SEC rules require us to show performance inside a logged-in account.
Every strategy runs live. Inside, you see every holding, every trade, and every move to cash.
An AI agent built on a chatbot writes each answer fresh, so the same question can produce two different portfolios. That is fine for brainstorming and risky for money. Here is why it happens and how fixed rules fix it.

An AI trading agent built on a chatbot writes each answer fresh, so the same question can produce two different portfolios. That is fine for brainstorming and risky for money. Fixed rules fix it.
Your AI gives different answers because it writes each reply from scratch, choosing words by probability. Small differences early in a reply can lead to a very different answer by the end.
For an essay, that variety is useful. For an AI trading agent, it means two identical requests can lead to two different sets of trades.
Chatbots vary their answers because they generate text one piece at a time, picking each next word from a range of likely options. Most apps add some randomness to that choice on purpose, so replies sound natural.
Even with randomness turned down, hosted AI services are not fully consistent. A 2025 study set 5 AI models to their most consistent settings and ran 8 tasks 10 times each. Accuracy still varied by up to 15% between runs.
The answers can change a lot. Ouyang and colleagues asked ChatGPT to write code for the same 829 problems many times. For 47% to 76% of tasks, depending on the test set, no two requests produced code with the same test results.
They also found that setting the model's randomness to zero reduced the problem but did not remove it. Trading questions are not code, but the same mechanism is at work.
In trading, a different answer becomes a different order. Ask an agent "what should my portfolio hold?" on Monday and again on Tuesday. You may get two different lists, even if nothing in the market changed.
FINRA is the Financial Industry Regulatory Authority. It lists hard-to-trace reasoning and hallucinations, or confident false statements, as risks of AI agents.
Rules fix the problem by moving the decision out of the chatbot. A fixed rule gives the same output for the same input every time. The agent's job shrinks to carrying out what the rule says.
Portfolio Lab sums it up: language models read, explain, and execute. They don't decide.
1. Give the agent a written plan with target holdings and sizes.
2. Ask it to compare your account to the plan, not to suggest ideas.
3. Use the same saved prompt every time.
4. Run it at the same time each day.
5. Ask it to list orders before placing them, and compare runs.
6. Tell it to stop and ask you if anything in the plan is unclear.
Timing helps too. See when to run your Portfolio Lab prompt. For the research on AI picking its own trades, read do AI trading bots actually work?
Portfolio Lab is a strategy research and simulation tool from alphaAI Capital. It lets you build investment strategies and test them in live markets with virtual money. You can connect it to AI apps like Claude or ChatGPT, so your agent has a strategy's plan to follow.
It writes each reply word by word, picking from likely options with some randomness. Small early differences grow into different answers.
You can reduce variation, but research shows hosted AI models still vary even on their most consistent settings.
An AI trading agent is an AI app, such as Claude or ChatGPT, connected to a broker so it can place orders for you.
For picking investments, yes. It means your strategy changes each time. For carrying out a fixed plan, it matters much less.
Yes. All chatbots built on large language models generate text this way, so all of them can vary.
Give it a fixed written plan and a saved prompt, and ask it to carry out the plan rather than make choices.
A hallucination is when an AI states something false with confidence. FINRA lists it as a risk of AI agents.
Build, test, and deploy your own strategies in Portfolio Lab. Free to start.
Educational & Research Disclosure: The content provided is for informational and educational purposes only and is not intended to constitute investment advice, a recommendation, solicitation, or offer to buy or sell any security. Any discussion of market trends, historical performance, academic research, models, examples, or illustrations is presented solely to explain general financial concepts and does not represent a prediction, guarantee, or assurance of future results. Past performance is not indicative of future results. All investing involves risk, including the possible loss of principal.