Write Like a Mad Man: Content That Actually Converts

Good writing is scarcer, not more abundant, in the AI era

The flood of AI-generated content has made good writing harder to find, not easier. Vague briefs get exactly what they ask for — "a generic input typically results in a fancy arrangement of average stuff," as MarTech puts it. Kantar warns that brands failing to offer "unique angles that set them apart in a sea of AI-generated sameness" get "optimized out."

Call it the input premium: when everyone wields the same model, the only durable edge is the quality of what feeds it — the brief, the persona, the human judgment about benefits and feeling. That edge compounds for the team disciplined enough to build it.

Benefits, not features: run the "so what?" test

The oldest rule in direct-response copy, framed as a test that's brutally simple to apply: if you can read a line of your own copy and ask "so what?", that's bad copy. A Toyota Tundra tows 12,000 lb — so what, so can a Ford. The benefit-led version: it "fits all heavy-duty lifestyles, whether you're hauling materials at a work site or getting the ATVs to the trails." Keep the specs, but lead with the benefit. In one widely cited Leadpages split test, rewriting a single headline from feature-flat to benefit-led lifted opt-in conversions 307% — proof the promise, not the spec sheet, goes first.

One mechanical point rides alongside it: always use active voice. Nielsen Norman Group's research on legibility is explicit that plain, direct language lowers reading effort and raises comprehension, and a 2024 review of nine studies found passive constructions consistently slowed processing versus active ones. Passive voice forces the brain to hold a sentence in memory until it learns who acted — a small tax that, repeated down a page, erodes attention. It's the rare optimization that costs nothing: rewrite the verbs.

Priming: how you start the emotional engine

This is the playbook's most original contribution: priming is how you start the emotional engine, and every word choice primes a feeling. A dentist advertising "painless dentistry" primes the very pain it's trying to deny. The fix is one word — "gentle dentistry": same promise, no negative anchor. Car dealers advertise a "hassle-free experience," reminding shoppers of exactly the hassle they don't enjoy; the correction is "a streamlined experience."

The language you're using is causing feelings in people. Make sure that feeling is positive.

The behavioral science backs this hard. The canonical attribute-framing study, a 1988 Journal of Consumer Research paper, found consumers rated ground beef labeled "75% lean" as tasting and looking better than the identical beef labeled "25% fat" — same product, same number, opposite frame, and the negative word measurably degraded perception. Eye-tracking research on message framing adds an honest caveat: negatively framed messages can grab more attention, but the same research finds purchase preference is higher under positive framing. Once you have the eyes, the threat word works against the yes.

Voice of customer: speak their language, not yours

If you're trying to persuade people, use the language they use every day, the language in which they think. Mine it from reviews, competitors' reviews, Reddit, Quora and People Also Ask. The payoff is twofold: algorithmically, it's a closer pattern match; humanly, it sounds more authentic — and authenticity is what earns trust in the age of AI.

This is the most validated technique in the whole playbook. Copyhackers, the shop that coined the term "conversion copywriter," tested a rehab center's control headline "Your addiction ends here" against "If you think you need rehab, you do" — a line lifted almost verbatim from how prospects described themselves in reviews. The customer-language version generated over 400% more clicks on the main CTA and over 20% more form submissions on the next page. The winning headline wasn't cleverer; it was theirs.

The content brief is where this becomes repeatable rather than accidental. A real brief covers the basics (client, offer, USPs, and whether the piece risks cannibalizing a previous one), background and context, past campaign performance, research and interviews, and — critically — the who: an incognito-mode audit of what's already ranking, so a text blog doesn't get built for a query Google is already answering with video. With generative-AI adoption at 87% of marketers (Salesforce, up from 51% in 2024), the model is table stakes; the brief is the only proprietary input left.

Write for scanners, choose depth over frequency

Website visitors scan, they don't read — so the execution question is whether a paragraph could be a chart, a bullet list, or a pull-quote instead. Nielsen Norman Group's research found 79% of users always scan a new page while only 16% read word-by-word, and visitors absorb at most 20–28% of the words on an average page. A 2026 Microsoft Research eye-tracking study found that when an AI Overview sits atop a results page, attention concentrates there first and users scroll even less into what's below — scanning got faster and shallower, not slower.

The same discipline decides how AI answers cite you. The Princeton-led "GEO: Generative Engine Optimization" study (KDD 2024) found citing external sources lifted a source's visibility in AI answers by roughly 40% on one metric, and adding statistics lifted it by around 37% on another — the biggest levers are exactly the ones this playbook already asks for: cite sources, back claims with numbers, in extractable chunks.

Depth matters more than volume, and the industry learned this the hard way. SE Ranking's 16-month experiment publishing 2,000 one-click AI-written articles across 20 brand-new domains looked promising early — 71% indexed within 36 days — then the floor fell out: by two and a half months the whole portfolio had drawn just 782 clicks, under one click per article, and visibility stayed flat a year on. Volume without depth didn't convert into traffic.

The AI era hasn't made writing obsolete. It has made good writing scarce, and therefore valuable.

Six things any team can start this week:

• Run the "so what?" test on every page. If a line states a feature with no answerable benefit, rewrite it — lead with the promise, then the specs.
• Audit copy for priming, one word at a time. Strip negations and threat words ("painless," "hassle-free") for their positive form ("gentle," "streamlined").
• Rewrite for active voice as a free optimization — no design change, no spend, just sharper verbs.
• Make voice-of-customer research mandatory. Mine reviews, competitor reviews, Reddit, Quora and People Also Ask before drafting a single line.
• Treat the brief as the product, not the chore — it's the input that determines both human and AI output quality.
• Write for scanners and choose depth over frequency. Bullets, charts, cited statistics and answer-first structure beat a high-volume calendar Google now devalues as scaled content.

The team that still knows how to make a promise, prime a feeling and deliver on the page will out-convert the ones racing to publish more.