Gemini

Verified against Gemini · 2026-07-22

Compare competing sources side by side without smoothing over where they disagree

A long-context prompt that loads several full reports, reviews, or papers at once and produces a structured comparison table instead of a blended, opinion-free summary that quietly hides where the sources actually disagree.

Gemini (Gemini 3 Pro)Gemini app6 fillable variables

The prompt

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

I'm giving you 4 sources about the health effects of intermittent fasting: a 2025 meta-analysis in a peer-reviewed journal, a hospital patient-education page, a popular-science book chapter, and a fitness influencer's long-form article

Read each one in full before comparing. Build a comparison, not a summary — a summary that blends four positions into one paragraph is exactly what this prompt is meant to avoid.

COMPARISON TABLE
Rows: claimed metabolic effects, recommended fasting window, cited study population size, stated caveats or limitations
Columns: one per source, using the source name/title as the column header.
For each cell, give the source's actual position, with a short quote or close paraphrase you could point back to if challenged — not your own averaged judgment of what the sources probably agree on.

AFTER THE TABLE
1. List every criterion where the sources actively disagree — not just differ slightly in emphasis or wording, but state a genuinely different position.
2. For each disagreement, classify it as one of: a factual dispute (they can't both be right), a difference in scope or timeframe (both may be correct for different situations), or a difference in methodology (they measured or defined something differently) — someone deciding whether to try it, with no medical background needs a different response to each of those three, so don't leave them lumped together as one undifferentiated "disagreement."
3. Do not resolve the disagreements for me by picking a winner unless I explicitly ask. Flag them and stop — presenting one position as the correct one defeats the purpose of asking for a comparison in the first place.
4. Name anything relevant to whether this is safe to try alongside an existing type 2 diabetes management plan that none of the 4 sources actually address, so I know where the evidence runs out.

SOURCE CREDIBILITY, NOTED NOT ENFORCED
Note anything relevant about each source's likely reliability — is it peer-reviewed, is it an official patient-education resource, is it an individual's personal account — as a separate line under the table, not folded into the table cells themselves. I want to weigh that myself once I can see all the positions side by side; don't let a source's perceived authority quietly influence how confidently you state its position in the comparison itself.

WHEN A SOURCE HEDGES
If a source itself states its finding with a caveat ("preliminary," "in a small sample," "further research needed"), carry that hedge into the table rather than smoothing it into an unqualified claim — a source that hedges its own conclusion is telling you something different than one that states it flatly, and collapsing that distinction in the retelling misrepresents what the source actually said.

IF TWO SOURCES AGREE BUT FOR DIFFERENT REASONS
Note it explicitly rather than treating agreement as automatically meaning the same underlying claim — two sources can land on the same recommendation while relying on different, even contradictory, reasoning, and that's worth surfacing rather than collapsing into simple agreement.

Customize

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

Why this works

Loading every full source into the same long context window, rather than pre-summarizing each one in a separate pass, is what makes an honest comparison possible — a summary of a summary loses the specific claim, quote, or number needed to say precisely where two sources diverge, and by the time four separate summaries are stitched together, the original wording that would let you verify the disagreement is already gone. The explicit ban on picking a winner unless asked directly counters the strongest pull on a long-context synthesis task: models tend to resolve conflicting inputs into one confident, harmonized answer, which reads well but erases the disagreement the person doing the comparison actually needs to see and weigh for themselves — a comparison that quietly picks a side has stopped being a comparison. Classifying each disagreement as factual, scope-based, or methodological gives the reader an actionable next step instead of just the fact that a conflict exists — a factual dispute needs a tie-breaker source or more evidence, a scope difference might mean both sources are simply correct for different populations or timeframes, and a methodology difference means the numbers were never actually measuring the same thing in the first place, which is a completely different problem to have than a factual contradiction. Requiring a quote or close paraphrase in every cell, rather than the model's restated version of the position, keeps the whole table auditable — anyone who doubts a cell can check it against the actual source text instead of trusting the model's characterization of what that source said. Keeping source-credibility notes in a separate row rather than folded into the comparison cells themselves prevents a subtle but real distortion: once a model starts weighting how confidently it states a position by its own judgment of the source's authority, the comparison stops being a neutral record of what each source claims and becomes a pre-filtered version of it, which defeats the purpose of asking to see all positions side by side in the first place. Preserving a source's own hedging language rather than smoothing it into an unqualified claim matters because a finding stated as preliminary and a finding stated as established are different pieces of evidence even when the headline claim reads identically, and losing that distinction in translation makes two sources sound more aligned in certainty than they actually are.

Verified against

Gemini Gemini 3 Pro · 2026-07-22

Changelog

  • 2026-07-22 Initial publish, verified against Gemini 3 Pro comparing four sources with genuinely conflicting claims.

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 stack of scattered documents into one sourced decision memoA long-context synthesis prompt that loads every related document into the same context window as one connected evidence base and returns a single decision memo with inline citations and flagged disagreements, instead of a chain of disconnected per-document summaries that lose the connections between them.Gemini (Gemini 3 Pro)Gemini app2026-07-20Diff a redlined contract against the original and flag every substantive changeA long-context prompt that loads both the original and the redlined version of a contract into the same window and returns a change-by-change risk analysis, instead of a generic first read that misses a clause quietly weakened somewhere the visible tracked-changes markup doesn't draw the eye.Gemini (Gemini 3 Pro)Gemini app2026-07-21Turn an entire small-to-mid codebase into an onboarding doc a new engineer can trustA long-context prompt that reads a whole repository — every file pasted or uploaded, not a handful of representative snippets — and produces an onboarding document that names actual file paths and real architectural seams, instead of generic advice that could describe any Next.js or Python project.Gemini (Gemini 3 Pro)Gemini app2026-07-23Turn an uploaded meeting or presentation recording into timestamped notes and action itemsA native multimodal prompt for Gemini's video understanding that returns timestamped highlights, decisions, and action items — including what was shown on screen, not just said out loud — instead of a flat transcript of everything that happened.Gemini (Gemini 3 Pro)Gemini app2026-07-24
All Gemini 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