Lesson 13 of 32 · Part One, the principles

The Search for Absolute Truth

Fact-Checking

These machines state a falsehood in exactly the same confident voice they use for a fact, because they have no way of telling the two apart. So build the checking into the asking: demand a source beside every claim, give the model explicit permission to say "I am not sure", ask which of its own statements it is least certain about - and then open the links yourself. Trust is pleasant. Verification is power.

Reach for it whenWhenever a number, a case name, a date or a quotation is about to leave your hands and appear in a memo, an article, a client's inbox or a slide behind you.

How to use The Search for Absolute Truth in Perplexity

This is the chapter where the answer engine stops being the awkward fit and becomes the right tool. Everything the chapter asks you to bolt on elsewhere - cite it, name the publisher, show me where sources disagree - is simply how this one works, so spend your effort on the next level of rigour instead. Insist on primary sources and name them: the statute on India Code, the court's own database, the disaster authority's bulletin, the review rather than the press release. Set Focus to Academic when you want research literature and Web when you want what was reported this season. Ask for the date printed on each page, for the exact line quoted rather than paraphrased, and for an explicit list of what could not be verified. Then keep the verified material in a Space so a long piece of work accumulates evidence instead of repeating searches. The follow-up is your real instrument: narrow to one disputed figure and keep narrowing until you reach the document.

  • Name the primary publisher you want: gazette, court database, authority's bulletin.
  • Academic focus for research, Web focus for this season's reporting.
  • Ask for the date printed on each page you are given.

From GenAI for Everyone: Prompt Engineering Handbook, First Edition 2025, by Raja Rajeshwari Seshadri and Shreeram Iyer.

This is the chapter where the answer engine stops being the awkward fit and becomes the right tool. Everything the chapter asks you to bolt on elsewhere - cite it, name the publisher, show me where sources disagree - is simply how this one works, so spend your effort on the next level of rigour instead. Insist on primary sources and name them: the statute on India Code, the court's own database, the disaster authority's bulletin, the review rather than the press release. Set Focus to Academic when you want research literature and Web when you want what was reported this season. Ask for the date printed on each page, for the exact line quoted rather than paraphrased, and for an explicit list of what could not be verified. Then keep the verified material in a Space so a long piece of work accumulates evidence instead of repeating searches. The follow-up is your real instrument: narrow to one disputed figure and keep narrowing until you reach the document.

Worth knowingIts one dangerous habit is the opposite of the others': the links are real, so you relax. But a citation can sit beside a sentence it does not support - the page exists, the claim is not on it - and nothing in the layout distinguishes a court record from an SEO page that scraped one. It also paraphrases by default, so a sentence in quotation marks is only a quotation if you asked for the verbatim line and then read the page. On a follow-up it can quietly drop the source that mattered. So rank the sources yourself, ask which single claim is weakest, and open the primary document. The fact that a machine finally shows you its receipts is not the same as the receipts being read.
  • Name the primary publisher you want: gazette, court database, authority's bulletin.
  • Academic focus for research, Web focus for this season's reporting.
  • Ask for the date printed on each page you are given.
  • Demand the verbatim line, not a paraphrase in quotation marks.
  • Keep verified evidence in a Space; narrow by follow-up to the document.
Ask for the Receipts

Instead of

Tell me cases on consumer rights for defective goods.

Perplexity version

I am a second-year law student in Kochi preparing a moot-court memo on defective goods under the Consumer Protection Act, 2019. Opposing counsel will check every citation, so I need sources I can open, not a summary.

Source hierarchy, and say which tier each answer came from. Tier one: the Act as published on India Code or in the gazette. Tier two: judgments on the Supreme Court, High Court or consumer commission's own database. Tier three: a law journal or a reported commentary. Nothing from a notes site, a coaching portal or a question bank - if that is all that exists for a point, say so explicitly.

1. Quote verbatim the provisions defining defect and setting out the remedies available for defective goods. Verbatim, not paraphrased, with the section number and the link.
2. The leading decisions applying those provisions: for each, the case name, the forum, the year, the paragraph number carrying the point, and the exact sentence from that paragraph. If you cannot reach the paragraph, say you cannot and give me the citation anyway so I can pull it myself.
3. Where the commissions have taken different views on manufacturer versus seller liability, set the decisions against each other and quote both.
4. A list of everything above you could not verify against tier one or tier two.

Give me the date on every page you used, and save this thread in a Space called "Moot court - defective goods".
Open Perplexity 1,415 characters

Then loop it

  1. For each case you cited, confirm it exists on the forum's own database and give me that link specifically. If a case appears only on commentary sites, tell me - I will treat it as unverified until I find the judgment.
  2. Narrow to the manufacturer-versus-seller conflict. Quote the key paragraph from each side of it and tell me which way the more recent and higher authority runs.
  3. Now list, in order, the three citations in this thread that are weakest - name the weakness, not just the case - and give me the one document I should read in full before I write.

Why it worksThe receipts arrive by default here, so the asking rises a level: a tier of sources, a verbatim line, a paragraph number - and an explicit list of what could not be verified at all.

Triangulate the Statistic

Instead of

How many people are affected by Assam floods every year?

Perplexity version

I am a journalism student in Guwahati writing on floods in Assam. I need five years of figures with sourcing strong enough for an editor, and I need to understand why published numbers differ.

Primary sources only for the figures themselves - the state disaster management authority's bulletins and annual reports, the central ministry's published data, answers given in Parliament, the central water commission. News reports are for cross-checking, not for the number.

1. A year-by-year table for the last five years: people affected, districts affected, the publisher, the link, and the date printed on that page.
2. For each year, say which document the figure comes from and whether it is a peak-day count or a cumulative figure for the season. This single distinction explains most of the gaps between reported numbers.
3. Set the official figure against the most-cited news figure for each year. Where they differ, quote both and explain the difference in terms of what was being counted - affected, displaced, in relief camps, revenue circles against districts.
4. Tell me which years are genuinely not available in a primary source. Leave the gap; do not interpolate.
5. Name the documents I should download and keep, and where they are published.

Switch to Academic focus for anything on long-term flood trends, and tell me when you have done so. Save the thread in a Space called "Assam floods".
Open Perplexity 1,408 characters

Then loop it

  1. Narrow to the most recent year. Quote the exact line from the authority's own bulletin that carries the affected figure, give me the bulletin number and its date, and link the PDF rather than a page describing it.
  2. Here is my draft paragraph. Against the primary sources in this thread only, mark each claim verified, unclear or unsupported, and give the link for the verified ones: [paste paragraph]
  3. Now tell me which of the figures in this thread is most likely to be challenged, and which single document would settle the challenge. I will read that one myself before filing.

Why it worksNumbers that disagree are the story here, so it goes past the citation to the thing the citation hides: what each source was counting, and in which document the real figure sits.

Question the Claim

Instead of

Is it true that eating after 8 pm makes you fat?

Perplexity version

I am a gym trainer in Amritsar. My clients repeat that eating after 8 pm makes you gain weight. I want the state of the evidence, not an opinion - and I will write the message to clients in another tool, because that part is not what this one is for.

Use Academic focus for the research and tell me when you switch away from it.

1. Start at the top of the evidence pyramid: systematic reviews and meta-analyses on meal timing and body weight. For each, the journal, the year, how many studies and participants it pooled, and its stated conclusion in its own words.
2. Then randomised controlled trials specifically on late eating or time-restricted eating: sample size, duration, population, what was actually measured, and who funded it.
3. Then where the field disagrees, and what the disagreement turns on - total calories against timing, shift workers against the general population, self-reported diet diaries against measured intake.
4. Then the 8 pm claim itself. Does any study anywhere test a clock time of 8 pm, or is a general finding being repeated as a rule? Cite whatever you find, including popular articles if that is where the rule actually lives, and label them as such.
5. Finally: where is the evidence genuinely weak or absent? Say so plainly. I would rather tell a client we do not know.

This is general information for clients, not medical or dietary advice, so mark anything that belongs with a doctor or dietitian.
Open Perplexity 1,442 characters

Then loop it

  1. Narrow to the two strongest trials. For each, quote the sentence from the paper reporting the primary outcome, and tell me whether the finding held after accounting for total calories.
  2. Which claim in this thread rests on the weakest source, and which rests on a popular article rather than a study? Name them, and do not soften it.
  3. Now give me a plain five-line summary of what is established, what is debated and what is unsupported, each with its citation, and save the thread in a Space called "Client questions". I will take those five lines to another tool to write the Punjabi note, with a line telling clients to see a doctor or dietitian.

Why it worksSorting a viral claim into established, debated and myth works best where every tier of evidence arrives with its source - and where the honest finding may be that nobody has studied the 8 pm rule at all.

Questions

What is The Search for Absolute Truth in prompt engineering?

These machines state a falsehood in exactly the same confident voice they use for a fact, because they have no way of telling the two apart. So build the checking into the asking: demand a source beside every claim, give the model explicit permission to say "I am not sure", ask which of its own statements it is least certain about - and then open the links yourself. Trust is pleasant. Verification is power.

When should I use this technique?

Whenever a number, a case name, a date or a quotation is about to leave your hands and appear in a memo, an article, a client's inbox or a slide behind you.

How do I apply it in Perplexity?

This is the chapter where the answer engine stops being the awkward fit and becomes the right tool. Everything the chapter asks you to bolt on elsewhere - cite it, name the publisher, show me where sources disagree - is simply how this one works, so spend your effort on the next level of rigour instead. Insist on primary sources and name them: the statute on India Code, the court's own database, the disaster authority's bulletin, the review rather than the press release. Set Focus to Academic when you want research literature and Web when you want what was reported this season. Ask for the date printed on each page, for the exact line quoted rather than paraphrased, and for an explicit list of what could not be verified. Then keep the verified material in a Space so a long piece of work accumulates evidence instead of repeating searches. The follow-up is your real instrument: narrow to one disputed figure and keep narrowing until you reach the document.

What are the limits of this technique in Perplexity?

Its one dangerous habit is the opposite of the others': the links are real, so you relax. But a citation can sit beside a sentence it does not support - the page exists, the claim is not on it - and nothing in the layout distinguishes a court record from an SEO page that scraped one. It also paraphrases by default, so a sentence in quotation marks is only a quotation if you asked for the verbatim line and then read the page. On a follow-up it can quietly drop the source that mattered. So rank the sources yourself, ask which single claim is weakest, and open the primary document. The fact that a machine finally shows you its receipts is not the same as the receipts being read.

How do I improve the answer after the first try?

Never stop at the first factual answer. Read it and underline every number, name, date and quote - these are where errors hide. Then loop: "Give the source for each of these," "Which of these are you least sure about?", "Search the web and confirm this." Gemini's double-check button and web search in all three apps help a great deal, but a source link is not proof until you open it. The final check belongs to you: read the original source yourself before you rely on it.

Loop it

Never stop at the first factual answer. Read it and underline every number, name, date and quote - these are where errors hide. Then loop: "Give the source for each of these," "Which of these are you least sure about?", "Search the web and confirm this." Gemini's double-check button and web search in all three apps help a great deal, but a source link is not proof until you open it. The final check belongs to you: read the original source yourself before you rely on it.

This lesson on the other tools

The technique is the same; the way you write it is not. ChatGPT, Claude, Gemini.