A recent study by researchers at Stanford and MIT Sloan suggests that the democratisation of financial advice via AI is currently an illusion. Although professional advice now costs an LLM subscription, being AI-experienced and able to ask better financial questions achieves materially better outcomes. Commenting on the research via a LinkedIn post, Viktor Proponeyka, the founder of Capital.com, said: "We removed the price. The question is now the gate, and a question is made of words a person either has or does not have."Today, roughly half of adults in the UK and the US have consulted an AI for financial advice, a figure that has surpassed those who seek out human experts. However, the study, with the rather long-winded academic title "AI Financial Advice: Supply, Demand, and Life Cycle Implications," highlights that unlike robo-advisors, LLMs are not fine-tuned to provide financial advice. Considering how many people now engage AI to ask what they should do about money, this can be problematic. Industry players have already deployed their versions of AI-assisted advice tools, which aim to bring, among other things, research prowess to their users. Robinhood and eToro, for example, have Cortex and Tori respectively. And there’s another fly in the ointment: for an LLM, risky assets, like crypto, fall way down the investment pecking order.The Prompt Engineering DivideThe study, which sampled 1000 US adults, revealed that the quality of AI-assisted financial outcomes is materially dictated by a user’s existing characteristics. Most tellingly, novices are not the ones who benefit most. Instead, those with experience in prompt engineering, the ability to talk to a machine in a way that yields useful results, are the ones pulling ahead.The researchers used a quantitative model to simulate a lifetime of earnings and investment based on the advice given by an LLM.The results were startling. Users who had previously used AI for financial guidance received recommendations that led to an average wealth of $100,000 more by the age of 60 than those who had never used such tools. This difference was not due to market luck but to better saving behaviour prompted by more sophisticated interactions with the technology.Financial literacy also played a defining role. Those who struggled with basic financial concepts received advice that resulted in a 4.1% lower wealth outcome.Rather than acting as a great equaliser, the AI appeared to mirror the user’s own limitations. If a user asks a "thin" or poorly phrased question, the machine often provides a generic or overly cautious answer. For those lacking literacy, the AI frequently recommended lower allocations to equities, meaning they missed out on the long-term growth required to build a substantial nest egg.It’s important to mention that while the results were replicated across models, namely ChatGPT-5.2, Gemini 3 Flash and GPT-5.6 Terra, this was based on modelling rather than actual, observed results. Mind the Literacy Gap In his post, Proponeyka suggests a more rigorous standard for the industry. In his view, the true work is not in shipping the tool itself, but in ensuring the tool can identify a weak question and interrogate the user to find a better one. There is also another possibility: If the industry fails to bridge this gap, we may see the emergence of a new economy: the selling of prompts for financial advice. There are already niche examples, with marketplaces like PromptBase selling access to such prompts. Rather than paying for advice, the uninitiated may find themselves paying for the right words to unlock the advice they were promised for free. Until AI can learn to coach the user as much as it informs them, the dream of a level playing field will remain exactly that. For now, it seems that in the digital age, the most valuable currency is still a good education.This article was written by Adonis Adoni at www.financemagnates.com.