AI Investing: Genuine Next Step or Robo Advice With a 2026 Sticker?

At the end of my robo investing update, I teased this post. Robo advisors automate the boring part, but AI supposedly goes further: it picks, it times, it trades for you. Whether AI investing in Belgium represents a true advancement in trading, just marketing, or even worse, a fast route to zero, that is what this post is about.

Before I get into it: I’ve been running my own experiment on AI investing in Belgium itself. I use Claude (Anthropic’s AI) via a desktop tool called Cowork, and part of what I use it for is investment research. Scanning earnings, asking what a certain valuation implies, and thinking through portfolio decisions. I’ll be transparent about what that’s actually like. Here’s a small hint: The sales pitches are just that, pitches 😉

First, Let’s Unravel What “AI Investing” Actually Means

When people say AI investing in 2026, they’re usually talking about one of three completely different things. Each carrying a different risk profile and a different level of hype attached to it:

• Type 1: Investing in AI companies. You want exposure to the AI industry itself: Nvidia, ASML, the cloud platforms, the chipmakers. This usually means an AI-themed ETF. It’s a legitimate thematic investment. It’s also just a sector bet with extra steps.
• Type 2: AI tools for investment research. You use an AI assistant to help you research companies, understand financial reports, and think through decisions. Claude, ChatGPT, Perplexity, and a handful of specialized tools like Danelfin fall into this bucket. The AI doesn’t execute trades; it helps you think.
• Type 3: AI-powered portfolio management. A platform that claims its AI model actively picks and trades your portfolio to beat the market. This is where most of the hype lives, and most of the danger.

As you can read, all very different beasts. Most of the marketing deliberately blurs the lines between them, which is exactly why the pitch sounds so compelling until you dig into what you’re actually buying.

The Hype, The promises

The pitch for Type 3 goes roughly like this: traditional fund managers can’t process millions of data points simultaneously, but AI can. It reads earnings calls, tracks sentiment, spots patterns in price data, and reacts in milliseconds. You, the retail investor, now have access to the same edge that hedge funds use. For €X or €Y per month.

It sounds good. It also sounded good when people said robo advisors would replace financial advisors in 2014.

Platforms like Danelfin give an “AI score” to individual stocks based on technical and fundamental signals. Others run fully automated portfolios claimed to be managed by machine learning models. A few have caught on: the Forbes-backed Q.ai in the US is probably the best-known name in the category. The track records are short, mixed, and almost universally worse than just buying VWCE.

The Reality Check

No publicly available AI model has consistently beaten a broad index fund over a long time horizon — and the AI investing Belgium picture is no different from anywhere else in the world on this front.

The AI ETFs themselves are a useful case study. Thematic funds that invest in AI companies have had strong recent performance, riding the Nvidia wave, the cloud infrastructure buildout, and the hyperscaler capex cycle. But over longer horizons, most of them have underperformed the S&P 500. The iShares Future AI and Tech ETF (ARTY) has been behind the broad market for most of its existence. The Robo Global Robotics and Automation Index ETF (ROBO) has underperformed the S&P 500 for much of its existence since its 2013 launch.

That’s thematic ETFs investing in AI. The AI tools that claim to use AI to beat the market have even shorter track records and less independent verification.

The reason is pretty simple: if an AI model reliably generated alpha, the people running it would not be selling you access for €9.99 a month. They’d be running a hedge fund and keeping it to themselves. Just like how a successful trader most likely isn’t on Instagram or TikTok selling a book or course to “help” you become successful too.

AI Investing Options for Belgian Investors

For the typical Belgian FIRE investor, the AI investing Belgium reality looks like this:

• AI-themed ETFs are accessible through any standard broker: DEGIRO, IBKR, you know, the usual options. The main funds are UCITS-compliant, so TOB applies, but they’re easy to buy. The question is whether the thematic bet makes sense when a good chunk of the AI exposure in these funds already sits in VWCE via Nvidia, Microsoft, and Alphabet.
• AI research tools are available to anyone with an internet connection and a few euros a month. Claude, ChatGPT, Perplexity. These are useful for research when used correctly (more on that below).
• AI-managed portfolios (Type 3 platforms) don’t yet have a real presence in the Belgian market. Most operate out of the US or the UK. A few are technically accessible to Belgian investors, but they fall outside Belgian regulatory oversight, which is a risk on top of the performance risk.

My Own Experience Using Claude for Investment Research

I’ve been using Claude for a few months as part of what I’d call my higher-risk investment research. Not the boring VWCE core (that runs on autopilot). The messier stuff: individual stocks, pre-IPO platforms, newsletter tips that need stress-testing.

A concrete example. I’m holding CAI (Caris Life Sciences), a precision oncology company I bought at $25.18. It dropped to around $15.94. I asked Claude whether to hold. The answer came back with a proper breakdown: a director had just bought $501,000 of his own money in shares at $16.15, Q1 revenue was up 79%, and the analyst consensus target was $28, based on 13 analysts. But also: the stock traded at 136x forward earnings, which is exactly the multiple that gets punished hard on any disappointment, and an unresolved Medicare reimbursement ruling was still hanging over it.

Useful research. What I did with it was sit on my hands and keep holding. Which may or may not be the right call.

The same pattern plays out on most questions. When I asked about RKLB (Rocket Lab) and ASTS (AST SpaceMobile), I got a solid analysis of why waiting made more sense than buying right then. Claude also flagged something I’d half-ignored: I was already carrying margin debt on two other losing positions, including GTLB. Ouch, good point AI.

What I haven’t done is actually buy something because Claude told me to.

On one hand because it doesn’t quite work that way: it gives you the picture and leaves the decision to you.

But on the other hand because there’s a practical problem I keep running into: I have a Cowork project for the high-risk part of my portfolio, a separate chat project called “Portfolio Analysis,” and individual research sessions that don’t know about each other. Ask Claude about space stocks in one window, and it has no idea you’re already carrying margin debt in another. The AI doesn’t have your full picture unless you paste it in every single time. Which is work.

So, there you have it: a pretty honest summary of where AI investing in Belgium stands in 2026 for retail investors, from my point of view. AIs like Claude (and Perplexity and grok as well, most likely) are very well-read research assistants who can read an earnings call in two minutes and never get irritated when you ask a basic question. That’s actually useful. It’s not a portfolio manager.

My Summary: Where AI Might Help (And Where It Doesn’t)

Useful: Research. Reading documents faster. Explaining financial concepts. Stress-testing an assumption. Working through tax implications. Comparing options without bias toward whichever one you already prefer. If you’ve been manually trawling through fund factsheets to compare TER and portfolio overlap, an AI assistant saves you real time.

Not useful (yet): Market timing. Stock picking with a verifiable edge over an index. “Should I buy this now or wait?” is a question, where the honest answer is that nobody knows, and the AI is pattern-matching on historical data that may not predict anything about tomorrow.

Actively dangerous: Treating AI output as financial advice without understanding it.