Verified against Perplexity Comet · 2026-08-08
Turn a pile of open tabs into a credibility scorecard instead of a gut-feel judgment call
A Comet browsing brief that scores each source you've gathered on independence, track record, and how directly it addresses the claim — instead of letting the most confidently written page win by default.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are scoring the sources currently open in my browser tabs (or the ones I list below) for how much weight each one should actually carry on the claim I'm evaluating — not for how well-written or authoritative each page sounds. CLAIM BEING EVALUATED That a particular third-party logistics vendor has a 98% on-time delivery rate for cold-chain shipments. SOURCES TO SCORE The vendor's own case study page, a trade-publication article citing the vendor, and a Reddit thread from logistics managers discussing the vendor. WHAT THIS DECISION AFFECTS Whether to sign a 2-year contract with this vendor for our perishable-goods line. SCORING RULES For each source, open it and check four things before assigning a score: (1) Independence — is this outlet/author reporting original findings, citing someone else's, or repackaging a press release with no added verification; (2) track record on this specific topic — has this source been right, wrong, or walked-back on similar claims before, based on what you can find; (3) directness — does the source actually address That a particular third-party logistics vendor has a 98% on-time delivery rate for cold-chain shipments., or does it address something adjacent that's being stretched to fit; (4) disclosed interest — does the source (or whoever it's citing) have a financial or reputational stake in the claim being true. A source that reads confidently but fails on directness or independence should score lower than a hedged source that's closer to the primary data — do not let writing quality substitute for evidentiary weight. SCORE EACH SOURCE Name | Independence (High/Med/Low) | Directness (High/Med/Low) | Disclosed Interest (None/Some/Significant) | One-line reason for the overall weight you're giving it. OVERALL VERDICT State which 1-2 sources should actually anchor a decision on Whether to sign a 2-year contract with this vendor for our perishable-goods line., and which sources — even if they showed up high in search results — should be treated as color or background only. If every source you found shares the same weakness (e.g. all trace back to one original report), say so explicitly, since that changes how much total confidence the claim deserves regardless of how many pages repeat it. If you cannot verify one of the four checks for a source (e.g. the author's track record isn't findable), say that plainly rather than assigning a score anyway — an unscoreable gap is different from a low score and the two should never look the same in the output.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Comet operates as a browser agent that actually opens and reads the tabs in front of you rather than only running a search-index query, which makes it possible to check independence and directness against the live page itself — including bylines, disclosure statements, and outbound citations a search snippet would never surface. The four-check structure exists because a language model's default failure mode on source evaluation is fluency bias: a well-organized, confidently written page scores as more credible than a messier one even when the messier one is closer to primary data, so forcing independence and directness to be scored as separate, named fields breaks the correlation between writing quality and evidentiary weight instead of letting one contaminate the other. Requiring the model to check whether every source traces back to the same original report matters specifically because repeated citation creates an illusion of corroboration — five articles citing one press release is one data point wearing five costumes, and a scorecard that scores each article independently without flagging shared ancestry would overstate confidence exactly where it shouldn't. The instruction to distinguish an unscoreable gap from an actual low score matters because a model under instruction to always produce a score will otherwise quietly convert 'I couldn't verify this' into a middling default score, which erases the single most useful signal in the whole exercise — that some claims simply cannot be checked with what's available, and a decision resting on them should carry more caveats, not fewer.
What you get back
Vendor case study page | Independence: Low | Directness: High | Disclosed Interest: Significant | Reason: directly addresses the claim but is the vendor's own marketing content. Trade-publication article | Independence: Med | Directness: Med | Disclosed Interest: None found | Reason: cites the vendor's stat without independent verification. Reddit thread | Independence: High | Directness: Low | Disclosed Interest: None | Reason: firsthand accounts but anecdotal, not the specific 98% figure. Verdict: none of the three sources independently verifies the 98% figure — treat it as vendor-claimed, not confirmed, before signing.
Verified against
Perplexity Comet 2026 browser agent · 2026-08-08
Changelog
- 2026-08-08 — Initial publish, verified against Perplexity Comet 2026 browser agent.
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
