QWE AI Academy Here’s a number that should reframe the entire index-funds-vs-individual-stocks debate before it starts: according to Hendrik Bessembinder’s research at ASU, just 86 stocks accounted for $16 trillion – half of all US stock market wealth creation – over the past 90 years, and the remaining 96% of stocks collectively matched one-month T-bills.
Read that again. If you pick stocks at random, the base rate says you’ll roughly tie a savings account.
Most tutorials on index funds vs individual stocks hand you a pros-and-cons list. This one won’t. Instead, we’ll look at the actual skewness math, the current SPIVA numbers, and – since this is an AI academy – how to use tools like ChatGPT to stress-test your own picks against the base rate rather than as a magic stock picker.
The problem with every “pros and cons” article
The standard framing goes: index funds are safer and cheaper, individual stocks have higher upside, choose based on your risk tolerance. It’s not wrong. It’s just useless, because it treats the two options as symmetrical bets with different volatility.
They aren’t symmetrical. The distribution of stock returns is violently skewed – a small number of massive winners drag the whole average up, while the median stock underperforms cash. Between 1926 and 2015, only 43% of equities returned more than Treasury Bills over their lifetime. The “average” 10% market return you hear about is real, but almost nobody experiences it by owning a handful of names.
So the real question isn’t “do I want more risk?” It’s “do I think I can consistently identify the top 4% before everyone else does?”
What the current data actually shows
Professional stock pickers – people who do this full-time with Bloomberg terminals and teams of analysts – mostly lose to the index. The 2024 SPIVA scorecard from S&P Dow Jones is brutal on this point.
| Category (2024) | % of active funds that lost to their index |
|---|---|
| US large-cap | 65% |
| US equity funds (15-year window) | Every single category – 22 out of 22 |
Over the 15-year period ending December 2024, there were no categories in which a majority of active managers outperformed – zero out of twenty-two. If pros can’t do it consistently over 15 years, the prior on you doing it in your brokerage account should be humble.
A caveat though: small-cap active strategies had their best year since SPIVA started tracking, with 70% of actively managed small-cap strategies outperforming the S&P SmallCap 600 Index in 2024. Sounds like a win for stock pickers. It isn’t – the 2024 small-cap anomaly traces to large-cap tilt inside those funds, not stock-picking skill. Those managers were winning because they were secretly holding large-caps during a mega-cap year.
The approach I actually recommend: core + curiosity
Purists on both sides will hate this, but a split portfolio makes practical sense: put the bulk of your money in a broad index fund and use a small “learning allocation” – say 5-10% – for individual stocks you actually want to research.
Why not go 100% index? Because you’ll never learn anything about companies, valuation, or your own psychology if you never own a single stock. Why not 100% stocks? Because Bessembinder’s math plus SPIVA together say the odds are stacked against you.
For the core, two funds keep coming up:
- VOO (S&P 500) – 0.03% expense ratio as of late 2025. Concentrated in mega-caps: top 10 holdings were 40.7% of assets at year-end 2025, with Information Technology at 34.4% of the fund. Recently crossed $860 billion, overtaking SPY as the world’s largest ETF by assets in 2025.
- VTI (Total US market) – same 0.03% headline fee as of late 2025, but there’s a hidden kicker. Securities lending fees get returned to fund shareholders, which means the effective cost of ownership is lower than the stated 0.03%. VTI benefits much more from securities lending than VOO does because small-cap stocks are harder to borrow and therefore command higher lending fees. Short sellers targeting obscure small-cap names pay a premium, and that premium flows back to VTI shareholders. VOO holds only large-caps, which are easy to borrow, so almost no lending revenue comes back to you.
That last point is a footnote in most comparisons but it’s real money over decades.
Where AI actually helps (and where it doesn’t)
One public test: a blogger gave ChatGPT a single prompt to pick 5 US stocks, invested $5,000 equally across them, with a matching $5,000 in VOO as the benchmark. Over 12 months starting January 2025, the AI portfolio returned 23.2% versus 20.5% for the index. Five stocks, one year, one prompt – the sample size tells you nothing about skill. It does illustrate something: AI is roughly as good as any other retail heuristic, which means indistinguishable from luck at small samples.
Better move: flip the prompt. Instead of “pick stocks for me,” ask AI to argue against your existing thesis:
I'm considering buying [TICKER]. My thesis is [your reason].
Do three things:
1. Steelman the bear case using publicly known concerns.
2. Identify what would have to be true for my thesis to work.
3. Given Bessembinder's finding that only ~4% of stocks generate
all market wealth above T-bills, estimate honestly whether this
name has characteristics of that group.
Do not tell me what to do. Just pressure-test the reasoning.This works because LLMs are terrible at prediction but reasonable at generating counterarguments – which is exactly the cognitive step retail investors skip.
Pro tip: Never let ChatGPT hallucinate prices, earnings figures, or recent news. It doesn’t have reliable real-time data. Use it for logic and stress-testing questions; use a broker or SEC EDGAR for numbers.
A real-world example: the two portfolios thought experiment
Person A puts $10,000 in VTI and forgets about it for 20 years.
Person B picks 8 stocks – a reasonable-sounding basket, maybe some tech, some healthcare, some consumer names. They spend 2 hours a week reading earnings reports.
Both start the same year. What does the math actually predict?
Person A gets the market return, minus 0.03%. Roughly guaranteed to match whatever the index does. No effort.
Person B is running an 8-stock lottery ticket against a distribution where only 4.3% of all stocks – 1,092 companies out of 25,967 – were responsible for the entirety of net wealth generated by the US market above T-bills. The expected value of a random 8-stock sample: match T-bills, with wide variance. To beat VTI, Person B needs at least one of those 8 to be a genuine outlier – and needs to hold it through the 50%+ drawdowns that outliers all experience on the way up.
Person B’s real return isn’t just the numerical return. It’s the return minus the value of 2 hours per week for 20 years. That’s roughly 2,000 hours. If those hours are worth $50 to you, Person B starts the race $100,000 behind.
Four things most comparisons miss
- The concentration you’re avoiding is already in your index. VOO isn’t 500 equally-weighted bets – at year-end 2025, Information Technology made up 34.4% of the fund, and the top 10 holdings accounted for 40.7% of total assets. When you buy “the market,” you’re buying a tech-heavy bet whether you meant to or not.
- International doesn’t fix the skewness problem. About 61% of non-US stocks underperformed Treasury bills in the 1990-2018 period, and just 1.3% of stocks contributed all of the net gain internationally. If anything the distribution is more extreme abroad.
- Track the counterfactual. If you insist on picking stocks, keep a spreadsheet comparing your actual portfolio to what you’d have if that money went into VTI on the same dates. After 3 years you’ll have real evidence about whether you have a skill or a hobby.
- Watch persistence data, not annual data. One good year proves nothing – the SPIVA research shows that even managers who outperform in year one rarely repeat. Annual rankings are noise; decade-long records are signal.
FAQ
Can I use ChatGPT to pick individual stocks?
Technically yes, practically no. It can’t access real-time prices, and the one public experiment showing a 2.7 percentage point edge over VOO was a single 5-stock sample – statistically meaningless. Use it to argue against your ideas, not to generate them.
If index funds are so obviously better, why does anyone buy individual stocks?
A small number of people really do get rich picking Amazon in 1998 or Nvidia in 2016, and those stories get told loudly. Owning individual companies is also more engaging in a real way than owning a fund – there’s psychological value in following a business, even when it costs money in returns. And most investors simply haven’t seen the Bessembinder numbers laid out plainly: when you know 96% of stocks collectively matched T-bills, the appeal of stock-picking shifts fast.
What about that 2024 small-cap active outperformance I read about?
Real, but misleading. The winning small-cap funds were mostly holding large-caps in disguise during a year mega-caps ran hot. It’s a style story, not a stock-picking story, and it’s not something you can replicate by picking small-cap stocks yourself.
What to do this week
Open your brokerage. Look at your current split between index funds and individual stocks. Write down the percentage. Then write down what you honestly think that percentage should be based on the numbers above – and if the two don’t match, make one trade that closes half the gap. Not all of it. Half.
Then set a calendar reminder for 90 days from now to close the other half or explain to yourself why you didn’t.