Don't Trust One AI Answer: Check the Sources and Ask a Second Model

Ask an AI model what the lot size of a Nifty contract is and you'll get a neat, confident answer. Sometimes it's right. Sometimes it's last year's number. The tone is exactly the same either way, and that's the problem.

We use AI every day, including for trading research, so we're not going to tell you to stop. We'd just suggest two small habits. They catch most of the mistakes, and PetAi happens to have a page for each.

Read the sources, not just the summary#

PetAi's Research page reads the web in a few rounds and writes you a report with numbered sources. The habit is simple: open two or three of those sources and look for the number or rule yourself. If it's not in there, treat the claim as unproven, however good the report sounds.

For Indian market rules, the source should usually be SEBI, NSE, BSE or your broker. A random blog repeating something isn't confirmation. Lot sizes, margin rules and expiry days change, and old articles stay online for years.

Ask more than one model, without the names#

The Compare page sends the same question to several models and shows their answers side by side with the names hidden. You read them and vote for the one you think is best. Hiding the names sounds like a gimmick, but it really does change how you read. You stop trusting the famous brand and start looking at what each answer actually says.

When the models agree, good. When they don't, that's your signal to go and check the facts yourself.

Our five-minute routine#

1. Ask the question in Research and read the summary.
2. Open at least two of the sources and confirm the key numbers.
3. Ask the same thing in Compare and see if the models agree.
4. Write the confirmed answer, with the link, into Notes or your trading journal.

Both pages are in the console preview on pet-ai.si. If you want to practise the habit without PetAi, our intraday rulebook is a good place to write down the rules you'll be checking answers against.

See Research and Compare on pet-ai.si →

Read: PetAi in beta, what it does today →