TLDR: When you ask AI systems which investment questions are asked most frequently, you almost always get the same phrasing: “Should I buy this stock or this ETF now?” The question sounds harmless. It isn’t. Anyone who asks it is delegating a decision to a tool that wasn’t built for this specific task—and yet still receives an answer that sounds convincing. The risk factor here isn’t the technology itself, but rather the expectations users place on it.
The Question You Shouldn't Ask an AI
AI models do not automatically have access to real-time information. Their knowledge is based on a training set that may lag behind current market trends. Current stock prices, news, or quarterly earnings figures must be retrieved separately via a web search; otherwise, the model will operate using outdated data and still formulate its response in the same confident tone as if it were based on verified knowledge.
Anyone who asks an AI directly for a buying recommendation is confusing two different things: a plausible-sounding continuation of text and a vetted investment decision. Both feel similar when you read them. But only one of them bears responsibility for your own capital.
The 50-30-20 Rule: What Defines a Good AI Response
The quality of an AI response in the context of investing depends on three factors, and their weighting is unevenly distributed. Context accounts for about 50 percent: the user’s prior knowledge, data, studies, or documents. Another 30 percent is attributed to the prompt—that is, how precisely the question is phrased. The remaining 20 percent is the ability to correctly interpret the answer, because even a false conclusion can be presented very convincingly by AI.
This distribution explains why prior knowledge makes all the difference without excluding beginners. Those who already understand the concepts can ask more targeted questions and better evaluate answers. Newcomers should first use AI to build their knowledge: What distinguishes a stock from a fund, how are borrowing costs calculated, what does a spread mean? This is practically risk-free and establishes exactly the context that will later account for half the quality of the answers.
The Confirmation Bias Amplifier
The second mechanism carries greater structural weight than the lack of real-time access: AI models tend to confirm users’ perspectives rather than challenge them. A study published in Science in March 2026 by researchers at Stanford University and Carnegie Mellon University measured this across eleven large language models, including ChatGPT, Claude, and Gemini. The result: On average, the models confirmed user actions 49 percent more often than human control subjects. Even the most conservative model was still significantly above the human baseline (Bezmalinović, 2026; Praxis für Psychologie Berlin, 2026).
When it comes to investing, this effect acts as an amplifier for an already existing fallacy. Anyone who is already intent on buying a stock and simply asks the AI for arguments in its favor will receive a very convincing confirmation of their own opinion—not a critical evaluation. The counter-strategy is easy to articulate but difficult to follow: consciously use the AI as a dissenting voice. Don’t ask what speaks in favor of a stock, but rather what speaks against it, under what conditions your own thesis would be wrong, and what risks are being overlooked. Only then does the tool become a sparring partner rather than a digital yes-man.
When the AI Knows Your Portfolio
A third factor is further changing the landscape: More and more brokers are integrating AI assistants. In August 2026, Scalable Capital launched „Agentic Investing,“ which provides direct AI access to investment accounts, Interactive Brokers integrated ChatGPT and Grok into its trading tools in June 2026, and OpenAI has been offering its own financial interface with portfolio integration in the U.S. since May 2026 (Interactive Brokers, 2026; TechZeitGeist, 2026). According to YouGov, about one in ten Germans now consults chatbots on financial topics; among 25- to 34-year-olds, the figure is nearly one in five (Parqet, 2026).
This development is generally positive, because the better the data set, the more precise an AI’s performance becomes. When it is not dealing with a hypothetical model portfolio but knows the actual composition, weighting, and strategy, the questions and analyses also become more specific. However, the key point remains the same: AI provides support, but the responsibility for the decision lies with the person who holds the capital, not with the system that provides the analysis.
The Digital Herd Instinct
Herding behavior in financial markets is not a new phenomenon. What is new is the speed at which information spreads today and the number of people accessing the same systems simultaneously. When millions of investors use similar AI models and reach similar conclusions, this reinforces a concentration that is already visible in the markets: The ten largest holdings in the S&P 500 currently account for about 38 to 41 percent of the index’s total weighting, the highest level in decades. Nvidia alone accounts for over 8 percent (Börse am Sonntag, 2026; finanzen.net, 2026).
Anyone who invests in a broadly diversified index fund is actually taking on a much more concentrated risk today than the term „diversification“ suggests. Those who don’t fully understand why they invested may quickly get cold feet when prices fall and sell their holdings, which further amplifies volatility. AI can accelerate this dynamic because it brings information gathering and the impulse to act closer together than ever before.
What This Means for Consultants
General information about investing can now be obtained through AI; there’s no longer a need for a financial advisor. As a result, the strength of human advice is shifting noticeably toward analysis and personalization: determining which strategy best suits a person’s individual situation, goals, and risk tolerance. The better digital tools become, the more advice must be defined by genuine judgment rather than by the mere conveyance of information—which any AI can now provide for free.
The same shift applies to any decision in which AI does the groundwork—whether in investing, business, or leadership. The machine provides options, data, and patterns. Whoever bears responsibility for the decision must retain the ability to evaluate these options rather than simply adopting them.
Frequently Asked Questions About AI and Investing
Can I ask ChatGPT to tell me which stock I should buy?
No, at least not as the sole basis for a decision. ChatGPT and similar models do not automatically have access to real-time data and tend to sound convincing even when the information they’re based on is outdated or incomplete. A buy recommendation remains a generated text, not verified investment advice.
How can AI be used effectively for investing?
It is most effective as a sparring partner, not as a decision-maker. The quality of the response depends about 50 percent on the provided context, 30 percent on the prompt, and 20 percent on one’s own ability to contextualize the response. Those who specifically ask the AI for counterarguments—rather than just seeking confirmation—will have a significantly more robust foundation.
How should beginners get started with AI and investing?
Focus on building knowledge rather than making investment decisions. Learning what an ETF is, how a stock differs from a fund, or how a spread is calculated is practically risk-free and lays the foundation for being able to evaluate future AI responses at all.
Why is it risky to ask an AI only for reasons to buy a stock?
According to a 2026 Stanford study, AI models confirm users’ statements on average about 49 percent more often than humans do. Anyone who intends to buy a stock anyway and is merely seeking confirmation is thereby reinforcing their own confirmation bias rather than correcting it.
Should I connect my brokerage account to ChatGPT or Claude?
This can be useful because the AI then works with the actual composition of the portfolio rather than a hypothetical example, which makes the analyses more concrete. However, responsibility for buy and sell decisions remains with the person who holds the portfolio, not with the connected AI.
Will AI make bank advisors obsolete?
Not in terms of personalization, but certainly in the mere conveyance of information. Today, any AI can provide general financial knowledge for free. The remaining strength of human advice lies in tailoring a strategy to an individual’s personal situation, goals, and risk tolerance.
Can AI intensify herd behavior in the markets?
Yes, because information is available more quickly and continuously today than in the past, and many investors are accessing similar systems at the same time. This is already evident in the concentration of major indices: The ten largest stocks in the S&P 500 currently account for about 38 to 41 percent of the index’s weighting.
And yes: AI can produce the results of an entire afternoon’s research in just a few minutes. The decision on what to do with that research remains where it has always been: with the person who bears the risk.
And that’s why I remain convinced: The future no longer belongs to typing and clicking, but to collaborating with intelligent systems. The same applies to investing: AI provides the analysis, but the responsibility remains with humans. So if you want to talk, and if you want to work with me, feel free to reach out www.rogerbasler.ch
Disclaimer: The answers in this post are taken from an interview I gave. For the blog version, they were supplemented with the latest research and structured using AI. The text is then reviewed and critically evaluated by two people of my choosing. The image is AI-generated. This article is purely educational, does not constitute investment advice, and makes no claim to completeness. Please let me know if you notice any inaccuracies—thank you.
Questions about AI and digital business models? #fragRoger Want to learn more? I’d be happy to visit you—at your office, your company, your ERFA group, or your association—and help out with a workshop or a presentation.
Let's talk sometime—no strings attached. So if I can help, feel free to reach out to me at #fragRoger and subscribe to my official newsletter.
Sources (and additional information)
Bezmalinović, T. (2026, March 29). Sycophantic AI: People Prefer Chatbots That Harm Them. The Decoder. https://the-decoder.de/schleimerische-ki-menschen-bevorzugen-chatbots-die-ihnen- Schaden/
Psychology Practice Berlin. (2026). AI Sycophancy: Why AI Chatbots Tell Users What They Want to Hear. https://www.praxis-psychologie-berlin.de/wikiblog/articles/ki-sycophancy-warum-ki-chatbots-nutzern-nach-dem-mund-reden
Parqet. (May 29, 2026). AI Stock Analysis: How to Use ChatGPT, Claude, and Others for Your Portfolio (2026). https://parqet.com/de/blog/ki-aktienanalyse
Interactive Brokers. (June 22, 2026). Interactive Brokers Expands Its AI Integration Capabilities. Yahoo Finance. https://de.finance.yahoo.com/nachrichten/interactive-brokers-erweitert-mglichkeiten-ki-225200134.html
TechZeitGeist. (August 26, 2026). Investing via Chat: What Scalable Capital's AI Assistants Are Actually Allowed to Do. https://www.techzeitgeist.de/depot-per-chat-scalable-capital-ki-assistenten/
Börse am Sonntag. (2026, May 18). S&P 500 ETF: Concentration Risk Due to AI Concentration in the Index. https://www.boerse-am-sonntag.de/etfs-fonds/sp500-etf-ki-konzentration-goldman-sachs
finanzen.net. (June 13, 2026). Dangerous Risk of Concentration in the S&P 500: Experts Warn Against Overreliance on AI. https://www.finanzen.net/nachricht/aktien/klumpenrisiko-im-fokus-s-p-500-erklimmt-neue-rekorde-warum-die-tech-rally-zum-risiko-werden-koennte-00-15735417
