What we lose when AI writes for us

Afonso Franco · email · 2026-05-25 Reader 🖥️

AI Slop Cognitive Surrender Writing as Thinking

When we let AI write for us, we don't merely outsource a technical task — we outsource the cognitive struggle through which we clarify beliefs, form judgment, and constitute the self, and at civilizational scale this risks converging humanity not just on a shared style but on shared thought.

practical · mixed — Mixed expository/argumentative essay with a strong practical thrust: Franco wants the reader to change behavior — to keep writing from a blank page, to use LLMs as co-creator rather than ghostwriter. The argumentative spine (writing-is-thinking; outsourcing writing outsources selfhood) is in service of a prescriptive conclusion.

Structure

Part	Role
Opening framing: 'The slow death of the Self' — distinguishes AI removing technical barriers (good) from AI removing cognitive struggle (bad).	Thesis statement and conceptual setup
'The empty page and the AI slop' — phenomenology of the blank page plus a taxonomy of LLM stylistic tells (false dichotomies, anaphora, epistrophe, concrete-to-abstract pivots, em-dashes).	Evidence by ostension — shows the homogenized style is already pervasive, making the threat concrete
'Human writing' — gallery of named authors (de Botton, Burkeman, Mollick, Ritson, Wodtke, Horowitz, Cutler, Pollan, Sapolsky, Torres) whose prose is recognizably theirs.	Counter-evidence: what is at stake — the perceptible signature of a Self in prose
Pivot to the deeper claim: 'The real risk is not that we all start sounding the same. It's that we all start thinking the same.'	Escalation from aesthetic to cognitive/epistemic stakes
Closing reframe of the blank page as a mirror — the page-as-Self-revelation metaphor.	Conclusion / rhetorical clincher

Key terms

Self Writing as Thinking
The crystallized set of beliefs, judgments, and identity that emerges through the act of articulating one's thoughts in writing; not a pre-existing essence but something constituted by the struggle of expression.
Cognitive struggle Writing as Thinking
The slow, uncomfortable mental work of turning vague feelings and ideas into coherent thoughts — distinct from a 'technical barrier' that AI may legitimately remove.
AI slop AI Slop
The recognizable homogenized LLM style — short sentences, false dichotomies, anaphora/epistrophe, concrete-to-abstract pivots, em-dashes — that Franco can now 'smell' and that has become 'humanity's writing style'.
Linguistic convergence (at civilizational scale) Writing as Thinking
A society-wide collapse of rhetorical diversity into a single LLM-derived style, with the further implication that thought itself converges along with rhetoric.
Co-creator Writing as Thinking
An acceptable mode of LLM use in which the human remains 'in the driver's seat,' goes for walks when stuck, and uses the LLM for sparring/improvement — contrasted with using LLMs to generate the text wholesale.
Writing Writing as Thinking
Not the output but the struggle: the slow process of turning feelings into coherent thoughts; equivalent to thinking ('writing is thinking').

Propositions

explicitdevelopedAI removing a technical barrier (e.g., coding, image generation) is good; AI removing cognitive struggle is bad.conf: high
refines ←Det skal være vanskelig · p12 — Franco frames AI as monolithically corrosive to cognitive struggle ('AI removing cognitive struggle is bad'). This piece refines that: the same model is an autopilot when it removes struggle but a simulator when it imposes it, so whether AI is corrosive is a design-and-incentive choice, not an intrinsic property.
explicitcorroboratedWriting is thinking — the cognitive struggle of articulation is not separable from the cognitive work of forming beliefs.conf: high
extends ←Teaching Coding When AI Can Write the Code · p1 — Generalizes the 'the medium is the evidence of thinking, and AI severs that link' claim from writing (Franco: writing is thinking) to code (Freeman: code was a window into thinking).
supports ←Writing vs. Thinking · p1 — Near-identical thesis independently argued: both claim writing is thinking, not a separable report of thought formed beforehand.
explicitcontestedThe Self is constituted (not merely reported) by the act of writing from a blank page.conf: high
contradicts ←How to write with style · p12 — Franco p3 insists the Self is 'constituted (not merely reported)' by the act of writing. Vonnegut occupies precisely the 'merely reported' position Franco rejects: style reveals a self that is largely pre-formed by origin and childhood speech ('which is what I am'), not created by the blank page. This is a genuine claim-to-claim contradiction on the report-vs-constitute axis — and it exposes an internal tension within Franco himself, since his p11 (style indexes an existing Self) leans toward the report side that his p3 disavows. Note: Vonnegut only occupies the report side; he offers no rival theory of self-formation.
extends ←Writing vs. Thinking · p12 — Franco argues the Self is constituted by writing from a blank page; this doc extends the same self-formation claim to code artifacts by arguing code-writing is not a substitute for prose-writing but an instance of the same activity.
explicitOutsourcing writing to AI outsources self-expression and ultimately self-formation.conf: high
explicitLLM-generated prose has identifiable stylistic markers (short sentences, false dichotomies, anaphora, epistrophe, concrete-to-abstract pivots, em-dashes) that have become pervasive in public writing.conf: high
explicitLinguistic convergence at civilizational scale will produce cognitive convergence — we will start thinking the same, not just sounding the same.conf: high
explicitestablishedRecognizable authorial voice (de Botton, Sapolsky, Cutler, etc.) is evidence of a Self being expressed and is what readers connect to.conf: high
supports ←The best response to AI slop, infinite advice, and online noise is from Robin Williams · p4 — Franco: recognizable authorial voice is evidence of a Self being expressed and is what readers connect to. Acunzo supplies the source of that voice — it comes from the creator re-living their own experience ('how YOU see the world separates you'), giving Franco's voice-as-Self claim its causal origin.
supports ←How to write with style · p6 — Vonnegut's 'sound like yourself' claim — that a writer is trusted most, by himself and by readers, when the prose sounds like the actual person he is — directly supports Franco's thesis that recognizable authorial voice is what readers connect to. Vonnegut's 'others seem to trust it most, too, when I sound most like a person from Indianapolis' is the reader-connection Franco asserts, stated 40 years earlier.
explicitcorroboratedUsing an LLM as a sparring partner / editor is acceptable; using it to generate the text is not.conf: high
supports ←Co-Existence and the End of Co-Intelligence · p5 — Mollick's own practice — AI as sparring partner/editor for the prose, never as drafter — matches Franco's line that using an LLM to edit is acceptable while using it to generate the text is not.
explicitcorroboratedThe friction and discomfort of writing are not bugs but the very mechanism of selfhood.conf: high
supports ←Det skal være vanskelig · p6 — Franco claims the friction and discomfort of writing are not bugs but the mechanism of selfhood. This piece supplies an independent learning-science grounding via Bjork's desirable difficulties: the effort is not a flaw in learning, it is the mechanism — and AI is a machine for removing exactly that effort.
implicitWriting is the privileged or unique site of selfhood-formation — other effortful cognitive activities (drawing, coding, conversation, journaling without writing) are not assigned the same role.conf: high
implicitcorroboratedAn author's prose style is a transparent and reliable index of their inner Self.conf: high
supports ←How to write with style · p8 — Vonnegut's founding move — style is a set of 'revelations' that tell the reader whether the writer is honest, informed, playful, etc. — is exactly Franco's (implicit) premise that prose style is a reliable index of the author's inner Self. Both treat style as diagnostic of character.
implicitCognitive effort and authentic selfhood require self-generated text — text co-produced or AI-edited cannot do the same work, even if the human deliberates over every word.conf: high
implicitStylistic diversity is a proxy for cognitive diversity — converging styles entail converging minds.conf: high
implicitPre-AI writing practice was already producing genuine selves at scale (i.e., there was a baseline of authentic selfhood now being eroded).conf: medium
implicitThe reader is, or could be, a person who writes from a blank page — i.e., the audience is people for whom writing is already a self-shaping practice.conf: medium

Arguments

Letting AI write for you erodes selfhood. strength: medium

Explicit premises:

Writing is the process by which we clarify beliefs, form judgment, and shape the self.
AI writing for you removes the cognitive struggle of that process.
Removing the cognitive struggle removes the mechanism of self-formation.

Implicit premises:

The Self is constituted by, not just expressed in, writing (a constructivist theory of selfhood).
No other cognitive activity (speaking, drawing, deliberating in conversation, etc.) can substitute for writing in this role.
AI-assisted or AI-edited writing is categorically different from AI-generated writing — only the latter is corrosive.

Inferences:

If selfhood is constituted by effortful articulation, then heavy LLM use across other articulation domains (slide-making, email drafting, slack messages) should be similarly corrosive — yet Franco doesn't extend the argument there.
If the Self is built through struggle, then practices that restore the struggle (Zettelkasten, longhand journaling, 'writing without LLM' rituals) become not preferences but ethical hygiene.

We are heading toward civilizational cognitive convergence. strength: low

Explicit premises:

LLM style has become 'humanity's writing style.'
Stylistic homogenization is empirically visible (LinkedIn, articles, books).
If we converge around rhetoric, we converge around thought.

Implicit premises:

Style and thought are tightly coupled — you cannot adopt a borrowed rhetoric without adopting (some of) its conceptual scaffolding.
Reading and writing in a homogenized style trains the reader's cognition into the same shape (a sort of Sapir-Whorf claim applied to LLM idiom).
The pre-LLM diversity of style reflected a pre-LLM diversity of thought (rather than, say, class/access markers).

Inferences:

If style trains thought, then deliberate stylistic resistance (writing 'against' the LLM idiom) is itself a cognitive intervention.
If true, this argument applies symmetrically to any era of stylistic dominance (e.g., AP style, MBA-speak) — but Franco treats LLM convergence as uniquely civilizational.

You should keep writing from a blank page and use LLMs only as a co-creator/sparring partner. strength: medium

Explicit premises:

Writing from a blank page is the mechanism of selfhood.
Outsourcing it loses the Self.
Co-creator use keeps the human 'in the driver's seat.'

Implicit premises:

The line between 'sparring partner' and 'ghostwriter' is in practice drawable by the user — a phenomenological boundary the essay does not specify.
Self-confidence and patience are individually cultivable goods rather than scaffolding the social environment is currently removing.

Inferences:

If the co-creator/ghostwriter line is hard to draw operationally, then well-intentioned users will gradually slide across it without noticing — which would mean Franco's prescription is unstable in practice.

Narratives

Franco describes his own return to writing from a blank page — staring at whiteness, the intimidation and refreshment, doing the essay 'using words chosen by me.'
→ Performative demonstration: the essay enacts its own thesis. The narrative implicitly supports the claim that the cognitive struggle is recoverable and valuable.
Catalogue of recognizable authors: de Botton's long sentences with paintings, Burkeman's quoted humour, Mollick's voice, Ritson's marketing giggle, Wodtke's opener, Horowitz quoting Notorious B.I.G., Cutler's speech-like prose, Pollan's curiosity, Sapolsky's audible cadence, Torres's confidence.
→ Implicitly argues that style = Self and that this is what is at stake. Functions as a 'before' picture for the loss Franco predicts.
Franco demonstrating the AI-slop techniques (anaphora, false dichotomy, concrete-to-abstract pivot) by deploying them on the page, then naming them.
→ Meta-rhetorical demonstration that the techniques are now legible — supports the empirical claim that the style is pervasive and recognizable.
The blank page reframed at the end as a mirror.
→ Metaphor doing the heavy lifting for the constructivist theory of selfhood — the page reveals/shapes the Self.
Franco's worry about his kids being unable to escape the blank-page struggle 'on their own.'
→ Ethical/intergenerational stakes — frames the threat as compounding over time.

Questions addressed

✅ What is the morally relevant difference between AI removing a technical barrier and AI doing the writing for you?
The former removes friction that doesn't form the Self; the latter removes the cognitive struggle that does.
✅ Why does it matter that everyone is starting to sound the same?
Because sounding the same will eventually mean thinking the same — civilizational cognitive convergence.
✅ What should we do about it?
Use LLMs as co-creator/sparring partner; preserve the practice of writing from a blank page.
⭕ Where exactly is the line between AI-as-editor/sparring partner (acceptable) and AI-as-author (corrosive)?
⭕ Why is writing — specifically — the privileged site of selfhood, rather than speech, drawing, building, deliberating with friends, or longhand journaling?
⭕ Is the claim that style-convergence entails thought-convergence empirical, conceptual, or a Sapir-Whorf-style hypothesis? What would falsify it?
⭕ Were pre-LLM writers actually forming richer selves through writing, or is Franco idealizing a baseline that was already homogenized by AP style, MBA-speak, SEO, etc.?

Gaps

incompleteFranco never distinguishes operationally between AI-as-editor and AI-as-author. He gestures at 'co-creator' and 'sparring partner' but offers no test for when LLM involvement has crossed from helping you think to thinking for you. Since the whole prescription hinges on this line, its absence is the essay's central structural weakness.
Read next: Andy Matuschak's notes on evergreen-note writing and 'understanding requires effortful engagement'; Sönke Ahrens, How to Take Smart Notes (the operational case for note-writing as thinking); literature on cognitive offloading (Risko & Gilbert) for empirical traction on which kinds of offloading degrade what kinds of cognition.
incompleteThe implicit theory of selfhood is constructivist (writing constitutes the Self) but Franco never engages alternatives: that selfhood is constituted in speech (Charles Taylor, dialogical self), in action (Aristotelian), in narrative across many modalities (Ricoeur), or that the 'Self' being expressed by his authorial gallery is a cultivated public persona rather than an inner essence.
Read next: Charles Taylor, Sources of the Self; Paul Ricoeur, Oneself as Another; Sönke Ahrens on writing-as-thinking; and contra: Cal Newport / classical-rhetoric arguments for speech and dialectic as primary.
incompleteThe argument from stylistic convergence to cognitive convergence is the most consequential claim in the essay and is unsupported. Franco asserts it ('we converge around rhetoric, what happens then?') but provides no mechanism. A Sapir-Whorf-style claim is doing the work uncited.
Read next: Linguistic relativity literature (Lera Boroditsky, Guy Deutscher); studies of LLM influence on user vocabulary and reasoning (e.g., recent work on 'epistemic homogenization' in LLM-assisted writing); Stolen Focus (Hari) on attention-environment shaping cognition.
incompleteThe argument privileges writing as the unique site of Self-formation without explaining why other effortful articulation practices (drawing, coding, deliberating aloud, longhand journaling, even careful prompting) cannot substitute. This matters because if other practices CAN substitute, the prescription changes from 'keep writing from a blank page' to 'preserve some effortful articulation practice.'
Read next: Embodied-cognition work (Andy Clark, Supersizing the Mind); the diary tradition (Sarah Manguso, Ongoingness) for non-public writing as Self-formation; sketchbook/visual-thinking traditions (Dan Roam).
closed byWriting vs. Thinking — That gap asks why writing is privileged as the unique site of effortful self-formation when other articulation practices — drawing, coding, deliberating aloud, prompting — might substitute for it. This document's "Code is another form of writing" section directly addresses the coding alternative, but by reframing rather than resolving it: it argues coding doesn't substitute for writing because it already is writing in the relevant cognitive sense (naming, tests, schemas all give implicit knowledge an inspectable, arguable form). This only partially closes the gap — it says nothing about the other listed alternatives (drawing, deliberating aloud, careful prompting), so the broader question of whether non-code, non-prose practices can substitute remains open.
incompleteNo engagement with the possibility that LLM-mediated reflection can be MORE rigorous than solo blank-page writing for some people — e.g., neurodivergent users for whom externalized dialogue scaffolds rather than replaces thought; or that writing from a blank page is itself a culturally specific Western Romantic ideal.
Read next: Literature on extended cognition; disability/neurodivergence and AI-assistance; the rhetoric tradition (which privileges dialogue and topoi over solitary blank-page composition).
illogicalSelf-undermining performance: Franco demonstrates LLM-style techniques 'on his own' (the anaphora caveat) to prove he can do them without an LLM. But this concedes that the techniques themselves are not the marker of AI authorship — humans use them too. If the stylistic tells are not in fact diagnostic, the empirical claim ('humanity's writing style') is weaker than asserted.
Read next: Studies on inter-rater reliability of human/AI text detection; the history of these rhetorical figures (which long predate LLMs).
uninformedThe essay treats LLM-induced stylistic homogenization as historically novel. Prior waves of stylistic dominance (AP style, MBA-speak, SEO writing, business-jargon, Strunk-and-White) produced their own convergences without producing the cognitive collapse Franco fears — suggesting style-to-thought transmission may be weaker than claimed.
Read next: History of plain-style movements (Orwell, 'Politics and the English Language'); style guide history; sociolinguistics of register convergence.

What we lose when AI writes for us

Structure

Key terms

Propositions

Arguments

Narratives

Questions addressed

Gaps

Related documents