What is the hidden cost of investment advice on social media?

A video titled How to Invest for Beginners in 2025 accumulated 2.5 million views in two months. Another, promising viewers they will be "worth millions" if they buy a particular stock, pulls 100,000 likes in a single day. Behind these numbers are real people making real decisions about their savings. However, research suggests that those people would have been better off switching off these videos.

The median portfolio built on following financial influencer recommendations underperformed the S&P 500 by 43 percentage points over a one-year holding period. It also carried a higher risk than the benchmark across all measures tested. And there was no observable pattern (across subscriber count, view volume, verified status, or any other variable) that would allow an investor to identify reliable voices from unreliable ones in advance.

The reason for this, according to University of Tübingen Professor of Marketing, Dominik Papies, is the hype cycle. Financial influencers are not identifying opportunities ahead of the market. They are latching onto stocks that are already rising, amplifying attention through their recommendations, and leaving their followers exposed when the cycle turns. The data, drawn from 231 YouTube channels and approximately 46,000 videos, identified an unambiguous pattern: the stocks influencers recommend have typically already outperformed the market before the video is published. After the recommendation, they underperform.

Why financial influencers have attracted serious attention

The rise of social media investment advice is not a niche phenomenon. A 2021 survey found that more than half of respondents reported relying on social media for their investment decisions. Retail investors are shifting away from institutional relationships toward self-guided investment, and financial influencer content is filling that gap at scale.

Prof. Papies observed that regulators have taken notice. In India, certain categories of investment advice on TikTok are no longer legal. The European Securities and Markets Authority has issued guidelines governing what financial influencers can and cannot do. The United States examined the issue early, though attention there has quietened. The common thread is concern: these recommendations reach millions of viewers, making real decisions about retirement savings, discretionary income and long-term wealth.

Prof. Papies, who presented the research at the UNSW Marketing Analytics Symposium Sydney (MASS) 2026, put it plainly: "So they start regulating this market because it is so financially consequential with the decisions that are made there, for example, if you think about retirement savings."

Photo gallery: The UNSW Marketing Analytics Symposium Sydney (MASS) 2026

What the data actually shows

Prof. Papies and his co-author, Jan Jacobsen, assembled a dataset of 231 YouTube channels that collectively published approximately 46,000 videos between 2016 and 2022. Channels required a minimum of 10,000 subscribers and at least 10 videos, with individual videos needing more than 10,000 views and running under 45 minutes. Four asset classes were examined: stocks, cryptocurrencies, ETFs (exchange-traded funds) and commodities.

Extracting recommendations from thousands of hours of transcripts required a purpose-built methodology. Using large language models (specifically a fine-tuned LLaMA model, benchmarked against GPT-4), the team identified named entities, classified sentiment and filtered out statements that sounded like buy recommendations but were not. A comment such as "if I had only bought Apple stock 20 years ago, I'd be a billionaire now," carries positive sentiment and names a stock, but it is not a recommendation; traditional NLP approaches used in earlier research could have easily misclassified it.

Portfolios were then constructed on the assumption that a retail investor had acted on each recommendation the day after it was published, holding positions for one year, rebalancing to equal weight and selling on any sell signal. The results were unambiguous. The median portfolio underperformed the S&P 500 by 43 percentage points. When adjusted for risk using the Sharpe ratio, the picture worsened: most recommended assets had higher volatility than the benchmark, meaning investors were accepting more risk for lower returns.

Learn more: From deepfakes to digital trust: Combating digital misinformation

"There is a 43 percentage point underperformance of the portfolios relative to the S&P 500, and this is a pretty robust pattern across different language models, across different portfolio constructions," Prof. Papies said.

The hype cycle mechanism

The most revealing finding was not the underperformance itself but the pattern surrounding it. When Prof. Papies examined asset performance in the period before a recommendation was published, a consistent picture emerged. Recommended stocks had typically outperformed the market in the weeks and months leading up to the video. After the recommendation, they underperformed. The pattern held regardless of the pre-recommendation window: 1 week, 3 months, or 1 year.

The interpretation is consistent with a reversion-to-mean effect following a period of market hype. Financial influencers are not identifying undervalued assets or demonstrating genuine analytical edge; they are amplifying momentum that is already fading.

"We see two distributions here,” said Prof. Papies, who referred to his presentation: “The blue distribution is the returns prior to the recommendation. What we see is that prior to the recommendation, the portfolios outperform the market. Post recommendation, they underperform. So what does it mean? It's a hype cycle. The financial influencers jump onto the train of a stock that is being hyped, they make that recommendation, but subsequent to the recommendation, this stock underperforms."

"Long story short, there is no real identifiable pattern. If you run regressions, most of the stuff is really insignificant"

DOMINIK PAPIES

Crypto-focused influencers showed comparatively stronger raw performance, driven by a small number of highly successful calls. Yet the risk metrics told a different story: the maximum drawdown (the gap between the peak and trough of returns) in cryptocurrency portfolios was severe, and the value-at-risk figures were correspondingly high. Whether certain crypto recommendations reflected genuine insight or the mechanics of a pump-and-dump scheme on low-liquidity assets remains an open question, though Prof. Papies noted it as a plausible explanation warranting further investigation.

Is there a way to identify the good ones?

The natural follow-up question, Prof. Papies acknowledged, is whether some influencers are simply better than others. A heterogeneity analysis examined every available variable: subscriber count, average views, verified status, investment focus and more. With one exception, the results were uniformly insignificant. No observable characteristic predicted which influencer would deliver better outcomes for followers.

The single significant variable was cryptocurrency focus, which correlated with higher returns, largely for the reasons noted above. Beyond that, there is no research-supported basis for distinguishing reliable financial influencers from unreliable ones using publicly available signals.

Subscribe to BusinessThink for the latest research, analysis and insights from UNSW Business School

"Long story short, there is no real identifiable pattern. If you run regressions, most of the stuff is really insignificant. The only thing that is significant is the crypto focus,” said Prof. Papies, who acknowledged this finding has consequences for retail investors who believe they can identify credible voices through proxies such as production quality, subscriber volume or confident delivery. The data does not support that belief.

What this means beyond finance

Prof. Papies framed the study deliberately within marketing scholarship, not just financial economics. This is because influencer marketing operates on a core assumption: that parasocial trust translates into genuine consumer benefit. Firms invest in influencer partnerships, measure success through engagement and clicks, and rarely, if ever, assess whether the recommended product or decision actually served the consumer well.

Financial advice is unusual in that it produces a measurable outcome. A stock either outperforms the market or it does not. In most other domains (whether influencer-recommended skincare, supplements, travel experiences or software tools), no equivalent outcome metric exists. Prof. Papies used finance precisely because it offered that clarity, and the finding has implications that extend well beyond investment portfolios.

"Influencers may be beneficial for the firms whose product they recommend, but not for the consumers,” he said. “So, one could also say that digital trust can be behaviourally powerful, but it is informationally unreliable."

The broader implication is a challenge for brands. A marketing strategy built on influencer endorsement may be producing persuasion without benefit: reach without reliability. For firms operating in categories where consumer outcomes are high-stakes, the research suggests that parasocial trust can drive purchase behaviour while delivering poor outcomes to consumers. That is not a foundation for durable brand equity, and it is increasingly the concern of regulators who are drawing the same conclusions, Prof. Papies affirmed.