Comparing two theories needs more than “they both talk about X”. A strong comparison surfaces the assumptions, scope conditions, evidence base, and predictive divergence that a reader cannot see from a summary. The 15 copy-ready prompts below turn a chat model into a methodologist that pushes you past surface similarity and into the actual disagreements.
TL;DR: Use the assumption matrix (template 1) and the treacherous-terms check (template 10) first — they catch the two errors that wreck most comparisons. Keep critiques symmetrical by always running templates 7 and 8 as a pair. Anchor every claim in primary texts; AI defaults to summarizing secondary critics. For long source PDFs, paste them into a 1M-token model (Gemini 3.1 Pro or Claude Opus 4.7) or NotebookLM so answers cite your sources, not the model’s memory.
Who this is for
Graduate students preparing comprehensive exams, theory-paper authors, tutorial and seminar leaders, and undergraduates writing comparative essays in philosophy, social science, or natural science.
When not to use these prompts
Skip if you have not read the primary sources for both theories — secondary summaries are a brittle base, and the model will happily extend a textbook caricature. Skip too for theories outside the model’s training data (very obscure or published after the cutoff); paste the source text in directly and verify outputs against it.
Prompt anatomy / structure formula
A comparative-theory prompt should always carry six elements:
- Role: who the AI plays — research tutor, peer reviewer, exam coach, debate partner, librarian.
- Context: your level, subject, deadline, paper count, target citation style, course or program.
- Goal: one concrete deliverable — 12 quiz items, a 1-page lit matrix, 5 counter-arguments, a 4-week revision plan.
- Constraints: word count, depth, source types allowed, what to skip, what to never claim.
- Output format: numbered list, table, JSON, or graded blocks (E / M / H) so you can paste into Notion / Anki / Word.
- Examples / signal: 1-2 reference paragraphs or anti-examples (“not the way Wikipedia explains it”).
Best for
- Comp exam preparation
- Theory paper / chapter drafting
- Tutorial discussion design
- Lit review theoretical section
- Undergraduate comparative essay
Which AI model to use (as of June 2026)
Model choice matters more here than for most prompt work, because theory comparison rewards two things: long context (so the model reasons over the actual texts, not a paraphrase) and disciplined reasoning (so it does not smooth over real disagreements).
| Model | Context window | Best for theory comparison | Notes |
|---|---|---|---|
| Gemini 3.1 Pro | 1M tokens | Load full PDFs of both primary texts in one pass | Released Feb 19 2026; 1M context standard; Deep Think mode for the hard assumption work |
| Claude Opus 4.7 | 1M tokens | Symmetrical critique, careful synthesis vs. incommensurability calls | Strong at holding two positions without collapsing them; workhorse Sonnet 4.6 is fine for drafting |
| ChatGPT (GPT-5.5) | ~320 pages in-app on Plus | Quick comparisons, essay paragraphs | Full 1M context only on the $200 Pro tier; pick Thinking in the model picker for assumption-level work |
| NotebookLM | Grounded to your uploads | Citations with page numbers back to your sources | Free; every answer links to the passage, so you can verify before you quote — it does not format APA/MLA for you |
Pricing for reference (June 2026): ChatGPT Plus $20/mo, Pro $100 or $200; Claude Pro $20/mo, Max $100 or $200; Google AI Pro $19.99/mo (the tier formerly called Gemini Advanced). For source-grounded work where every claim must trace to a text, NotebookLM is the safest default and is free. For raw reasoning over two long texts, Gemini 3.1 Pro and Claude Opus 4.7 both take 1M tokens.
15 copy-ready prompt templates
1. Assumption-by-assumption matrix
Default comparison; reveals the real disagreements.
You are a theory professor. Compare {Theory A} and {Theory B} on their core assumptions. Output a 5-row table: assumption topic | Theory A position | Theory B position | where they overlap | where they diverge. Cite one foundational text per theory.Variables to swap: Theory A, Theory B
Optimization: If overlap is overstated, add: “Be precise: surface convergence often hides deep divergence. Note any cases where both use the same word for different concepts.”
2. Scope conditions comparison
For {Theory A} and {Theory B}, identify the scope conditions under which each is meant to apply. Where do they overlap, where does only one apply, where does neither claim authority? Cite the theorists or texts that explicitly bound the theory.3. Evidence-base contrast
What kinds of evidence does {Theory A} typically marshal? What about {Theory B}? Compare on: empirical type (experiment, fieldwork, archive, formal model), prototypical cases, and what counts as a refutation.4. Predictive divergence
On the question "{question}", what does {Theory A} predict and what does {Theory B} predict? Where do the predictions diverge most sharply? What empirical observation could decide between them?5. Common-cases test
Take 3 well-known cases in {field}: {case 1, case 2, case 3}. For each, how does {Theory A} explain it? {Theory B}? Where do their explanations differ, where do they converge?6. Lineage and influence
Sketch the intellectual lineage of {Theory A} and {Theory B}: predecessor thinkers, the problem each was reacting against, and the contemporary descendants. Note any place where Theory A descends from a critic of Theory B (or vice versa).7. Critique-from-A-against-B
Write a 200-word critique of {Theory B} from the standpoint of {Theory A}: which assumptions of B does A reject, why, what implications follow. Use the vocabulary a serious proponent of A would use.8. Reciprocal critique
Now reverse it: write a 200-word critique of {Theory A} from {Theory B}'s standpoint. Maintain the symmetry; do not stack the deck.9. Synthesis or incommensurability
Can {Theory A} and {Theory B} be synthesized, or are they fundamentally incommensurable? Give the case for each answer in 100 words each. End with which case is stronger and why.10. Glossary alignment
List 6 key terms used by both {Theory A} and {Theory B}. For each: definition in A, definition in B, whether the terms mean the same thing. Flag any words that look like cognates but refer to different concepts (treacherous terms).11. Empirical-test design
Design a study that would distinguish {Theory A} from {Theory B}: the question, the empirical setting, the measurable outcomes, and what result would favor each theory. Constraints: feasible for a {grad student / lab} to run.12. Visual diagram outline
Describe a Venn / 2x2 / cladogram diagram showing the relationship of {Theory A} and {Theory B}. List the regions / cells / branches and what should be placed in each. The reader should be able to draw it from your description.13. Common misreadings
List 4 common misreadings of {Theory A} (especially as filtered through critics or textbooks) and 4 common misreadings of {Theory B}. For each: the misreading, why it persists, and the corrective text or passage to read.14. Comparative essay paragraph
Write a 200-word body paragraph for a comparative essay arguing "{thesis}" using {Theory A} and {Theory B}. Show me how to acknowledge the strengths of each while supporting my thesis. Cite at least 2 sources by author + year.15. Comp-exam ready summary
Produce a 1-page exam-ready summary comparing {Theory A} and {Theory B}: 3 assumptions of A, 3 of B, 3 points of agreement, 3 points of conflict, 2 cases each best explains, and 1 sentence on why this comparison matters for {field}.Common mistakes
- Comparing only the headline claims, not the underlying assumptions.
- Using the same word in both theories without checking whether the meaning matches (template 10 catches this).
- Stacking the deck — making one theory look strong by paraphrasing fairly and the other shallow.
- Forcing synthesis when incommensurability is the honest answer.
- Citing only secondary critics instead of the primary texts.
- Treating “more recent” as “better”; older theories often hold on important domains.
- Skipping scope conditions; many “disagreements” are really about different domains.
How to push results further
- Always pair template 7 and template 8 to keep the critique symmetrical.
- Use template 10 early; treacherous terms cause most of the confusion in comparative work.
- Anchor every comparison in primary texts; AI summarizes secondary by default.
- For comp exams, run template 15 for every theory pair on your reading list.
- When the answer is “incommensurable”, say so; honesty beats forced synthesis.
- Pair theoretical comparison with empirical-test design (template 11) when possible.
- Save the diagram outline (template 12) as a visual for tutorials.
FAQ
- How accurate is AI on theory comparison?: Good on well-known pairs (utilitarianism vs Kantian ethics, Keynesian vs monetarist, realism vs constructivism in IR). It is weaker on obscure or post-cutoff debates, where it will confabulate plausible-sounding positions. The fix is to paste the primary texts in and force the model to quote them — that is exactly what NotebookLM’s page-level citations are built for.
- Which model should I use?: For reasoning over both full texts at once, Gemini 3.1 Pro or Claude Opus 4.7 (both 1M-token context as of June 2026). For grounded answers that cite your uploaded PDFs by page, use the free NotebookLM. For quick essay paragraphs, ChatGPT (GPT-5.5) on the Thinking setting is fine.
- How long should a comparison be?: For an essay paragraph, ~200 words. For a comp exam answer, 800-1200. For a thesis chapter, 5000+ across multiple sections.
- Should I argue for one side?: For an essay, yes. For a lit review, usually no — map the field first. For comp exams, name the strengths of each, then position your stance.
- Can AI judge which theory is correct?: No. It can describe positions and pinpoint where they diverge, but the verdict depends on evidence and values you supply. Treat it as a research tutor, not a referee.
- What if the two theories barely overlap?: Run template 2 (scope conditions). Comparison sometimes reduces to “these answer different questions”, and that is a legitimate, citable finding.
Related
- Theory simplification prompts
- Concept comparison prompts
- Research summary prompts
- Thesis topic prompts
- Counterargument Generation Prompts for Stress-Testing Essays
- Learning & Research Prompts hub
For primary-source grounding, cross-check the model against the Stanford Encyclopedia of Philosophy (peer-reviewed entries with full bibliographies) before you quote any position.