How to make the argument land honestly — classical appeals, dialectical moves, voice and framing devices, AI anti-patterns, and fallacies. From the Rhetoric Playbook.
Make the argument's logical structure visible and defensible — premises, inferences, conclusion.
Outputs
Explicit reasoning chain with named inference moves, evidence anchors, and stated assumptions.
How it differs
Logos in writing is not just being correct — it is making the shape of the argument legible to the reader. A sound argument that hides its structure reads as assertion.
Best for
Counterintuitive claims, technical arguments, posts where the reader needs to follow you to a non-obvious conclusion.
Failure mode
"Logical-sounding" prose that uses the vocabulary of logic (therefore, thus, it follows) without doing the work — AI's most common forgery.
Connect the claim to a specific emotional state the reader is in or could be moved to.
Outputs
Named emotion, the trigger that produces it, and a concrete moment in the draft where the emotional register changes.
How it differs
Pathos is not "make it feel-y." It is identifying a specific emotion (frustration, hope, vertigo, recognition) and earning it through specific detail rather than generic warmth.
Best for
Topics where the reader's lived experience is more persuasive than the data; opening hooks; closing moves.
Failure mode
Generic warmth or programmed empathy ("this can be challenging") that reads as AI comfort copy.
Acknowledge specific valid points in the opposing position before pressing your own.
Outputs
A specific, narrowly-scoped concession that surrenders something real but doesn't surrender the larger claim.
How it differs
Generic balance ("there are good points on both sides") concedes nothing. Real concession names a specific point and grants it. The result is increased credibility on the rest.
Best for
Posts where the audience expects you to acknowledge complexity; arguments where one part is genuinely contested.
Failure mode
Concession so broad it dissolves the argument; or so narrow it reads as performative.
Advance an argument through a sequence of pointed questions that expose assumptions in the opposing view.
Outputs
A 3-5 question sequence where each question forces a clarification or concession.
How it differs
Direct refutation tells the reader the opposition is wrong. Socratic questioning lets them arrive at the same conclusion themselves — which makes the conclusion stickier.
Best for
Posts about contested premises, hidden assumptions, or domains where directly stating the conclusion would feel preachy.
Failure mode
Loaded questions disguised as Socratic — questions that smuggle the conclusion into the premise.
Change the category, level, or vocabulary in which an argument is being conducted.
Outputs
A reframed version of the debate where the original opposition no longer makes sense or becomes a special case.
How it differs
Most arguments are conducted inside a frame both sides accept. Reframing rejects the frame. The risk is that you sound evasive; the reward is that you change the question.
Best for
Debates that have become stuck in unproductive opposition; posts that need to introduce a new category or concept.
Failure mode
Reframe that reads as dodging the question rather than dissolving it.
Pair a real concession with a sharp refutation, in that order.
Outputs
A paragraph or section that grants something to the opposition then takes it back at a higher level.
How it differs
Most posts either concede broadly (and dissolve their claim) or refuse to concede (and read as defensive). The architectural move is to concede narrowly and then raise the stakes.
Best for
Sections where the reader is most likely to push back; transitions from objection-handling to forward argument.
Failure mode
Concession that is so weak it functions as a setup for an attack — readers smell the trick.
Set two ideas in deliberate contrast to sharpen the distinction between them.
Outputs
A paired construction — "X is one thing; Y is another" — that reveals what would otherwise be implicit.
How it differs
AI uses antithesis as decoration ("not just X, but Y"). Productive antithesis distinguishes two things the reader was conflating. The test is whether the contrast does work — does it make a distinction that changes how the reader thinks?
Best for
Defining terms; clarifying confused debates; establishing the move from one frame to another.
Failure mode
Antithesis that contrasts two things nobody was confusing, producing rhythm without insight. See anti:false-reframe.
Refute a position by extending it to a conclusion the opposition cannot accept.
Outputs
A logical chain showing that the opposition's premise leads to an absurd or untenable result.
How it differs
Direct refutation argues the opposition is wrong. Reductio shows that following their reasoning produces something even they would reject — a sharper move.
Best for
Premises that look reasonable on their own but have unacceptable implications.
Failure mode
Slippery slope masquerading as reductio — you don't actually show the implication, you just gesture at a worse outcome. See fallacy:slippery-slope.
State the opposing claim and explain precisely why it is wrong.
Outputs
A paragraph that names the opposing claim, identifies the specific failure (false premise, invalid inference, missing evidence), and replaces it.
How it differs
Most rhetorical responses dodge — reframe, change topic, attack character. Direct refutation does the work: here is the claim, here is why it fails, here is what is true instead.
Best for
Specific factual or logical errors that need to be named; arguments where the reader expects you to engage rather than evade.
Failure mode
Refutation that beats up a weak version (fallacy:strawman) or attacks the speaker (fallacy:ad-hominem) instead of the claim.
Open a section with a strong claim, not a topic sentence.
Outputs
An opening sentence that commits to a position rather than announcing what will be discussed.
How it differs
"This section discusses AI writing voice" is a topic sentence. "There is a particular voice that AI writes in" is a claim. The latter pulls the reader in; the former tells them what they're about to read.
Best for
Section openings; the first sentence after a heading; anywhere the prose is at risk of becoming summary.
Failure mode
Declarative-sounding sentence that doesn't actually claim anything ("AI is changing the way we write" — claims nothing specific).
Use first person to take positions, not just to soften claims.
Outputs
First-person sentences where the "I" carries weight — "I think this is the most underappreciated finding," "I want to argue that."
How it differs
Most academic writing avoids "I" to feel objective. Most AI writing avoids "I" because the model has no self. Strategic first person announces that a person is making a judgment, which the reader can take or leave.
Best for
The turns of an argument; moments of evaluation; closing paragraphs where the writer is asking the reader to agree.
Failure mode
First-person filler ("I believe that") used as hedging rather than commitment. The first person should make the claim stronger, not softer.
Speak directly to the reader using "you" at the turns of the argument.
Outputs
Sentences that engage the reader's perspective at moments of stake-raising or transition.
How it differs
Like first person, "you" should appear at deliberate moments — where the argument shifts, where the stakes rise, where you want the reader to pause. Constant "you" reads as a sales pitch.
Best for
Section transitions; moments where the implication shifts from abstract to personal; closing moves.
Failure mode
"You" everywhere, eroding the effect; or "you" at the wrong moment, accusing the reader rather than engaging them.
Use a question to mark a transition, not to fill space.
Outputs
A question that sits at a section break or argument shift, inviting the reader forward into the next move.
How it differs
AI uses rhetorical questions as filler ("What does this mean?"). Productive use places the question at a deliberate turn where the reader is themselves about to ask it.
Best for
Between sections; before a counterintuitive claim; before a structural shift.
Failure mode
Question with no payoff — the prose answers it generically or doesn't answer it at all.
Cite quantified findings with the actual numbers, not summaries.
Outputs
Statements with specific values — "32% vs 90%," "Cohen's d 1.5 vs 0.2," "87% correct rationales but 64% correct actions."
How it differs
"Many" and "most" and "significant" are AI's vocabulary. Specific numbers say a person looked at the data. Even when the number isn't load-bearing, the specificity is a credibility signal.
Best for
Claims grounded in research; arguments that depend on magnitude.
Failure mode
Numbers cited without context, becoming decorative; or false precision (single-digit precision on noisy data).
Adopt a specific cognitive and tonal stance before writing or analyzing a passage.
Outputs
Output colored by the selected stance — vocabulary, attention pattern, default moves all shift.
How it differs
Persona activation that changes the kind of attention being applied, not just the surface tone. Skeptical stance asks different questions than empathetic stance.
Best for
Shifting between drafts, matching platform expectations, recovering from a flat passage by changing the underlying lens.
Failure mode
Persona that overrides substance — style without content. Or stance that contradicts the topic (e.g., provocative on a sensitive topic).
Lead with what the reader stands to lose if they ignore the argument.
Outputs
Opening hook or section frame organized around stakes-of-inaction.
How it differs
Behavioral economics shows loss aversion: people respond more strongly to potential losses than equivalent gains. Loss framing leverages this — but it can read as alarmist if overdone.
Best for
Arguments where the cost of ignoring the issue is real and concrete; warning-shaped posts.
Failure mode
Manufactured urgency ("if you don't read this, you'll fall behind") that reads as marketing copy.
Lead with what the reader gains from understanding or acting on the argument.
Outputs
Opening or section frame organized around opportunity, capability, or insight.
How it differs
Often weaker than loss frames in emotional pull but more sustainable for long-form arguments. Gain framing is the natural fit for posts about new capabilities or unlocked moves.
Best for
Posts introducing a new framework or tool; arguments about expanded possibility.
Failure mode
Vague benefit language ("unlock your potential") indistinguishable from self-help marketing.
Frame the argument as a contrast between two positions, then defend the second.
Outputs
A frame of the form "Most people think X. The truth is Y" — but earned, not performed.
How it differs
AI overuses this frame as decoration ("Not just X, but Y"). Productive antithesis identifies a real misconception and replaces it. The test: does the reader actually hold the X position?
Best for
Posts that aim to change a specific belief held by the audience.
Failure mode
Strawmanning the X position (fallacy:strawman) — making it weak so the Y position looks strong by contrast.
Open with two true things that seem to contradict, then resolve them.
Outputs
A hook or section that names a tension and then dissolves it.
How it differs
Paradox creates productive cognitive dissonance — the reader wants the resolution. Done well, this is the strongest opening in long-form. Done poorly, it reads as clever-for-its-own-sake.
Best for
Counterintuitive arguments; posts integrating two domains; hooks for long-form.
Failure mode
False paradox where the two "contradictory" things don't actually contradict (the reader smells the trick).
Frame the argument around who the reader is or who they want to be.
Outputs
A frame that activates an aspirational, in-group, or contrarian identity.
How it differs
Most arguments target beliefs. Identity frames target self-concept. This is high-leverage but high-risk: get it right and the reader adopts the position because it fits who they are; get it wrong and you trigger identity threat.
Best for
Topics that touch professional or intellectual identity; readers who already self-identify with a group.
Failure mode
Flattering the reader without earning it; or accidentally activating an out-group threat that closes them off.
Frame the argument as time-sensitive without manufacturing urgency.
Outputs
A frame that establishes why this argument matters now — specific to current discourse, recent events, or emerging conditions.
How it differs
Real urgency identifies a window. Manufactured urgency uses words like "now more than ever" without specifying what changed. The former is kairos; the latter is filler.
Best for
Commentary; posts entering active debates; arguments that have become unavoidable due to recent events.
Failure mode
Generic urgency vocabulary ("more important than ever," "in today's rapidly changing landscape") that signals AI.
Structure the argument as a story arc — discovery, warning, paradox, or journey.
Outputs
Argument cast as narrative with arc beats: setup, complication, resolution.
How it differs
Pure exposition states facts. Narrative draws the reader through a sequence. The risk is that narrative can over-coherence — making the messy real story too neat.
Best for
Long-form posts; arguments where the reader needs to follow a path; complex causal claims.
Failure mode
Fake journey ("I used to think X, then I discovered Y") when the writer never actually held X. Performed narrative.
Triadic enumeration ("fast, scalable, secure") used reflexively rather than purposefully.
Failure mode
The rule of three is genuinely effective when each item adds something. AI uses it to fake comprehensiveness — three adjectives where one would do. Readers trained on AI prose now register triadic structure as filler.
Ascending pattern that escalates a claim through "not just X, but Y," typically reaching a grand third term.
Failure mode
Done occasionally by skilled writers, this can land. Done by AI, it lands every paragraph. The structure creates an illusion of depth (mild → moderate → grand) without actual complexity.
Symmetrical "while X has benefits, it also carries risks" constructions that avoid commitment.
Failure mode
Reads as risk-averse design rather than considered judgment. Substitutes vague neutrality for actual evaluation. The reader cannot tell what the writer thinks because the writer hasn't taken a position.
Climactic-sounding sentences that promise transformation or revelation — "It's not just about X. It's about becoming Y."
Failure mode
Once or twice in a post, this lands. Multiple times, it reads as TED-talk template. The structure (antithesis + parallel + climactic progression) becomes recognizable independent of content.
Templated emotional warmth — "It's completely normal to feel uncertain," "Many others have felt this way."
Failure mode
Programmed empathy substitutes for real engagement with the reader's situation. Recognizable as a tone borrowed from self-help, HR copy, and therapy chatbots.
Announcing what the post will do rather than doing it — "In this section, I will explore three reasons..."
Failure mode
Reads as textbook scaffolding. The reader wants the content, not the meta-content. AI writes signposts because it cannot trust itself to land the move; humans should trust themselves.
Cluster of words AI overuses: delve, leverage, robust, comprehensive, intricate, tapestry, landscape, pivotal, vibrant, meticulously, elucidate, garner, foster, underscore, navigate (as a metaphor), bolster, harness, illuminate.
Failure mode
Individually, none of these words is wrong. In clusters, they are AI's signature. Studies of post-2022 academic writing show measurable spikes in these words. Readers attuned to AI prose register them as a fingerprint.
Avoiding repetition by using awkward synonyms — "the protagonist," "the eponymous character," "the key player."
Failure mode
AI's repetition penalty makes it substitute progressively stranger synonyms. Repeating the actual name is usually better than the synonym chain. Journalists call these "popular orange vegetables."
Constructions that retroactively challenge an unstated misconception — "It's not just X, it's also Y," "However, it's important to note...".
Failure mode
Reads as preemptive defense against an objection the reader wasn't making. The "however" or "but" pivot is overused as connective tissue rather than as a real turn.
Inflating the importance of the topic with phrases like "stands as a testament," "plays a vital role," "marks a turning point," "represents a paradigm shift."
Failure mode
AI puffs up significance because it has no judgment about what is actually significant. The vocabulary of importance becomes filler. Readers register it as marketing.
Vague attribution borrows credibility without earning it. Readers cannot verify; the attribution shields the writer from accountability while still gesturing at authority.
Studies show LLM-era prose has measurably less use of "is" and "are." The substitutes feel grandiose. "He served as a candidate" instead of "He was a candidate" reads as inflated.
Trailing "-ing" phrases that gesture at significance without doing analytical work — "highlighting the importance of...," "reflecting broader trends...," "contributing to..."
Failure mode
The participial tail looks like analysis but is decoration. The reader cannot tell what was added. Often paired with vague attributions to make it sound substantive.
