Every generation of advertising technology arrives with the same panicked question attached to it: what happens to creativity now? Radio asked it, television asked it, digital asked it, and AI is only the latest version of a question the industry has been answering, correctly, for over a century.
The answer came from a man most people in advertising have never heard of. His name was John E. Powers. In the 1880s, working for department store owner John Wanamaker and later for Lord and Taylor, Powers became known as the world's first full-time copywriter. He earned that reputation not by writing beautifully, but by refusing to let style stand in for substance.
Powers built his entire reputation on a single, blunt principle: fine writing is offensive if it gets between the reader and the truth. Advertising in his era was dominated by flowery, exaggerated language, the kind of copy that promised miracle cures and described ordinary merchandise in the language of poetry, and Powers threw all of it out. His ads read like a clerk telling a customer exactly what was wrong with the product and why they should still buy it anyway.
Asked to move a surplus of damaged fabric that Wanamaker's could not sell, Powers wrote an ad that said the store had a lot of rotten gossamers it wanted to get rid of. The stock sold out by noon. That was the whole method: identify what is actually true about the offer, say it as simply as possible, and let the truth do the persuading. If the truth was not tellable, the job was to fix the product or the offer, not to dress up the language around it.
This was radical at the time because it treated advertising as a testable claim rather than an art form. Powers wrote copy, watched what sold, and adjusted. He was running split tests decades before anyone had a name for it.
Clever Is Fine. Hollow Is Not.
It would be a mistake to read Powers as an argument against clever, stylized, or genuinely artful advertising, and it is worth saying directly: some of the most effective campaigns in the industry's history are clever precisely because they are clever. Volkswagen's "Think Small" worked because it was a surprising, well-crafted idea, not a flat recitation of features. A great tagline can carry a brand for decades in a way no plain statement of benefit ever could.
The actual distinction Powers drew was narrower and more durable than "plain beats clever." It was: the cleverness has to sit on top of something real. "Think Small" worked because the car actually was small, at a time when American cars were not, and the ad turned an honest fact into a memorable idea rather than using style to obscure a weak one. Powers would have approved of that ad, not in spite of its wit but because the wit was in service of a true, specific claim rather than a substitute for one. What he refused to do was let beautiful language cover for a product or claim that could not stand on its own.
A Principle That Outlasts Every Technology Built to Test It
Most agency people today have never read a word Powers wrote, and that is exactly the point. His method did not survive as a style. It survived as an operating principle, and every advertising technology since has been, in one form or another, a faster way of proving him right.
Direct mail proved it with response rates. Television proved it with recall studies. Digital proved it with click-through data. AI is simply the newest and fastest instrument for running the same test Powers was running by hand in a Philadelphia department store in 1880. Every modern ad platform built around machine learning, whether it is optimizing a search campaign, a social campaign, or a generative tool producing dozens of headline variants overnight, tends to reward the version that is clearest about what it is offering and to whom. Clarity about the real claim produces a measurable response more reliably than ornamentation does, and that has been true across every major shift in advertising technology so far.
A few specific patterns are worth watching for, regardless of what platform an agency happens to be using this year.
A vague claim rarely survives real testing, however well it is written. A headline built entirely around style, with nothing specific underneath it, tends to lose to a headline built around a genuine, concrete claim, whether the judge is a 19th-century newspaper reader or a 21st-century optimization algorithm. That is not an argument against craft. It is an argument against using craft to disguise the absence of a real claim.
The offer gets fixed, not just the language. When a claim does not hold up under real testing, the best answer has rarely been better wording alone. More often it has been a better offer. Powers understood this with a stack of newspapers and a set of weekly sales figures. Every platform since has understood it with better and faster data, and the underlying instinct has held up for well over a century.
Volume replaced instinct with evidence, and that shift did not start with AI. It started the moment advertising became measurable at all. What has changed is only the speed. Powers could test one headline against reality at a time, in one newspaper, over one selling season. Modern platforms can test dozens of variants against real audience response in days. The scale keeps getting faster. The underlying belief, that the market is a meaningful judge of what copy works and that no amount of polish fully overrides a weak underlying claim, has held up remarkably well.
What This Means for an Agency Right Now
The specific technology matters less than the discipline it demands, and AI is currently a sharp test of that discipline.
The warning is real: AI tools make it easier than ever to produce polished, well-written copy that says very little. Generative tools default toward smooth, agreeable, adjective-rich language because that is what most training data rewards stylistically. Left unchecked, AI-assisted copywriting can drift toward exactly the kind of ornamentation Powers spent his career fighting, using style to paper over a claim nobody has actually sharpened. An agency that lets any tool write unsupervised, today's or whatever replaces it in five years, risks producing more copy, not better copy.
The opportunity is just as real. A great idea built around a real, specific truth about the offer is still the strongest thing an agency can produce, and AI tools can help sharpen that idea rather than replace it, if an agency uses them that way. Feed a tool a real, specific, defensible claim about a client's offer, and ask it to help express that claim as sharply and memorably as possible, whether that means plain or genuinely clever. Then use whatever testing data is available the way Powers used his own sales figures: as a meaningful check on whether the idea is actually landing, not the final word on whether craft has a place in the work.
What This Means for a B2B Agency, Too
None of this is only a consumer-marketing story. It applies just as directly, arguably more directly, to B2B. A B2B buyer reading a vague, adjective-heavy pitch about transformative solutions and unlocking synergy reacts to it much the way a Wanamaker's customer reacted to flowery Victorian ad copy: with suspicion. B2B sales cycles are longer and the testing looks different, an email subject line, a LinkedIn ad, a gated white paper headline, rather than a same-day retail sellout, but the underlying test is similar. Does the copy say something specific and true about the offer, dressed up as cleverly as the agency can manage, or does it hide behind polish because the offer itself is not sharp enough yet? A vague claim tends to underperform whether it is chasing a retail sale in an afternoon or a signed contract in six months.
The agencies that get the most out of any new advertising technology, this one included, will not be the ones with the most sophisticated tools. They will be the ones that treat every piece of output the same way Powers treated his own copy: as a claim worth testing against reality, built on something true, and expressed as well as the agency's creative talent can manage. That was true in a department store in 1880. It will be true whatever comes after AI, too.
