Studying with AI
Use an AI assistant like a study partner, never like an oracle.
AI tools can explain a term five different ways at midnight. They can also be confidently wrong. This page covers useful prompts, a checking routine, the errors to expect and what never to paste in.
What can an AI assistant actually help me with?
It is good at explaining, rephrasing and quizzing. It is poor at facts that change, UK-specific rules and anything about your own situation.
- Explaining a term in simpler words, or with an analogy
- Breaking a long document paragraph into plain steps
- Generating practice questions on a topic you studied
- Listing questions to ask before reading a fund document
- Current rates, allowances, limits or tax rules
- Choosing investments, providers or timing
- Guessing where markets or prices will go
- Anything involving your personal or account data
AI language tools produce text by continuing patterns learned from very large amounts of writing. That makes them fluent, but fluency is not the same as accuracy, and they rarely show which parts of an answer they are unsure about.
Verify it: the provider's own help pages usually state the date range of the model's training information and its known limits.
How do I ask for a good explanation?
Four prompt patterns. Each shows a vague request, a sharper version and why the second works better.
Define, with limits
"What is a bond?"
"Explain what a bond is to a beginner in the UK, in under 150 words. Then list two risks of holding bonds, and tell me which official UK source I could use to check your explanation."
A length limit keeps it focused, "in the UK" reduces US-centred answers, and asking for a source to check builds verification into the reply.
Ask for an example
"Explain liquidity."
"Explain liquidity using one everyday example and one investment example. Use simple round numbers and say clearly that the numbers are invented."
Examples make abstract ideas concrete. Asking for invented, labelled numbers stops the reply from presenting made-up figures as real data.
Ask it to test you
"Teach me about diversification."
"I have just read about diversification. Ask me three questions, one at a time, to check whether I understand it. Wait for my answer before giving the next question, and correct me gently."
Answering questions makes you retrieve what you learned, which helps it stick far better than rereading.
Ask for the other side
"Why are low-cost funds better?"
"What are the arguments for and against focusing on fund costs? Give the strongest point on each side, without recommending anything."
A leading question invites agreement. A neutral framing gives you a fuller picture and reduces confirmation bias.
Verify it: whatever the reply says, check the key facts with the routine in the next question.
How do I verify what the AI told me?
A five-step routine that takes a few minutes and catches most problems.
-
Split the answer into claims
Underline every statement of fact: names of bodies, rules, numbers, dates. Explanations of ideas need less checking than facts.
-
Match each claim to a source type
Use the table below to decide where each kind of claim should be confirmed.
-
Go to the source directly
Type the official site's address yourself rather than following links offered in the reply, which can be wrong or invented.
-
Check the date
Rates, allowances and rules change. Make sure the source page is current and the AI's figure is not from an earlier year.
-
Note what did not match
If something was wrong, keep a short note. Patterns in the errors tell you which topics to double-check next time.
| If the claim is about… | Check it with |
|---|---|
| Interest rates, Bank Rate | Bank of England |
| Inflation, prices, earnings data | Office for National Statistics |
| Whether a firm is authorised | FCA Financial Services Register |
| Consumer rights, investment basics | FCA consumer pages, MoneyHelper |
| Tax, ISAs, pensions rules | GOV.UK and HMRC guidance |
| Compensation limits | Financial Services Compensation Scheme |
| A specific fund's costs or risks | Its Key Information Document and factsheet |
What mistakes do AI tools typically make on finance?
Six patterns appear again and again in the samples our editors review.
-
Out-of-date figures
Allowances, thresholds and rates quoted from an earlier year, stated as if current.
-
Wrong country
US account types, tax rules or regulators applied to a UK question, often without saying so.
-
Blurred institutions
Mixing up who measures, who targets and who sets: the ONS, the Government and the Bank of England have different roles.
-
Invented sources
Plausible report titles, page names or web addresses that do not exist.
-
Confident arithmetic slips
Percentages and compounding worked out with a small error early on that changes the final figure.
-
Advice by the back door
A neutral question answered with a suggested product or "a good option for most people", which no tool can know.
Verify it: try the "Spot the AI mistake" samples on our home page to practise noticing these patterns.
What should I never paste into an AI tool?
Anything that identifies you or gives access to your money. Treat every chat as if a stranger might one day read it.
- Account numbers, sort codes and card numbers
- Passwords, PINs, security answers and one-time codes
- National Insurance numbers, passport or driving licence details
- Full statements, payslips or tax documents
- Your name and address alongside your financial details
What you type may be stored by the provider and, depending on settings, used to improve their systems. If you want help with a confusing line in a statement, copy only that line, remove anything personal, and ask what the term means in general.
Verify it: read the provider's privacy notice, and the Information Commissioner's Office guidance on your data rights.
So where are the limits, in the end?
An AI assistant does not know you, cannot see the future and is not accountable for what it says. Those three limits do not go away with better prompts.
It does not know your income, debts, plans or how you would feel during a long market fall, so it cannot judge what is suitable for you. Nobody can tell what markets will do next, and a tool that sounds sure about it is guessing fluently. And unlike an authorised adviser, a chat tool carries no professional duty to you and offers no route to complain to the Financial Ombudsman Service about its answer.
Inside those limits, AI can be a patient and tireless tutor. Ask it to explain, rephrase and test you; then let official sources and, where you need it, a regulated professional handle the facts and the decisions.
Verify it: MoneyHelper explains when free guidance is enough and when regulated advice may be worth considering.