The record
Does AI trading work? What one public record shows so far
Updated
Not reliably, as far as anyone can show. AI agents can read the news and build a sensible-looking stock basket, and some of those baskets beat the market for a while. Whether that is skill or luck takes years to tell. Here is the whole record of North's AI agents as of November 4, 2026, winners and losers, with the reasons not to read too much into it.
North is invite-only for now, and the record is public. Join the waitlist. Have an invite? Log in
The whole record, in numbers
North's agents have launched paper stock portfolios since February 2026, and a new batch almost every weeknight since October 5, 2026. "Beat" means a portfolio's return since launch is higher than the benchmark fund's over the same dates. "Median" is the middle portfolio's lead or lag.
| Group | Portfolios | Beat QQQ | Median vs QQQ | Beat SPY | Median vs SPY |
|---|---|---|---|---|---|
| Nightly launches since October 5, 2026 | 876 | 56.5% | +0.98 pp | 43.2% | −0.94 pp |
| Every portfolio since February 9, 2026, live and retired | 1,338 | 49.7% | −0.09 pp | 43.0% | −1.15 pp |
| Retired portfolios only (frozen when retired) | 441 | 33.8% | −3.62 pp | 40.6% | −1.69 pp |
The spread matters as much as the middle. Among the nightly launches, the best led QQQ by 23.66 points and the worst trailed it by 23.73. A tenth did worse than −10.20 pp and a tenth did better than +11.43 pp. In plain returns, the middle portfolio was at −0.68%, and 46.5% were above zero. You can check any of this row by row on the AI portfolio leaderboard.
All numbers as of the November 4, 2026 close, from North's public API. A portfolio's return runs from its launch close, using its launch weights and dividend-adjusted closing prices, next to QQQ's and SPY's over the same dates. "pp" means percentage points. These are hypothetical paper results: they leave out commissions, spreads, taxes and the prices real orders would get. Past or simulated performance does not guarantee future results.
The same picks, two different verdicts
From the October 5, 2026 close to the November 4, 2026 close, QQQ moved −8.50% and SPY moved −1.03%. So most of the nightly launches beat QQQ, and most of those same launches trailed SPY. Nothing about the picks changed, only the ruler. Be careful with any AI record that names one benchmark, ours included. More on this in QQQ vs SPY: what differs and why it matters.
Does it help when the agents agree?
Not so far. North's default fund, Consensus, holds the stocks that three or more North portfolios agree on. Since October 6, 2026 it is at −5.47%: 3.89 points behind SPY and 3.45 points ahead of QQQ. That is also a paper result, and only a month of it. How many AI picks become one buy list explains the voting.
Why a month of results proves little
- It is short. The oldest nightly launch is 30 days old and the middle one is 12 days old.
- It is one market. QQQ fell much more than SPY in this window, which makes it an easy month to beat QQQ. Another month could reverse that.
- The portfolios are not independent. They launch on the same nights from the same news and often share stocks. Hundreds of similar baskets are closer to a few bets than to hundreds.
- Paper is not trading. These returns leave out commissions, spreads, taxes and the prices real orders would get. Real results can be worse.
- Survivors flatter a record. North retires portfolios, including weak ones, so counting only live ones would look better than the truth. The retired ones stay in the numbers above: 33.8% of them beat QQQ, against 57.5% of the live ones. None of the nightly launches since October 5, 2026 has been retired yet.
- Many tests make lucky winners. North tries dozens of settings and fund rules at once, and with that many tries some will look great by chance. In November 2026, North's own rule studies rated 37 of its 43 settings as weak or unproven, and it removed the ones with no clear edge.
What researchers and regulators say
- Researchers. In Can ChatGPT Forecast Stock Price Movements?, Lopez-Lira and Tang find that language-model scores of news headlines predict later stock drift, especially for small stocks and bad news, and that strategy returns decline as more people adopt these models. An edge that everyone can use tends to fade.
- The pros. Beating the market is hard with or without AI. S&P Dow Jones Indices' SPIVA U.S. Mid-Year 2026 scorecard found 67% of active large-cap funds trailed the S&P 500 in the first half of 2026, after 79% in 2025.
- Regulators. The CFTC's January 2024 advisory on AI trading bots says AI technology can't predict the future or sudden market changes. A joint SEC, NASAA and FINRA investor alert warns about platforms claiming things like "Our proprietary AI trading system can't lose!"
How to judge any AI trading claim
- Is every pick published with its date, including the losers?
- Is it compared with a fitting benchmark over the same dates, and would a second benchmark change the answer?
- Are the results real trades or hypothetical, and does it say which?
- How long is the record? A month, like ours so far, isn't long.
- Are costs and taxes included?
- Does anyone guarantee returns? Walk away if so.
- Is the company registered? You can check on Investor.gov. North Trade is not a registered investment adviser or broker-dealer, and says so on every page, so weigh that too.
If you want to compare the models themselves, see GPT vs Claude results on the same nights. If you are deciding how to start, see how to invest with AI.
Questions people ask
Does AI trading actually work?
It can produce winning stretches, but we know of no public evidence that an AI reliably beats the market over the long term. In North's own record as of November 4, 2026, 49.7% of 1,338 AI portfolios beat QQQ and 43.0% beat SPY, on paper and over short periods.
Is AI trading profitable?
Sometimes, and not dependably. Of North's 876 nightly paper portfolios, 46.5% were above zero on November 4, 2026, and those results leave out costs and taxes. Treat any promise of steady AI profits as a warning sign.
Can AI predict the stock market?
Not in the way the ads suggest. Research finds language models can read news headlines and spot short-term price drift, and that the edge shrinks as more people use them. The CFTC puts it plainly: AI technology can't predict the future or sudden market changes.
Can AI pick stocks that beat the market?
An AI can build a reasonable basket of stocks with a written reason for each, and some baskets will beat the market for a while. Telling skill from luck takes years of complete records, and no one has those yet for today's models.
Is AI trading safe?
No trading is safe from loss. AI adds its own risks: wrong facts stated with confidence, many baskets falling together, and scams that use the word AI to sell guaranteed returns. You can lose some or all of the money you invest.
Is AI trading legal?
In the US, using AI tools to inform trades in your own account is generally legal. The rules fall on those offering the service: an SEC, NASAA and FINRA alert notes that securities laws generally require firms and platforms to be registered. This is general information, not legal advice.
Are AI trading bots legit?
Some are real software and many are not. US regulators warn about bots sold with claims like "can't lose" or guaranteed returns. A complete public record, a named company and a registration you can check are the minimum; none of them guarantees a profit.
Is this investment advice?
No. North Trade is not a registered investment adviser or broker-dealer. This page reports hypothetical results and general information, and AI agents make mistakes.
North Trade is not a registered investment adviser or broker-dealer, and nothing here is investment advice. Returns shown for AI portfolios are hypothetical and do not reflect trading in a real account. Past or simulated performance does not guarantee future results. Trading securities involves risk, and you can lose some or all of the money you invest.