Who Let the Bots Out?: 2026 AI Portfolio Performance Showdown

As artificial intelligence transitions from a conversational tool to a strategic partner, its ability to navigate complex financial markets has become a key area of interest for investors and technologists alike.

Can AI effectively manage risk and deliver returns in real-world markets? To answer this question, we have deployed seven of the most advanced AI platforms such as ChatGPT, Gemini, Copilot, DeepSeek, Perplexity, Meta AI and Grok to act as portfolio managers.

What is the 2026 AI Portfolio Performance Showdown?

The 2026 AI Portfolio Performance Showdown is a year-long experiment running from January 2026 to December 2026. It is designed to test how different artificial intelligence models make real-world investment decisions over a full market cycle.

The project aims to assess how AI interprets market conditions, balances risk and return and adapts its strategy as economic and market data evolves throughout the year. By tracking decisions over 12 months rather than relying on short-term predictions or hypothetical scenarios, the project provides a more realistic view of AI-driven investment behaviour.

All portfolios are constructed and managed under consistent rules, allowing for direct comparison between models and risk profiles within an Australian market context. The results displayed are updated monthly, with rankings based on total return adjusted for the specific risk parameters of each portfolio profile.

How are the portfolios built and tested over time?

At the start of the project in January 2026, each AI model is asked to construct four separate 10-stock portfolios and assign a percentage weighting to each stock. Each portfolio reflects a different risk appetite commonly seen among Australian investors and is designed to operate within the realities of the local share market.

The four risk profiles are:

• Conservative, focused on capital preservation and reliable income

• Moderate, balancing long-term growth with steadier returns

• Aggressive, targeting higher growth through more volatile sectors and smaller companies

• Neutral, designed to broadly track overall market performance

Once the initial portfolios are established, the project moves into a monthly review cycle that runs through to December 2026. At each review point, the AI models must decide whether to keep existing stocks unchanged, adjust position sizes to increase or reduce exposure, or replace stocks if they no longer align with the portfolio's objectives. All decisions must remain consistent with the portfolio's original risk profile.

This combined process tests both the quality of each AI model's initial stock selection and its ability to adapt strategy over the course of the year.

How are the results measured and reported?

Portfolio performance is tracked from January 2026 through December 2026 using total return, which includes both price movements and income such as dividends. Results are updated monthly to reflect portfolio changes and market performance over the same period.

Portfolios are ranked within their respective risk categories rather than against all portfolios combined.

At the conclusion of the project in December 2026, the full year of results will provide a complete view of how each AI model performed across different market conditions.

Monthly Leaderboard: March

Which AI model is navigating the market most effectively? This leaderboard ranks AI models based on the performance of their portfolios over the month. Results are calculated using the percentage return generated by each portfolio, providing a clear comparison of how each model's investment strategy has performed.

Conservative

All AI portfolios recorded negative returns in March 2026, showing a broad decline compared to February.

DeepSeek was the weakest performer at about -24.0%, followed by Copilot and Meta AI with losses of around -19.3% and -16.8%. ChatGPT also declined by about -10.8%, while Gemini and Perplexity fell by roughly -4.9%. Grok had the smallest drop at around -1.9%.

The benchmark portfolios also declined but performed relatively better, with the S&P 500 Low Volatility Index down about -5.7% and the iShares Core US Aggregate Bond ETF falling by around -1.4%.

Moderate

All AI portfolios posted negative returns in March 2026 under the moderate risk profile.
DeepSeek was the weakest performer at about -27.1%, followed by Copilot at around -12.3%. Gemini and ChatGPT declined by roughly -8.1% and -7.8%, while Meta AI and Grok fell by about -5.8% and -5.0%. Perplexity had a smaller loss of around -4.2%.

The benchmark portfolios also declined but performed better overall. The S&P 500 fell by about -5.1%, while the S&P/ASX 200 recorded a slight gain of around 1.0%, making it the strongest performer for March.

Aggressive

Performance was mixed across the AI portfolios in March 2026 under the aggressive risk profile.

Gemini was the only positive performer, gaining about 2.1%, while ChatGPT had a small decline of around -0.6%. Meta AI and Copilot posted moderate losses of about -3.8% and -10.4%, with Grok close behind at around -9.9%.

DeepSeek declined further by about -16.9%, while Perplexity recorded the largest drop at roughly -25.8%.

The benchmark portfolios also fell, with the Russell 2000 down around -3.0% and the Vanguard MSCI International Shares ETF declining by about -4.0%.

Neutral

All AI portfolios recorded negative returns in March 2026 under the neutral risk profile.

Copilot and Perplexity saw the largest declines at around -14.7% and -13.3%, followed by DeepSeek at about -9.6%. Meta AI and Grok posted losses of roughly -9.0% and -8.9%, while ChatGPT and Gemini declined by around -7.8% and -7.3%.

The benchmark portfolios also fell, with the S&P 500 down about -6.7% and the S&P/ASX 200 declining by around -5.1%, performing slightly better than most AI portfolios.

What stocks did each AI model choose?

The AI models were asked to build a 10-stock portfolio in January 2026. The portfolios are reviewed monthly, giving each model the opportunity to add or remove stocks and adjust portfolio weightings based on its assessment of current market conditions.

• ChatGPT

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

• Gemini Portfolio

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

• Copilot Portfolio

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

• DeepSeek Portfolio

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

• Perplexity Portfolio

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

• MetaAI Portfolio

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

• Grok Portfolio

  • Conservative
  • Moderate
  • Aggressive
  • Neutral

Written by

William Capada

Insights Analyst

William Capada is an insights analyst at Finder. With years of experience as an analyst, he has honed his skills in analysing complex datasets and extracting actionable insights. Proficient in various analytical tools, he has a proven track record of delivering meaningful insights that drive strategic decision-making. William conducts research related to economic data and is also responsible for updating the insights statistics pages. He also assists in ensuring that the scoring makes sense for the Finder Retail Awards. See full bio

William's expertise

William has written 4 Finder guides across topics including:

• Data Analysis
• Data Visualization
• Retail Awards