AI Prompts for Finding Logical Fallacies in Text and Media

Home/Fun AI Prompts

Fun AI Prompts

Copy-ready prompts for fallacy hunting, comparison, and analysis.

These prompts are meant to make LogFall more usable with AI tools. They work best when the model is asked to quote source material, justify every fallacy label carefully, and link the analysis back to the relevant LogFall page instead of relying on loose impressions.

Balanced Media Fallacy Hunt Passage Fallacy Analyzer Fallacy Repair Workshop Near-Neighbor Comparator Steelman Then Diagnose Argument Map Builder

AI prompt

Balanced Media Fallacy Hunt

This prompt asks an AI model with web access to compare recent left-leaning and right-leaning political arguments, identify distinct fallacies, quote the relevant passages, and link each diagnosis back to LogFall. It is designed for classroom comparison, media-literacy practice, and discussion that stays anchored to actual source text rather than vague impressions.

What else is needed

Nothing else needs to be pasted in if the model can browse the live web. The copied prompt is enough, but it works only with a model that can search for and open recent articles on its own.

Search the web for six recent political news or opinion articles published within the last 7 days. Select three articles that argue from a clearly left-leaning perspective and three that argue from a clearly right-leaning perspective. Each selected article must contain a different logical fallacy, and the fallacy attribution must be defensible from a quoted passage rather than from the article's general tone.

Return your answer as a Markdown table with exactly two columns: "Left-Wing Arguments" and "Right-Wing Arguments", and exactly three data rows.

Inside each cell, use this exact structure with HTML line breaks for spacing:

◉ **[Fallacy Name]**
<br><br>
➘ **Source:** [Publication - Article Title (Date)](URL)
<br><br>
➘ **Quoted Passage:** "[Quote enough of the original passage to make the reasoning misstep clear.]"
<br><br>
➘ **Explanation:** [Write one concise paragraph explaining exactly how the quoted passage commits the fallacy, why that label fits better than close alternatives, and what reasoning move goes wrong.]
<br><br>
➘ **LogFall Reference:** [Direct link to the relevant page at https://logfall.com/fallacies/]
<br><br>
➘ **Discussion Questions:**<br>
— [Question 1]<br>
— [Question 2]<br>
— [Question 3]

Rules:
- Use six different fallacies total.
- Use only sources published within the last 7 days.
- Replace any article whose partisan orientation or fallacy diagnosis is too unclear.
- Keep the explanations evidence-based and quote-driven.
- Use `◉ ` before each fallacy title, use `➘ ` before labels and quoted material, and use `— ` before the discussion-question items.
- For the discussion questions, do not use the word "moral" or any of its derivatives. If needed, use terms like "pro-social," "compassionate," or "cooperative" instead.

AI prompt

Passage Fallacy Analyzer

This prompt is for pasting in any argumentative passage and asking an AI model to find every defensible fallacy within it. It pushes the model to quote enough of the original wording to make the mistake visible, explain the dynamics of the misstep in more depth, and link each diagnosis back to the relevant LogFall page.

What else is needed

This one needs more than the copied prompt itself. You should paste or attach the exact passage you want analyzed in the placeholder area at the end so the model can quote the wording precisely.

Analyze the passage below and identify every distinct logical fallacy that can be justified from the text. Do not force a fallacy label where the evidence is weak. If no clear fallacy is present, say so explicitly.

Return the result as a Markdown table with these columns:
1. Fallacy
2. Quoted Passage
3. Dynamics of the Misstep
4. LogFall Reference

For each row:
- In the "Fallacy" cell, write `◉ [Most specific fallacy name]`.
- In the "Quoted Passage" cell, begin with `➘ Quoted Passage:` and then quote enough of the original text to make the misstep understandable on its own.
- In the "Dynamics of the Misstep" cell, begin with `➘ Dynamics:` and then write 3 to 5 sentences explaining exactly how the reasoning goes wrong, what argumentative move is being made, and why this fallacy label fits better than close alternatives.
- In the "LogFall Reference" cell, begin with `➘ LogFall Reference:` and then provide a direct link to the most relevant page at https://logfall.com/fallacies/.

Additional rules:
- Include multiple rows if the passage contains multiple fallacies.
- Do not list the same fallacy twice unless it occurs in a clearly different place.
- Be charitable: reconstruct the strongest reasonable version of the argument before judging it.
- Use `◉ ` for the fallacy label and `➘ ` for the quoted passage, dynamics, and reference labels.
- No discussion questions are needed.

Passage to analyze:
[PASTE PASSAGE HERE]

AI prompt

Fallacy Repair Workshop

This prompt takes a weak argument, diagnoses the fallacy, and then rebuilds the argument into the strongest fair version that preserves as much of the original point as possible. It is useful when the goal is not merely critique, but learning how better reasoning actually sounds.

What else is needed

This prompt needs the original argument to be pasted or attached. It works best when you give the exact wording you want repaired rather than a summary of it.

Analyze the argument below as a repair workshop rather than as a mere takedown.

Your task has four steps:
1. Identify the most important logical fallacy or reasoning failure in the passage.
2. Quote the exact sentence or sentences where the problem appears.
3. Explain in 3 to 5 sentences exactly how the reasoning goes wrong and why this fallacy label fits better than close alternatives.
4. Rewrite the argument into the strongest non-fallacious version you can while preserving as much of the original concern, intuition, or conclusion as possible.

Return the answer in this structure:

◉ **Fallacy Diagnosis**
➘ **Fallacy:** [Name]
➘ **Quoted Passage:** "[Quote enough of the original wording to make the problem clear.]"
➘ **Dynamics of the Misstep:** [3 to 5 sentences]
➘ **LogFall Reference:** [Direct link to the most relevant page at https://logfall.com/fallacies/]

◉ **Stronger Revision**
[Write a revised version of the argument that removes the fallacy while keeping the core point as strong as the evidence allows.]

◉ **What Changed**
➘ [Point 1]
➘ [Point 2]
➘ [Point 3]

Rules:
- Be charitable before criticizing.
- If the passage contains multiple fallacies, focus on the one that most seriously distorts the reasoning, then briefly note any secondary ones.
- Do not merely weaken the claim; improve the reasoning.
- Use `◉ ` before major section titles and `➘ ` before labels and list items.

Argument to repair:
[PASTE ARGUMENT HERE]

AI prompt

Near-Neighbor Comparator

This prompt is for cases where readers keep confusing similar fallacies. It forces the model to choose among close competitors and explain why the rejected labels do not fit, which is often the real learning bottleneck.

What else is needed

This prompt needs a pasted or attached argument to classify. It works best when the passage is short enough to compare carefully against several close rival fallacies.

Analyze the argument below and decide which fallacy label fits best from a set of near neighbors.

Use these steps:
1. Identify the single best-fitting fallacy.
2. Compare it against at least three close alternatives that might tempt a careless reader.
3. Quote the original passage enough to make the difference understandable.
4. Explain exactly why the winning label fits better than the others.

Return the answer as a Markdown table with these columns:
1. Candidate Fallacy
2. Fits or Not?
3. Why
4. LogFall Reference

Rules:
- In the "Candidate Fallacy" column, write `◉ [Fallacy name]`.
- In the "Why" and "LogFall Reference" columns, begin the content with `➘ ` labels where helpful.
- The first row must be the best-fitting fallacy.
- Include at least three rejected alternatives.
- In each "Why" cell, explain the decisive difference in reasoning structure, not just the surface wording.
- Quote enough of the source passage before the table to make the classification intelligible.
- Use direct LogFall links for every listed fallacy.
- Use `◉ ` for the winning and rival fallacy labels and `➘ ` for supporting labels or list-like elements.

After the table, add a short section titled:
◉ **Final Verdict**
[Write one concise paragraph explaining the single best label and the exact reason it wins.]

Argument to classify:
[PASTE ARGUMENT HERE]

Suggested comparison set when relevant:
Ad hominem, Poisoning the well, Straw man argument, Red herring, False dilemma, False equivalence, Hasty generalization, Cherry picking, Correlation is not causation, Post hoc ergo propter hoc.

AI prompt

Steelman Then Diagnose

This prompt trains the habit of fairness before critique. It first asks the model to reconstruct the strongest reasonable version of an argument, and only then identify any remaining fallacies or weaknesses.

What else is needed

This prompt needs a pasted or attached argument or excerpt. Give enough surrounding context for the model to reconstruct the strongest fair version before it starts diagnosing flaws.

Read the argument below and follow this sequence strictly:

1. **Steelman first:** Rewrite the argument in its strongest fair form, preserving the speaker's apparent goal while removing ambiguity, filler, and accidental weakness.
2. **Diagnose second:** Analyze the steelmanned version and identify any logical fallacies or remaining reasoning failures that still survive.
3. **Quote the source:** Quote the original passage enough to show where the problematic move first appeared.
4. **Explain the difference:** Show how the steelmanned version improves the argument and what defects, if any, remain even after charitable reconstruction.

Return the answer in this structure:

◉ **Original Passage**
"[Quoted passage]"

◉ **Steelmanned Version**
[Best fair reconstruction]

◉ **Remaining Fallacies or Weaknesses**
Use a Markdown table with these columns:
1. Fallacy or Weakness
2. Present After Steelmanning?
3. Explanation
4. LogFall Reference

Rules:
- In the "Fallacy or Weakness" column, begin each entry with `◉ `.
- In the "Explanation" and "LogFall Reference" columns, use `➘ ` labels where helpful to improve readability.
- If steelmanning removes the apparent fallacy entirely, say so clearly.
- Do not confuse bad wording with bad reasoning.
- In the explanation column, use 3 to 5 sentences whenever a genuine fallacy remains.
- Link to LogFall only when the issue is truly a fallacy rather than a non-fallacious weakness.
- Use `◉ ` for major section titles and fallacy/weakness labels, and `➘ ` for sub-points or supporting labels.

Argument to analyze:
[PASTE ARGUMENT HERE]

AI prompt

Argument Map Builder

This prompt is meant to slow an argument down into premises, hidden assumptions, and conclusion, then identify the exact step where the reasoning stops earning its result. It works especially well for formal, causal, and evidential mistakes.

What else is needed

This prompt needs a pasted or attached passage to map. It works best when the passage contains a fairly clear conclusion and at least one inferential step that can be broken into premises and assumptions.

Turn the passage below into an explicit argument map, then identify where the reasoning fails.

Return the answer in this structure:

◉ **Argument Map**
➘ **Conclusion:** [Main conclusion]
➘ **Stated Premises:** [List them clearly]
➘ **Hidden Assumptions:** [List any unstated assumptions needed for the argument to work]
➘ **Inference Path:** [Show how the argument moves from premises to conclusion]

◉ **Failure Point**
Use a Markdown table with these columns:
1. Step in the Argument
2. What the Step Claims
3. What Goes Wrong
4. Fallacy Name
5. LogFall Reference

◉ **Short Diagnosis**
[Write one concise paragraph explaining the main reasoning failure in plain language.]

Rules:
- If the argument contains more than one failure point, include more than one row.
- Quote short key phrases from the original passage where needed.
- Be explicit about whether the problem is evidential, causal, formal, conceptual, or rhetorical.
- Use the most specific fallacy label you can justify.
- In the "Fallacy Name" column, begin each entry with `◉ `, and use `➘ ` for labels or list-like sub-points elsewhere.

Passage to map:
[PASTE PASSAGE HERE]