In Defense of AI Slop for Go-to-Market

Everyone is writing about AI slop. Almost nobody is writing about the kind of writing sales teams do.

So sales leaders are getting bad advice on AI in their outreach. The voices are credible, the anxiety is real, and the advice is still wrong.

This is a defense of the writing they're calling slop. It isn't a claim that quality stopped mattering. Quality still sorts the good operators from the bad. The slop panic just misreads where that sorting happens.

Ethan Mollick wrote on X that AI writing keeps getting harder to detect. His longer Substack piece, "Choosing to Stay Human," describes AI output as "meaning-shaped attention vampires." Posts that take effort to read and give nothing back.

Cal Newport, in "Easy Is Overrated," cited new data from an Organization Science AI Task Force. Heavy-AI papers get desk-rejected at nearly 70%, while low-AI papers get rejected at 44%.

Rebecca Winthrop covered Adam Green's study of 370,000 college essays in the New York Times. AI-assisted essays produced up to eight times fewer new ideas than human ones.

The New Yorker reported a statistic hard to forget in Jill Lepore's prehistory of AI slop. By fall 2024, machines wrote roughly half of all English-language articles on the internet.

LinkedIn rebuilt its algorithm this spring around a model called 360Brew. Research from AuthoredUp shows median impressions dropped 47% on posts that read as AI-generated. A cottage industry of "humanize your AI" tools sprung up around the penalty. Humanizer, one of them, sits on GitHub with more than 20,000 stars.

But almost all of this writing is about the same kind of writing. Op-eds, essays, academic papers, voice-driven thought leadership, LinkedIn posts where the writer's voice is the product.

Almost none of it is about teams sending prospecting emails or producing marketing content at scale. Operational writing where volume and personalization matter more than craft per piece.

That's the gap this piece fills. What is slop, really, and where does it apply?

Slop isn't an aesthetic property of AI writing. It's a specific reader experience.

That experience needs two things firing at the same time: AI writing signals (em dashes, familiar cadence, meaningless similes, "not X, but Y" constructions) and empty ideas.

Either alone is tolerable. Beautiful prose with sharp ideas gets a pass. Rough prose with sharp ideas gets a pass. AI-flavored writing with sharp ideas gets a pass.

Slop is the specific feeling of investing attention and getting nothing back. The reader feels cheated, looks for something to blame, and the AI signals are right there, easy to point at.

Mollick captures part of this in a footnote. Humans assign meaning to difficult writing even when none is there, and AI signals interrupt that generous reading.

Winthrop's finding sharpens the picture. Human judges rated AI-assisted essays as more creative than human ones, until they looked at the underlying ideas. The polish fools readers on first pass. The emptiness shows up on closer inspection.

LinkedIn's new algorithm operationalizes the same reaction at platform scale. The 47% reach penalty is a measure of reader behavior. The system tracks dwell time, click bounces, saves, and shares. Generic AI content fails because readers leave fast.

The signals are the smoke. The empty ideas are the fire. Most of the current discourse fights the smoke.

Slop = empty ideas + AI writing signals.

Mollick is right that we used word counts and prose effort as proxies for thinking. He's wrong that the proxy was ever reliable.

Dostoyevsky dictated his late novels to Anna Grigorievna. PhD theses are partly synthesis of work other people did. Executives have always used ghostwriters and chiefs of staff. Lawyers dictate to associates and paralegals.

The writer isn't always the writer. Dostoyevsky didn't type every word of his late novels. The senior partner doesn't draft the brief. The CEO doesn't write the shareholder letter from scratch.

The person who owns the ideas is the writer. The person pressing the keys is the typist. Those two roles have been distinct for centuries.

Length never equaled quality. Polish never equaled depth. AI didn't break the proxy. AI made the proxy's weakness visible at scale, and it gave every writer their own scribe.

You can outsource your thinking. You can't outsource your understanding. That line is the whole argument, and the rest of this piece is what it means for the work your team does every day.

If AI is a scribe, then AI writes what it's told. A bad prompt yields bad output. A thoughtful prompt yields thoughtful output. The primacy of ideas hasn't changed.

What changed is the cost of producing polished-sounding volume. That cost has collapsed, and the burden of evaluating ideas now sits with the reader.

Newport's Organization Science finding makes the dynamic concrete. Researchers used to be gatekept by the cost of writing a paper. That cost dropped, and the reviewer pool absorbs the new volume.

Every easier workflow for the producer creates a harder experience for someone downstream. The cost doesn't disappear. It moves.

The current discourse conflates two very different kinds of writing.

Category one is writing where the act of writing produces the value. Essays, novels, op-eds, academic papers, original analysis, voice-driven thought leadership. The thinking happens in the typing. Craft develops through practice. The writer is the product.

This is what Mollick, Newport, Winthrop, and most of the X discourse are writing about, and they're mostly right about it.

Category two is writing where the act of writing is a bottleneck. Prospecting outreach, sales sequences, marketing emails, press releases, ad copy, content at scale. The thinking happens upstream in strategy, segmentation, and message design. Volume and personalization matter more than craft per piece. The writer is a node in a system.

The platforms themselves are now drawing the same line. LinkedIn's algorithm measures whether readers stay: dwell time, click bounces, saves, shares. Generic AI content tanks because readers leave fast. The 47% reach penalty is a behavioral signal operating at platform scale.

Prospecting emails don't have a depth score. They have a reply rate. Different distribution model, different reader relationship, different rules. The same content that loses 47% of LinkedIn reach can book a meeting if the targeting and message logic are sound.

The slop frame applies cleanly to category one. It mostly doesn't apply to category two.

The baseline matters. Most prospecting emails were bad before AI. Generic, unresearched, ignored. We called them spam, and we were right. In category two, AI raises a floor that was already on the ground.

The mistake is generalizing from category one to all writing. That's where the bad advice for sales leaders is coming from.

Within either kind of writing, there are three ways to deploy AI. Only one of them produces slop.

Mode A: Human owns the ideas. AI converts to prose. The writer has the argument, the structure, the examples. AI is the typist.

Mode B: Rough prompt in. Minimal edit. Ship it. The writer has a topic. AI generates the rest. The writer doesn't push back on the output and may not be able to defend it.

Mode C: Iterative thinking partnership. The writer brings ideas. AI generates a draft. The writer pushes back. AI surfaces angles the writer didn't see. The writer decides what stays. Every claim can be defended because the writer made every decision.

Newport's data captures what Mode B looks like at scale. The desk-reject rate is what happens when Mode B scales across thousands of writers.

The Anthropic programmer study Mollick cites supports the same distinction. Programmers who let AI do the work couldn't answer questions about what they had done. Programmers who asked AI to explain its work, or used AI for only part of the task, kept their understanding intact. Mode B versus Modes A and C, in someone else's data.

Even Winthrop concedes the distinction at the end of her essay. Workers with deep craft knowledge can use AI to streamline administrative tasks so they can focus on where originality lives. That's Mode A and Mode C, conceded by a critic.

The two frameworks stack. Categories tell you what kind of writing you're doing. Modes tell you how you're using AI on it. The combinations matter:

We blame the tool. The problem sits in the hands holding it.

Mollick, Newport, and Winthrop are right that writing helps you think. There's a real learning cost when AI does the typing.

For category-two work, the cost is real. The question is where the learning should happen.

For a BDR, the relevant learning lives below the prose, in the strategy underneath the message. Why this sequence cadence. Why this subject line construction. How to match a case study to a stated pain point. Why A/B testing matters because even informed guesses fail.

For a junior marketer producing content at scale, the same logic holds. Why this campaign theme. Why this audience segmentation. Why this distribution sequence. Why this hook structure.

A rep or marketer who understands those things is doing the same cognitive work the critics care about, on the strategy layer rather than the typing layer.

The real risk lives one level above the tool. Leaders deploying AI without investing in the strategic understanding their teams need. Without that investment, AI scales whatever the team doesn't understand.

AI lets a BDR write above their weight on the page. The email signals pain-point sophistication and business acumen the BDR may not have actually internalized.

The prospect reads the email and thinks: this person gets me. They take the call expecting the writer of that email.

When the conversation doesn't deliver the depth the email implied, the gap collapses the deal. The prospect feels cheated. It's the two-factor reaction again, surfaced live on the call: AI signals in the email, plus empty understanding in the rep.

This is Newport's externality in B2B form. The seller's easier workflow creates a harder experience for the prospect, the BDR who has to defend the email, and the sales manager who inherits the bad-fit pipeline.

The cost moved. The cost didn't disappear.

A BDR who can defend the strategy on the call closes the loop the email opened. The earlier investment in understanding why this approach beats the alternatives pays off here, on the phone, in the moment that decides the deal.

The BDR outsourced the writing. The discovery call is where the prospect finds out whether they also outsourced the understanding.

Mollick prescribes individual intentionality. Newport prescribes individual effort. Winthrop prescribes individual practice. All three prescriptions work for category-one writing.

For sales and marketing teams producing category-two writing at scale, the answer is process design and strategic enablement. Individual willpower doesn't scale across a team.

Deploy AI heavily on category-two work. The baseline was low to begin with, and AI raises the floor.

Invest the saved time in strategic understanding. Strategic understanding is what lets your team defend whatever the system produces in their name. Why this sequence. Why this case study match. Why this subject line. Why this segmentation. Playbooks, message houses, manager coaching, and enablement build that understanding.

Your job is building the depth of understanding that lets your reps stand behind whatever shows up in the prospect's inbox under their name. AI detection is the wrong target.

Working human process must come before AI deployment. AI captures and scales what the team already does well. Deployed onto broken or undocumented strategy, AI produces broken outputs at scale.

Slop isn't what AI produces. Slop is what happens when the human in the loop can't back up what got produced.

The slop frame fits category-one writing, where the critics are right and the stakes are real.

The slop frame breaks down in category-two writing, where it produces bad advice that leaves sales teams competing with one hand tied.

The signals will keep getting harder to detect, as Mollick predicts. The depth gap will keep getting harder to hide. Newport's reviewers are discovering it now. Your prospects are about to.

The work that matters is upstream of the prompt. Always was.

There's a lot of criticism about using AI for writing, and some of it is valid. But not all writing is the same, and the rules need to reflect that.

What people are really objecting to isn't AI itself. It's writers who write a bad prompt, let the AI run with it, and put out whatever comes back. That's a real problem. But it gets lumped in with every other form of AI-assisted writing, and that's not a fair read.

At the end of the day, the writer is still responsible for what they put out. The tool doesn't change that.