What 18 months running my marketing in Claude actually changed, and what nobody's saying out loud.
I got an email last week from a brand I used to like. The subject line was personalized. The opening referenced a recent piece of content I'd shared. The body was structurally perfect, with a clear hook, a tight value proposition, and a single low-friction CTA. It read like it had been generated by Claude. It probably had been. And I closed it without finishing.
The problem wasn't that the email was AI-written. The problem was that I could tell. Not because of the writing. The writing was fine. Because of the shape. The same hook structure I'd seen in fourteen other emails that week. The same "personalization, value, CTA" rhythm. The same restraint applied to the same length. It had been optimized into a category instead of into a voice.
This is what AI did to marketing. Not "transformed it." Not "unlocked new possibilities." It made the slop easier to produce, and now we're all swimming in it.
I run my marketing in Claude. I built a paid prompt library aimed at marketers using Claude. I should be the last person making the case against AI in marketing. But this is the case I want to make. The marketers who will win the next five years are the ones using AI to ship LESS, not more.
The pitch every AI marketing tool makes is the same. Save time, scale output, free yourself up for "the strategic work." It's a comforting story because it lets us avoid the harder question, which is whether we were doing strategy before, or whether we were doing assembly.
Most marketing teams I know took the time AI saved them and used it to make more stuff. Not better stuff. Just more. Same content cadence at 3× volume. Same email program with 4× more sends. Same blog with 10× more posts. The metric they're measuring went up. The metric the audience cares about, which is whether the content was worth their attention, went down.
The pitch promised AI would free us up for strategy. The reality is most marketing teams used AI to outsource the part of the job that was already the easiest, and kept neglecting the part that was always hardest. Deciding what NOT to ship.
The work AI made easier isn't the work that mattered.
When I started rebuilding my marketing around Claude two years ago, I expected the big change to be drafting time. It was. A 1,500-word article that used to take me three hours of typing now takes thirty minutes of editing. Reporting that took two days takes ten minutes. Brand voice analysis I would have charged a client $4,000 for runs in Claude in twenty minutes against the same source material.
Here's the part nobody talks about. The first draft used to be 80% of the work. Now it's 10%. And the work that used to be the other 20%. Judgment, taste, defending decisions, knowing when a campaign should be killed instead of polished. That work is still there. Still hard. Still slow. Still the actual job.
Most marketing teams haven't reorganized around this. They still measure the easy thing. Output volume. Number of pieces shipped. Posts per quarter. Emails sent. They optimized their own measurement system around the part of the work AI made easy, then they wondered why their results didn't get better.
The work that compounds didn't get easier. The work that compounds is still: who do we refuse to talk to, what do we refuse to say, which campaign do we kill because it would dilute the brand, what do we choose NOT to publish this quarter. Those decisions need a human with taste and the authority to lose an argument for them. Claude can help draft the deck. Claude cannot make the call.
Here is the part of this conversation no marketing vendor will run a webinar on.
Every piece of AI-generated content shipped at scale has an externality cost. Your audience pays it. They pay it in attention they used to give your industry, that they're now withdrawing because too much of what reaches their inbox or feed feels generated. They pay it in trust, because they've started discounting messages from your category by default. They pay it in time spent learning to filter, getting better at smelling AI faster than the supply side is improving the AI.
This is the slop tax. The audience didn't agree to pay it. We just started shipping more, and they responded by trusting us less.
The teams that win the next five years will figure out that the marketing that breaks through is the marketing that's visibly not slop. Visibly considered. Visibly specific. Visibly hand-edited. Visibly willing to say one true thing and stop. Not because AI is bad, but because AI made more the default, and the audience is now starved for fewer.
Restraint is a strategy. It used to be a tasteful choice. Now it's a survival mechanism.
The lever isn't speed. It's judgment. The marketers who'll thrive in the next five years aren't the ones who type fastest into Claude. They're the ones who can look at a Claude output, see that it's technically correct, and still say "no, we're not shipping this. It's not on-brand for us, it's not honest, it's not the move." That's a hard skill. AI doesn't teach it. AI atrophies it if you let it.
The marketers who'll get quietly fired in the next five years are the ones whose only contribution is "I can run prompts faster." That's a commodity skill now. It was a commodity skill twelve months ago. It will be a commodity skill that two-year-olds can do in eighteen months. Prompts are a commodity. Taste isn't.
The honest summary of what AI changed about marketing is this. It raised the floor and lowered the ceiling. It raised the floor because the worst marketing output is now technically competent, with clean grammar, coherent structure, on-brief. It lowered the ceiling because the best output is now harder to distinguish from the floor. The teams that win will be the ones who realize the new game is to raise the ceiling back up, and that doing so will require slowing down, shipping less, and getting opinionated.
Audit your last month of marketing output. Highlight every piece that's just AI slop with your name on it. Cut the bottom half. Don't replace it. The space you create is where the better work will go.
Stop measuring "we shipped X assets." Start measuring "we made N decisions a competitor wouldn't have made." The first metric pretends activity is value. The second is closer to the truth.
Find the one campaign on your roadmap that you secretly know shouldn't ship. Kill it. Write a one-paragraph internal memo explaining why. Send it to your CEO. That memo is more strategic work than the campaign would have been.
The most overlooked way to differentiate from your competition this year isn't a new tool. It's restraint.
Zara
This essay named the tax. The free Slop Sniffer lets you measure it. Paste any copy and get a 0 to 100 score with inline highlights of AI tell-words, hedging filler, and horoscope phrases. Runs in your browser. No email gate.
Open the Slop Sniffer →This essay sits inside the Marketing Prompt Hub: 30 tested Claude prompts across 8 marketing disciplines, 6 playbooks, 5 agent blueprints, full reporting kit. Pay once, own the files. No email gate.
See the hub →