Verified against Perplexity Pro · 2026-07-24
Force independent corroboration before you trust a claim, not citation-cluster agreement
A verification prompt that requires a stated minimum of genuinely independent sources for a claim, explicitly ruling out citations that all trace back to the same original press release, study, or wire report.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Check the following claim against independent sources. Do not treat this as answered until you find at least 3 sources that are independent of each other — meaning none of them is simply reporting on, quoting, summarizing, or citing one of the others as its own evidence. CLAIM TO CHECK A recent survey found that 40% of enterprise buyers now require an AI usage policy from vendors before signing a contract. WHERE I ENCOUNTERED IT It's being quoted in a vendor's sales deck as a justification for a new compliance feature. RULES FOR WHAT COUNTS AS INDEPENDENT - A news article that cites a press release, and the press release itself, are the same source, not two. - Two outlets that both cite the same single study, survey, or wire report are the same source counted twice, not two independent confirmations. - A company's own blog post and a "sponsored" or "in partnership with" piece written using that company's talking points count as one source, regardless of the byline. - Wikipedia, the vendor's own press release, single-author blog posts may be cited for context but does not count toward the independent-source minimum on its own. WORKING THROUGH THE CITATION CHAIN For each candidate source, briefly check what it cites as its own evidence before counting it. If a source doesn't name where its figure came from at all, flag it as unsourced rather than assuming it's independent just because it doesn't explicitly credit another outlet. OUTPUT 1. For each independent source found: what it actually says, in its own words or its own data — not a paraphrase that smooths over a difference between two sources' exact wording or numbers. 2. A one-line note on whether the independent sources agree, partially agree, or conflict. If they conflict, say specifically what the disagreement is about — methodology, timeframe, definition, sample — not just "sources disagree." 3. If you cannot find 3 truly independent sources, say so plainly and state how many you actually found, rather than padding the list with derivative coverage to hit the requested number. 4. A one-line verdict: is this claim currently well-corroborated, thinly corroborated, or effectively single-sourced dressed up as consensus?
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Search-grounded answer engines have a documented failure mode of citation clustering — returning several citations that look like independent corroboration but are actually all downstream of one wire story or one press release repeating the same original claim in slightly different words. Explicitly defining what counts as independent, rather than just asking for "multiple sources," closes the exact loophole a citation count alone leaves open: a naive count treats five outlets quoting the same PR blast as five sources instead of one, and a model asked only for "a few sources" has no reason to notice they're the same underlying claim wearing five different bylines. Instructing the model to check each candidate source's own citation before counting it turns triangulation from a surface-level tally into an actual chain-walk, which is the mechanism that catches the case where an article never explicitly says "according to X" but is nonetheless just paraphrasing X without attribution — a pattern plain citation-counting misses entirely because there's no visible citation to flag. Setting a numeric minimum forces the retrieval step to keep searching past the first agreeing cluster it finds, since stopping early would visibly fail the stated bar rather than just under-deliver quietly, and requiring the model to admit when it cannot reach that minimum prevents the alternative and more common failure mode: padding a thin result with derivative coverage just to appear to have satisfied the requirement, which is worse than an honest shortfall because it manufactures false confidence in exactly the place a reader is relying on the count to mean something.
What you get back
Source 1 (survey publisher, primary data): reports 41% of 500 surveyed enterprise buyers. Source 2 (independent industry outlet, own reporting): cites the same underlying survey but adds context from three named buyers confirming the policy requirement independently. Source 3: could not be verified as independent — it cites Source 1's press release only, with no additional reporting of its own. Verdict: only 2 independent sources found, not the 3 requested. The 40% figure traces to one survey; corroboration beyond that survey is currently thin, so treat this as thinly corroborated rather than confirmed.
Verified against
Perplexity Pro Sonar Pro · 2026-07-24
Changelog
- 2026-07-24 — Initial publish, verified against Perplexity Pro Sonar Pro search.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
