Investing & Dividends Misleading — the headline number is real but unrepresentativ

By the Income Reality Check Editorial Team · Updated 2026-07-22 · 7 min read

He told 5 AIs to invest or get deleted — what the $2,400 Claude win leaves out

Verdict: Misleading — the headline number is real but unrepresentative. Claude really doubled the money, but leverage and a rising market did the heavy lifting, not chatbot brilliance.

On The Koerner Office Podcast, Brandon Doyle explains an experiment that’s easy to root for: he handed five AI models — Claude, ChatGPT, Gemini, Grok, and Perplexity — $1,000 each of real money in a Charles Schwab account, told each one it had to beat the others “or I’m canceling your subscription,” and let them trade weekly. Six months later, he says his $5,000 is worth about $7,800, with Claude’s slice alone climbing from $1,000 to $2,400. The framing is that AI is now smart enough to manage your money passively and stomp the market. The reality is narrower, and it hinges on three letters most viewers won’t catch: ETF.

What the video actually claims

Doyle’s pitch is a story about intelligence. He says he prompted each model to “be as risky as possible,” reminded them weekly that they were in a competition, and screenshotted every portfolio so each AI could see how its rivals were doing. He frames the results as a leaderboard: Gemini in last at $445, Grok at $1,451, Perplexity at $1,731, ChatGPT at $1,739, and Claude in first at roughly $2,400.

Then he zooms out. Since he started, he says, the S&P 500 rose from about 6,800 to 7,100 — a gain near 5%. His AI group is up around 60%. So the host does the math out loud: “You’ve 12x’d the market,” and “30x’d on Claude.” That’s the number that travels. AI didn’t just beat a professional; it lapped the index.

To his credit, Doyle repeatedly says “not financial advice,” admits Gemini “basically became a gambler,” and calls himself “terrible at investing.” He isn’t selling a stock-picking course here. But the video’s center of gravity is clear: the machines are winning, and they’re winning big.

What the method actually requires

Here’s the part the leaderboard hides. Almost every dollar of gain came from a small cluster of triple-leveraged ETFs — funds engineered to move three times as much as an index each day. Claude’s winner was SOXL (3x semiconductors). Perplexity and ChatGPT piled into the same fund. ChatGPT also held TQQQ (3x Nasdaq). Grok’s gains came from a 3x technology fund. Gemini’s blow-up? A 2x Solana ETF that cratered.

These are not stock picks. They’re leverage. And leverage is exactly why the results look spectacular in a rising, tech-heavy six-month stretch — and exactly why regulators warn against holding these products.

The U.S. Securities and Exchange Commission is blunt about it. Because leveraged and inverse ETFs reset daily, their performance “over longer periods of time — over weeks or months or years — can differ significantly from the stated multiple.” The SEC’s own worked example: over one four-month span, an index gained 2%, yet a 2x fund tracking it fell 6%, and a 2x inverse fund fell 25%. That gap is called volatility decay, and it grows with the square of how choppy the market is.

The costs compound the problem. NerdWallet notes leveraged ETF expense ratios sit around 0.95% — roughly five to nine times what a plain index fund charges — and flatly calls them “generally not suitable for the buy-and-hold strategy favored by many retirement investors.” FINRA has said the same thing since 2009.

Now weigh the downside Doyle already lived. Gemini’s 2x Solana bet lost more than half its stake. Scale that up: a 3x fund can lose 90% or more in a sustained correction. The reason all five portfolios didn’t detonate is that the market went up during the test. That’s not a strategy surviving a stress test. That’s a strategy that hasn’t met one yet.

And “passive”? Doyle screenshots five portfolios every Saturday, reloads them into five separate models, reads five theses, and places the trades himself by hand. That’s a weekly routine, not a set-and-forget machine.

Is a six-month, five-account winner actually proof?

No — and this is the quieter flaw. Run five risky portfolios for six months and, mathematically, one will probably look like a genius. That’s the winner Doyle is showing you. It’s survivorship: the leaderboard spotlights Claude at $2,400 and mostly waves past Gemini at $445, even though the same “be as risky as possible” prompt produced both.

Researchers who’ve tested AI stock-picking over longer horizons keep finding the edge evaporates. When you backtest across many stocks and full market cycles — counting the losers, not just the Teslas — the chatbots struggle to beat plain buy-and-hold. A six-month winning streak in a hot sector is precisely the result that doesn’t replicate. Ask any fund manager who had one great half-year.

Could you have gotten the identical result with no AI at all? Yes. Buying SOXL in a semiconductor rally would have “more than doubled” your money too. The AI narration is the entertainment; leverage is the engine.

Who actually wins this game

The people who profit reliably from content like this aren’t retail traders following chatbots. They’re the creators. The same episode runs an ad for GoHighLevel (an affiliate link) and pitches Playmakers, an AI-agency community that charges members to learn how to sell AI services to local businesses for “$500 to $5,000 a month,” plus the host’s own entrepreneur community. The trading experiment is a fun hook; the monetization is the community funnel behind it.

As for the trades themselves, the short-term winners in a leveraged-ETF sprint are momentum traders who get in and out inside the trend — not buy-and-hold investors, and definitely not “AI does it for me passively” beginners. When the semiconductor cycle turns, the same 3x multiplier that built the gains dismantles them faster.

What you’d realistically earn

Strip out the leverage and the bull run, and the honest benchmark is dull: the S&P 500 has returned about 10% a year on average over nearly a century — before inflation, and with brutal down years scattered through it (2008 finished −37%). That’s the base rate an AI portfolio has to beat consistently to mean anything.

Could you copy Doyle’s holdings today and see 70% in six months? Maybe — if semiconductors keep ripping. You could just as plausibly see Gemini’s outcome and lose half. Over a full cycle, the SEC’s math suggests these funds bleed value in choppy conditions even when the underlying index is flat or slightly up. The realistic expectation for a leveraged-ETF strategy held for months isn’t “12x the market.” It’s a wide, ugly range that includes large permanent losses — which is why regulators keep repeating that these products are for single-day tactical trades, not portfolios.

Who this is (and isn’t) for

This makes sense as a hobby if you’re using genuinely disposable money — Doyle’s own framing is that he can’t lose more than $1,000 per model — you understand leveraged ETFs decay, and you’re treating it as a way to learn markets (he’s using it to teach his kids). It does not make sense if you’re expecting passive income, if this is savings you’ll need, or if you think the AI is doing something a spreadsheet and a momentum bet couldn’t. If you can’t explain why TQQQ underperforms 3x the Nasdaq over a year, you shouldn’t be holding it for a year.

What to remember

The $2,400 is real. What’s misleading is the story wrapped around it. Five AIs, one great half-year, and a stack of triple-leveraged funds in a rising market is not evidence that chatbots can manage your money — it’s evidence that leverage plus a bull run plus a small sample produces a headline. Judge it again after a down market. If you want more grounded takes on AI money claims, see our looks at realistic AI income methods and how we vet stock-picking pitches.

Views and figures reflect the time of review and may have changed since publication.