Perplexity

Verified against Perplexity Comet · 2026-08-14

Run a pre-publish verification pass so a factual mistake doesn't ship because nobody re-checked the draft

A final-pass checklist prompt for Comet that re-verifies every checkable claim in a near-final draft against live sources, rather than trusting whatever was verified when the draft was first written.

Perplexity Comet3 fillable variables

The prompt

Ready to copy — highlighted parts are example details you can swap.

Before this piece goes out, run a verification checklist against the draft below. Treat every version-1 fact-check as potentially stale — re-verify against current sources now, not against whatever was true when the draft was first written.

DRAFT TO CHECK
a 900-word blog post citing three industry statistics and two named executives' public statements

WHERE THIS IS PUBLISHING
the company's public blog, which gets cited by trade press

DEADLINE PRESSURE
publishing in 40 minutes, this is the last check before it goes live

CHECKLIST — GO CLAIM BY CLAIM
For every factual claim in the draft (a number, a name, a date, a quote, a status like "currently" or "as of now"), do the following: (1) check whether it's still accurate as of today, not just whether it was accurate when likely written; (2) check whether it's attributed to the right source in the draft, or whether the draft's citation doesn't actually match what that source says; (3) flag any claim stated more strongly in the draft than the underlying source actually supports (a source that says "likely" being reported in the draft as certain); (4) flag any claim that has no clear source at all, even if it reads as common knowledge.

OUTPUT
A numbered list matching the order claims appear in the draft: Claim | Status (Still accurate / Now outdated / Misattributed / Overstated / Unsourced) | What to fix, in one sentence. Only include claims that need a flag — don't pad the list with claims that checked out cleanly, since publishing in 40 minutes, this is the last check before it goes live means the list needs to be scannable, not exhaustive busywork.

End with a single line: publish as-is, publish after fixing the flagged items, or hold — do not soften this into "looks mostly good" if there's a flagged item serious enough to actually delay publication.

Customize

Optional — swap in your own details for the highlighted parts above.

Why this works

The instruction to treat every earlier fact-check as potentially stale exists because the single most common way a true claim becomes false in a published piece isn't sloppy original research — it's simply time passing between when a fact was checked and when the piece actually ships, and a verification pass that only re-reads the draft without re-checking against current sources will happily confirm a number that was accurate weeks ago and isn't anymore. Splitting the check into four distinct failure modes — outdated, misattributed, overstated, unsourced — matters because these are mechanistically different problems with different fixes, and collapsing them into one generic 'fact-check this' instruction tends to catch only the most obvious kind (outright wrong) while missing the subtler and arguably more common ones, like a source that hedges with 'likely' being flattened into a confident claim somewhere in the editing process. The overstatement check specifically matters for anything assembled from multiple drafts or contributors, since certainty tends to accumulate through revision even when the underlying evidence hasn't gotten any stronger — each pass smooths out a hedge that felt unnecessary in isolation, and only a check that compares the final wording back against the actual source will catch the cumulative drift. Limiting the output to only flagged claims, rather than a full annotated pass over every sentence, is a deliberate proportionality choice for a time-pressured final check — a long list of items that all checked out fine buries the one item that actually needs fixing, which is the opposite of what a pre-publish gate is for.

What you get back

3. Claim: "the executive said adoption would triple by year-end" | Status: Overstated | Fix: the source quote actually says adoption "could plausibly" triple, not that it will — soften the draft's wording to match. 5. Claim: "the platform serves 2 million users" | Status: Now outdated | Fix: the company's own site now states 2.4 million as of last week; update the figure or add an as-of date. Verdict: publish after fixing the two flagged items — neither requires new reporting, just a wording correction and a figure update.

Verified against

Perplexity Comet 2026 browser agent · 2026-08-14

Changelog

  • 2026-08-14 — 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
Turn a vague research question into a Deep Research report with a real evidence trailA brief for Perplexity Deep Research mode that forces a stated research plan, named source priorities, a recency window, and a Finding/Source/Date/Confidence table instead of a fluent-sounding single-pass answer.Perplexity Pro (Deep Research mode)Perplexity Pro Search2026-07-22Give a Perplexity Space persistent instructions so every new thread doesn't start from zeroA custom-instructions document you paste into a Perplexity Space once, so role, source priorities, file-grounding rules, and output format apply automatically to every future thread created inside that Space.Perplexity SpacesPerplexity Pro2026-07-20Force independent corroboration before you trust a claim, not citation-cluster agreementA 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.Perplexity Pro (Sonar Pro)Perplexity Pro Search2026-07-24Audit a Perplexity answer's citations before you rely on any of themA follow-up prompt that re-checks every numbered citation in a prior Perplexity answer against the actual source text, catching claims attached to a citation that only loosely or partially supports them.Perplexity Pro (Sonar Pro)Perplexity Pro Search2026-07-26
All Perplexity prompts

Check your AI visibility

One URL in, a 0–100 score and the exact fixes out.

RUN THE CHECK

Browse all the tools

15 tools across six categories
13 of them never send your data anywhere

Free · No signup · No trial clock

SEE THE DIRECTORY