can see what comes next.
On their own pages on November 15, 2026: Betterment’s Digital plan is $5 a month, or 0.25% a year once you set up deposits of $200 a month or reach a $24,000 balance. Wealthfront charges 0.25% a year for its automated portfolio. Vanguard Digital Advisor says “you’ll pay no more than $20 per $10,000 annually” for its all-index option, with no advisory fee for the first 90 days (fund costs still apply). Prices change, so check each firm’s own pricing page before you sign up.
Not on the record so far. Research finds that most professional stock pickers don’t beat their benchmark once luck and costs are counted, and AI pickers face the same test. AIEQ, an ETF picked by AI since 2017, returned 3.13% a year over the five years to October 31, 2026, against 11.93% for SPY, an S&P 500 fund. AI can still help you research and stay disciplined.
A robo-advisor is one kind of it. The SEC describes robo-advisers as registered investment advisers that use computer algorithms to manage money online with limited human contact, and those algorithms are often rules, such as a target mix that is rebalanced when it drifts. AI investing also covers assistants that read filings and your accounts, AI stock scores, and funds whose holdings a model picks.
On its own, it can explain them, not manage them. ChatGPT’s Finances feature links your accounts through Plaid and reads balances, transactions and investment data, but OpenAI says it can’t make any changes to your accounts and is not a replacement for professional financial advice. Use it to ask questions about what you own, and check its numbers against your statements.
It is the term for claiming to use artificial intelligence that a firm doesn’t use, or doesn’t use the way it says. In March 2024 the SEC settled charges against two investment advisers, Delphia and Global Predictions, over false AI claims, with $400,000 in penalties between them. Its exam staff now checks whether firms describe their algorithms fairly and accurately.
No. Ethy is an AI trading agent, not a robo-advisor: it doesn’t build or rebalance a portfolio, and it doesn’t hold shares for you at a broker. It helps around the decision, with why a market moved, a chart read on four timeframes, backtests against buy and hold, alerts, and scheduled buys of tokens on Base. It is software for research and analysis and doesn’t give investment advice.
Trading involves risk, including the loss of your money. Ethy AI is a research and trading tool and does not give investment advice or recommendations. Backtests are simulations on historical data and do not predict future results.
Deploy your AI trading agent in one click.
