Capital.com’s Prokopenya Wants To Remove the AI Vocabulary Tax

Viktor Prokopenya, the founder of Capital.com, has called on the industry to address what has been described as an "AI vocabulary tax" in financial advice, after new research from Stanford and MIT Sloan found that users' financial literacy and prompt-engineering ability materially shape the quality of AI-generated recommendations.

In a LinkedIn post commenting on the study, Prokopenya wrote: "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."

The study, titled "AI Financial Advice: Supply, Demand, and Life Cycle Implications," surveyed 1,000 US adults and used a quantitative model to simulate lifetime earnings and investment outcomes based on LLM-generated advice. It found that roughly half of adults in the UK and US have already consulted AI for financial guidance, surpassing the share who turn to human advisors. Unlike robo-advisors, however, the researchers found LLMs are not fine-tuned specifically for financial advice, leaving room for inconsistent results.

The modelling showed that users who had previously used AI for financial guidance received recommendations leading to an average of $100,000 more wealth by age 60 than those who had never used such tools. By contrast, users who struggled with basic financial concepts ended up with 4.1% lower wealth outcomes. The AI, the authors concluded, often mirrors the user's own limitations: "thin" or poorly phrased questions generate generic or overly cautious answers, and less literate users tend to receive lower equity allocations that can curtail long-term growth. The results were replicated across several models, including ChatGPT-5.2, Gemini 3 Flash and GPT-5.6 Terra, though they come from modelling rather than observed outcomes.

The findings arrive as brokers race to deploy AI assistants. Robinhood has Cortex and eToro has Tori, among other industry efforts. Prokopenya argues the real task is not just shipping the tool but making it capable of spotting a weak question and interrogating the user to arrive at a better one.

Should the industry fail to close that gap, he suggested, a new marketplace could emerge around selling the right prompts. Niche prompt marketplaces such as PromptBase already exist, offering pre-written prompts for investment-related queries. Until AI learns to coach users as much as it informs them, the level playing field promised by low-cost AI advice will remain out of reach.