How to check a fact, a company or a figure in five minutes

How to check a fact, a company or a figure in five minutes

The answer arrives confident and with references, and then it turns out one leads elsewhere and a third does not exist. There will be no signal saying "I am unsure here". Below is why a search tool is built differently and four techniques that make it usable for work.

The answer arrives confident and with references, and then it turns out one leads elsewhere and a third does not exist. There will be no signal saying "I am unsure here". Below is why a search tool is built differently and four techniques that make it usable for work.

The problem is not that AI does not know. It is that it does not doubt

You ask and get a confident answer with references. It looks convincing. Then it turns out one reference leads elsewhere, a second to a page saying something different, and a third does not exist.

An error looks exactly like a correct answer. There will be no signal saying "I am unsure here": a number looks like a number, a date like a date, a reference like a reference.

For an everyday question that does not matter. For somebody whose mistaken fact costs a reputation — a lawyer, a doctor, a finance professional, a writer, an entrepreneur — it makes the tool unusable. Right up until a verification process appears.

The process exists, and it takes minutes rather than hours.

Why search AI works differently from a chat

The key distinction, from which everything else follows.

An ordinary model answers from memory. It does not search — it reconstructs from what it saw in training. Hence the invented references: there is nowhere to take them from, and the answer has to be complete.

A search tool searches first and answers second. It has sources and relies on them.

Hence a non-obvious consequence: long detailed requests work worse here than short ones. In an ordinary model context helps. In a search tool it gets in the way: a long request throws the search off and you get an answer based on the wrong selection.

The rule: separate the search intent from the output intent. First a short precise query to find the right thing. Then separately, what to do with what was found.

Four techniques

1. Precise search instead of general

The general web gives what is best optimised rather than what is most correct. A substantial part of the skill is being able to narrow the field: search specific sources, limit by time, go to databases rather than to articles about databases.

One properly narrowed query saves an hour of reading paraphrases.

2. The zero-trust protocol

The main technique and the main principle. It is stated like this: make the tool prove every statement.

A direct quotation rather than a paraphrase. An indication of exactly where from. And separately, an analysis of contradictions: where the sources diverge.

The last is the most valuable. Sources agreeing proves nothing — they may have copied each other. A divergence always shows where to dig.

3. Separate facts from conclusions

Ask for the result in two parts: first a list of statements with their source, then the interpretation separately.

That way you get a ready list of what needs checking instead of continuous text in which everything needs checking.

4. Multi-stage research

A complex question is not settled by one query. It is settled by a chain: found the basics → saw what was missing → refined it → found a contradiction → worked it out.

In that scheme you are the manager and the tool is the research team. Your work: setting the direction and deciding what to do with the findings.

What checking looks like in practice

Three minutes per fact, no more.

Step one — isolate the statement. Not "check the text" but take one specific thing: a figure, a date, a number, a name. Checking wholesale is impossible; checking is done point by point.

Step two — find the primary source rather than a paraphrase. An official document, a register, the organisation's site, the original publication. An article citing an article citing an article is not verification.

Step three — compare it word for word. Not "roughly this" but is it exactly that figure, that period, that definition. More than half the errors here are not invented facts but correct facts about the wrong thing: the figure exists but relates to another year or another country.

Step four — record it. The reference to the primary source goes beside the fact, in your own document. Otherwise in a month it will have to be checked again.

And the main rule of discipline: the checking is done by whoever puts their name to the result. If material goes out under your name, responsibility for every figure is yours, regardless of who found it.

What always gets checked

A short list worth keeping in mind.

Figures and dates. Any of them. Especially the plausible-looking ones.

Numbers of provisions, documents, standards. The most common category of invention.

References. Open them. Not "there is a reference" but go there and see what is claimed.

Quotations. Check the wording and the context.

Names and proper nouns. Does it exist, is it that person, is it that company.

What need not be checked: the structure, the logic of the reasoning, the phrasing. You judge those with your own expertise.

Three signs an answer cannot be trusted

Not a guarantee, but a reason to be wary before you start checking.

Figures that are too round. Real data is rarely even. "Roughly thirty percent" may be true, but more often it is a sign the figure was assembled rather than found.

A reference with no specifics. "According to studies" and "according to experts" is an absence of a source. A real reference leads to a specific document with a specific page.

An answer that is too convenient. If what was found perfectly confirms what you wanted to hear, check twice as hard. The tool adjusts to the phrasing of the question, and a leading question gives a leading answer.

Hence a technique: ask the same question in the reverse phrasing. If the answer flips to the opposite, what you have is not a fact but a reflection of your own request.

Where this applies in practice

Checking a client before a deal. How long they have existed, whether there are traces of other projects, whether what is written in the listing matches what is visible from outside. A minute of checking against an hour on a free "test task".

Analysing somebody's business before a conversation. Three precise observations gathered before the first message work better than any CV.

Checking data in your own material. Before putting a figure in a presentation for a client.

Competitive intelligence. What they are doing, what they are silent about, where anything is free.

Where to start

The “Perplexity as an investigative tool” express course — five modules on turning AI search into an analyst. Inside: the "search first" approach and why short queries work better than long ones, precise search techniques and working with databases instead of the general web, structured synthesis, zero-trust verification protocols and multi-stage research.

Who for: those who need accuracy — experts, writers, specialists and entrepreneurs for whom a mistaken fact costs a reputation and manual rechecking costs time.

The result: you stop wondering whether a reference was invented and get material you can put into work with no redoing.

The cost is $9. The course is included in the Basic plan, and registration is free and opens three days of full Basic access — that is, it can simply be worked through with nothing paid.

Inside with it are five more foundational courses, the ten-lesson sales module and chats with 200+ AI assistants across 14 categories, including analytics and legal.

Do one thing today: take the last answer you got from a model and used somewhere. Find every verifiable statement in it — figures, dates, references, numbers. Check them against sources. If everything matched, you have acquired a habit. If not, you have just understood what a protocol is for.

The next level: verification inside the process

The four techniques above are manual work. There is an approach with rechecking built in.

An investigative agent from the template catalogue — the “Autonomous deep research and analytics agent” — works differently from ordinary search: from your request it generates its own search queries, collects data from several sources (web search, Jina AI, Wikipedia) and builds a full analytical report.

Plus the lesson “Prompt engineering. Logic and reasoning” in the “Advanced prompting” course — how to make a model show its working so an error is visible in the process rather than only in the result.

The course is Basic; the template is Creator, Make, N8N and Full.

The course that makes a profession of this

“Perplexity as an investigative tool” — $9 or the Basic plan. Five modules closing the main pain of working with AI: distrust of the result.

The "search first" approach — why long detailed prompts work worse than short ones: they mix search intent with output intent and both come out worse. They have to be separated.

Precise search techniques — operators, source filters, working with databases instead of the general web.

Structured synthesis — how to make a model work as an analyst rather than a paraphraser.

The phrasing from the programme: the course turns a student into a senior research analyst.

Assistants for verification, Basic plan, the catalogue: Analytical journalist — subject analysis and working with sources. Academic editor — where precision of wording matters. Brainstorm AI: Devil's Advocacy — it looks for the weak points in a conclusion you liked. Paralegal — if what needs checking is a counterparty rather than a fact.

Skills for verification

What to pick up on the platform for this task

What to read next

How to verify a model's output — the protocol in general form.

AI for lawyers — a profession where this is critical.

How to find trends — another use of precise search.

Analysing a competitor by figures — verifiable data instead of impressions.

Best open LLMs 2026 — the full guide to open models.

Also on AI SKILLS