PortfolioGPT

Positioning

PortfolioGPT is a small, cheap, oddly honest tool from Unprompted Labs: you answer a short questionnaire about how much you want to invest, how much risk you can stomach, and when you need the money, and an LLM writes you a portfolio. It is not a robo-advisor and does not pretend to be one — no brokerage connection, no money movement, no ongoing management. What it sells is the first fifteen minutes of portfolio construction: the allocation sketch that most people never get to because they either freeze or copy a random forum post. At $3 per month or $27 per year, with a genuinely usable free tier, it is priced like a novelty and behaves more like a research worksheet. My expectation for tools in this bracket is low: either they spit out generic ETF mush, or they hallucinate tickers. PortfolioGPT does some of both, but it does enough right that I kept using it after the review window closed.

What it actually does

The input form is the product. You enter an investment amount, pick risk tolerance from three buckets (conservative, balanced, aggressive), set a time horizon (under a year, one to three years, three to five, or ten-plus), add your age and a goal, and check which asset classes you want considered — ETFs, stocks, bonds, real estate, commodities, and cryptocurrencies. On submission, the app pulls real-time market context and analyst sentiment from Yahoo Finance data, feeds your questionnaire into OpenAI's GPT-4.1-mini, and returns a portfolio with per-asset allocations, dollar amounts at your stated principal, a backtested performance estimate over a trailing window matching your horizon, and forward-looking guidance estimates. Vendor documentation states that balanced portfolios come back with an equal-weighted mix of ETFs, stocks and bonds, which tells you the allocation philosophy is deliberately simple.

The free tier is ad-sponsored and allows five portfolio generations per day, and it includes the surrounding calculators — retirement, interest and budgeting. The paid plans (at time of writing, $3/month or $27/year per the vendor's own blog and homepage) add unlimited generations, backtesting detail, forward guidance, market context, and the sibling tools: Index GPT for index portfolios (S&P 500, NASDAQ, international) and Sector GPT for sector plays such as technology, healthcare and energy. The vendor has also published Real Estate GPT, for REIT and property-adjacent portfolios, and has signposted an Income GPT.

My testing notes

Three things stood out. First, the questionnaire-to-output loop is genuinely fast and the risk input actually moves the result: I generated matched pairs of portfolios at identical amounts and horizons with conservative versus aggressive risk settings, and the aggressive outputs tilted toward individual equities and crypto while the conservative runs pulled toward bonds and broad ETFs. The reasoning text changed tone accordingly — more capital-protection language on the conservative side. That is the correct behavior and not all tools in this class manage it.

Second, and less flattering: the goal field is decorative. The vendor's own getting-started documentation notes that the goal filter 'does not have any impact on the portfolio strategy outcome,' and my test confirmed it — a portfolio generated for 'retirement' and one for 'a house down payment' with identical other inputs came back materially the same. I would rather the field were removed than left in as a placebo knob, because it invites users to believe the output is goal-aware when it is not.

Third, the output is an idea board, not a plan. The backtested performance numbers arrive with no visible methodology — no benchmark, no fee assumptions, no drawdown figures — and the forward guidance estimates are LLM prose summarizing analyst sentiment rather than quantified forecasts. When I probed for why a particular stock was chosen, the responses were the kind of general diversification language you could write yourself. There is no per-pick rationale of the depth a human advisor would give. Still, iterating is cheap and fast: changing the horizon from three years to ten visibly shifted the mix toward growth assets, and the five-free-generations-a-day allowance meant I could run a full matrix of scenarios without paying anything.

On the surrounding experience: sign-up is Google or magic-link, and the dashboard is closer to a toy than a workspace — there is no portfolio tracker, no saved-scenario comparison view, and no export I could find beyond reading the generated page. I ran Index GPT as well, and it follows the same pattern for index portfolios; Sector GPT is the same form with a sector picker. The consistent design across the suite suggests a small team templating one good interaction repeatedly, which is a reasonable way to build cheap tools but means none of them deepen into a workflow. One practical plus: because the asset-class checkboxes include real estate and commodities, I could generate an all-ETFs portfolio and a diversified one side by side and see how the model treats alternatives at small principal amounts — at $500 the allocations still came back in whole shares and sane dollar amounts, which is a small thing that many portfolio generators get wrong.

Limits

Start with the disclaimers the product itself implies: PortfolioGPT does not provide licensed financial advice, does not execute trades, does not connect to brokerages, and does not rebalance or harvest tax losses. Once a portfolio is generated you cannot edit it — no swapping a ticker you dislike — so the tool gives you a take-it-or-leave-it snapshot. The underlying model is GPT-4.1-mini with a mid-2024 knowledge cutoff, which means anything requiring post-cutoff market knowledge leans entirely on the Yahoo Finance feed layered on top; the synthesis between the two is sometimes visible as stale framing around current numbers.

Scale and maturity signals are thin. The homepage counters disagree with each other — one block claims 1,640+ portfolios generated, another claims 2,160+ investors — and both are unverified vendor claims. There is no public community, no evidence of a support organization behind the product, and third-party coverage is sparse; the most detailed writeups are affiliate-adjacent blogs rather than independent audits. Support, from what I could see, is an email address and a blog — fine for a $3 tool, but it means no one is holding your hand when a portfolio looks wrong. Nobody publishes a track record for these generated allocations either: because you cannot save or monitor a portfolio inside the product, there is no way to check how its suggestions performed after the fact. For a tool that shapes investment decisions, the absence of published allocation methodology is the biggest trust gap: I could not find a document explaining how backtests are computed or what 'forward guidance estimates' concretely mean. And the naming collision — several unrelated products use 'PortfolioGPT' or near-variants — makes due diligence harder than it should be.

Verdict

Treat PortfolioGPT as a five-minute allocation brainstorm, not advice. On my rubric it scores 3.0 overall: the core function — questionnaire to plausible portfolio — works and the risk sensitivity is real, which carries the 40% core-functionality weight further than I expected; value is excellent at $3/month; usability is a clean, short form; but support/ecosystem and reliability signals are minimal, and the shallow rationale plus decorative goal field cap the depth score. For a beginner who wants a concrete starting allocation to research further, it earns its price and then some — the free tier alone answers the question 'what does a sensible mix for my situation even look like'. For anyone making actual investment decisions, pair it with real market data and, where the sums matter, a human advisor — and re-run any allocation it suggests through your own reasoning before a dollar moves.

Sources

• https://portfoliogpt.xyz/
• https://portfoliogpt.xyz/blog/ai-investment-portfolio-generator
• https://blog.portfoliogpt.xyz/tools/ai-gpt-portfolio-generator-stocks
• https://blog.portfoliogpt.xyz/getting-started-ai-portfolio-generator
• https://portfoliogpt.xyz/blog/ai-real-estate-portfolio-generator
• https://rightaichoice.com/tools/portfoliogpt
• https://rightaichoice.com/tools/portfoliogpt/sentiment
• https://automateed.com/portfoliogpt-review
• https://gistdam.com/portfoliogpt
• https://challengingvoice.com/tools/portfoliogpt
• https://lizaai.blog/portfoliogpt-ai-review