Give it the source material
An answer grounded in your uploaded document is far more reliable than one from memory. Put the files in a Project.
Can I trust AI answers?
Can I trust AI answers? Trust the ones you can check. Treat the rest as drafts.
The question is not whether AI is trustworthy in general; it is which answer, for which purpose, with what at stake. A draft email needs no checking. A number in a business plan needs a source. Code that touches money needs a review and a test. Medical, legal or financial advice needs a professional, with the AI making that conversation shorter. This page is a practical trust scale: how much to check, based on what happens if the answer is wrong.
Free to start. The latest models and Auto Verification are on paid plans.
The trust scale
Match the checking to the cost of a mistake
| If the answer is wrong… | Example | Check with |
|---|---|---|
| Nothing happens | A draft email, a brainstorm, an explanation for yourself | Nothing. Use it. |
| You waste an hour | A recipe, a how-to, a first draft of a document | Skim for the obvious; fix as you go. |
| You waste a week or embarrass yourself | A number in a report, a claim in a post, a plan you will follow | A source you open, or a second model that agrees for its own reasons. |
| You lose money or customers | Code touching payments or data, a price, a contract clause | A second model's review, a test, and your own reading of the source. |
| You risk health, legal standing or savings | Medication, a legal position, an investment | A professional. Use the AI to prepare the questions, not to answer them. |
What makes an answer more trustworthy
Four things you can control
Ask for them, then open them
A search-native model with citations for anything factual. A citation you have not opened is decoration.
A second model, cold
Different family, fresh chat, same Project. Where they agree for their own reasons, trust rises. Where they differ, look.
Ask it to argue against
Models agree by default. "What is wrong with this?" gets you more truth than "is this right?".
When to stop and ask a person
AI narrows the question; a professional answers it
For health, law, tax and investments, use the AI to understand the vocabulary, list the options, prepare the questions and summarise the documents, then take that to someone accountable. The preparation makes the appointment shorter and cheaper, which is a real benefit. Treating the AI as the final answer in these areas is where people get hurt, and no amount of cross-checking between models fixes it, because the models share the same limitation: none of them is responsible for you.
The fix
How to check an answer in MultipleChat
The same method works for facts, code, advice and writing. It costs one or two messages in MultipleChat because every model reads the same conversation, files and Project.
Ask a second model the same question
Switch the model from the selector and ask again; the history carries over. Pick a model from a different family: ChatGPT then Claude, or Claude then Gemini. Same-family models share blind spots.
Ask it to attack, not to agree
"Argue against this answer as strongly as you can" produces more than "do you agree?". Models are agreeable by default; you have to invite disagreement.
Read the disagreement as a map
Where the two agree, confidence is warranted. Where they differ, that is the claim, the assumption or the line of code to check. Usually it is the one that mattered.
Get a source or a test for anything you will act on
Facts get a source you open (Perplexity cites; Auto Verification on paid plans runs an independent pass). Code gets a test you run. Advice gets a third model or a person.
Keep it in a Project
The question, both answers and the sources stay together, so the next time the topic comes up the checking is already done.
Free is the start. The latest models are the difference.
Checking answers is where the latest models earn their price.
The free plan gives you the major model families and enough messages to make second opinions a habit, which is the whole point of this page. The paid plans add the latest model in each family, which disagree in more useful ways and invent less; Perplexity Sonar Pro for sourced facts; and Auto Verification, an independent verification pass that runs when factual confidence matters, without you orchestrating it.
Newer models hold your whole Project in view instead of the last few messages, reason through trade-offs instead of picking the first plausible answer, and are wrong less often and with less confidence. For the questions on this page, that is the difference between advice that sounds right and advice you can act on.
Pro from $20 a month, cancel any time. The free plan stays free.
- Your first Project and file uploads
- The major AI families to try the workflow
- A small daily message allowance
- The latest ChatGPT, Claude, Gemini, Grok and Perplexity models, all on the same Project
- Far more messages a day, so a working session does not stop halfway
- Modes where several models draft, challenge and verify each other's answers
- Auto Verification: an independent check on factual claims
- Perplexity Sonar Pro for sourced, current facts
- Collaboration modes where models draft, challenge and verify each other
Start here
Messages that raise trust
Use before you act on an answer.
Ground itAnswer using only the documents in this Project. Quote the passage each claim rests on.
Source itGive me a source I can open for each factual claim.
Cold checkAnswer independently: [question]. Then I will compare with another model.
AttackWhat is wrong with this answer? Argue against it.
StakesIf this answer were wrong, what would it cost me? What should I verify before acting?
PrepareI am seeing a [lawyer/doctor/accountant] about this. What questions should I ask, and what documents should I bring?
Honest answers
What people ask
Is AI trustworthy for research?
For finding and summarising, yes, with sources you open. For a number you will repeat, only after you have seen the source. See research with AI.
Is AI trustworthy for code?
For writing a first version, yes. For shipping it, after a second model's review and a test of the case nobody mentioned. See coding with AI.
Is AI trustworthy for medical or legal questions?
For understanding and preparing, yes. For deciding, no. Take the preparation to a professional.
Does a second model make an answer trustworthy?
It makes it more trustworthy, and it shows you where the uncertainty is. It does not replace a source or a test for high-stakes answers.
Why is MultipleChat better for this than one chatbot?
Because the checks are cheap: the same context is read by several models, Perplexity supplies sources, and Auto Verification on paid plans runs an independent pass. Checking becomes a habit instead of a chore.
What is free?
The major model families and a daily message allowance. The latest models, Sonar Pro and Auto Verification are on paid plans.
Decide what a wrong answer would cost. Check accordingly.
Context, sources, a second model, and a person when it matters most. Free to start.
Also: how to know if AI is wrong · can AI lie to me? · should I ask two AIs?