How to use AI for investing: what it can and can't do for you

What can AI assistants do well for investors?

AI assistants are good at language work: explaining terms in plain words, summarising long documents you give them, listing questions to ask, comparing how two kinds of product work, arguing the case against an idea and organising your notes. They're fast and patient, which makes them useful for learning concepts, as long as you check the facts.

What are AI assistants bad at in investing?

AI assistants are unreliable on current prices and data, on predicting markets, on your personal situation and on exact numbers. They can state a wrong figure, a made-up rule or an invented source with complete confidence, an error often called a hallucination. Treat every number and fact from an assistant as a lead to check, not an answer.

The detail behind the prediction problem is in the next lesson: can AI predict stock prices?

Which investing tasks can you safely give to AI?

Sort tasks into three groups. Use AI freely to explain, summarise and draft questions. Use it, then check every detail, for sums, comparisons and anything about current prices or rules. Never use it to judge whether a firm is genuine, to pick what to buy, to predict prices or to touch your accounts.

What is a safe way to use AI for investment research?

Use five steps every time: ask the assistant to explain, verify what it says in the original document, ask it for the counter-case, make the decision yourself, and write the decision down. The assistant speeds up understanding; the original source, your own judgement and your written record are what keep the decision safe.

1. Explain

   Ask the assistant to explain the product, term or document section in plain words, and to say which parts of its answer it's unsure about.

2. Verify in the original

   Find each key fact, figure and charge in the original source: the key information document, the terms, the annual report or the official register. If you can't find it, treat it as unconfirmed.

3. Ask for the counter-case

   Ask for the strongest reasons not to go ahead, the main risks, and what would have to happen for this to go badly.

4. You decide

   Apply your own checklist: your goal, your timeline, the fall you could live with and who holds the money. The assistant doesn't know these, and it doesn't carry the consequences.

5. Write it down

   Record what you decided, why, what you checked and what you couldn't confirm, in a decision journal.

Step 2 is the easiest to skip, and the one that matters most. An assistant's summary of a document is only as good as your check of it.

What does the workflow look like on a real question?

Here is an illustration. A learner wants to understand the costs of a fund they're considering. They paste the costs section of its key information document into an assistant, ask for a plain explanation, check every figure against the document, ask what could make costs higher, then record the total. The assistant never says whether to buy.

1. Explain. They ask: “Explain this costs section in plain English. List every charge, what it pays for and when it's taken. Don't tell me whether to invest.”
2. Verify. The assistant lists four charges. The learner finds three in the document; the fourth, a performance fee, isn't there. They mark it as an error.
3. Counter-case. They ask: “What could make my real costs higher than these figures, and what should I ask the firm?” The answer prompts them to look up the provider's dealing charges.
4. Decide. They add the provider's charges, work out the all-in cost in money using the method in how investment fees work, and judge whether the cost is acceptable for their goal.
5. Write it down. Their journal records the total, the source of each figure and the assistant's error.

Illustration only: a fictional learner and fund.

What should you never share with an AI assistant?

Never paste passwords, PINs, one-time codes, full account or card numbers, identity documents or anything that would let someone sign in as you. Be careful with names, addresses and exact balances too. Conversations may be stored or reviewed, depending on the tool and its settings, so share only what the question needs.

• Never: passwords, PINs, security answers, one-time codes, full card or account numbers, scans of passports or ID cards.
• Remove first: names, addresses, dates of birth and account numbers from any statement or document before you paste it in.
• Round it: use 'about $20,000' instead of an exact balance when the exact number doesn't matter.
• Check the settings: read how the tool stores conversations, and whether you can turn off chat history or the use of your chats for training.
• Don't connect it: never link an assistant or add-on to your bank or investment account, or give it permission to trade.

No genuine AI tool needs your banking password to explain an investment. Anything that asks for it, or for money to 'activate' trading, is a scam; AI trading bot scams shows how they work.

Why shouldn't you ask AI what to buy?

Because 'what should I buy?' asks for the thing AI assistants do worst: a confident pick based on possibly stale data, with no knowledge of your situation and no responsibility for the outcome. Ask what to understand, what to check and what to ask instead. That uses the assistant's strengths and keeps the decision with you.

A useful habit is to add one guard line to every investing prompt: “Don't tell me what to buy or sell; explain, and list what I should verify in the original document.” Our library of AI prompts for investment research is built around guard lines like this one.

How do scammers use AI against investors?

Scammers use AI to fake celebrity videos and voices, write convincing messages without mistakes, build polished websites quickly and sell fake 'AI trading bots' that promise steady daily gains. The word 'AI' is itself bait, because it suggests a clever system that ordinary investors can't inspect. The ordinary checks on the firm and the payment still work.

Whatever the technology, the money has to go somewhere. Check the firm on the official register, check where the payment would go, and treat any promise of fixed daily or weekly returns as a red flag. The patterns are set out in AI trading bot scams and investment fraud red flags.

Using AI for investing: common questions