The Efficiency Pivot: Why Human Copy Wins Back Ground

One of the searches that brings people to this page is “is the new Colgate advert AI generated.” It and several variations keep appearing in the page’s own search data. Nobody used to google whether a toothpaste ad was made by a person. People do it today because they have learned to look for the tells.

That search is the most useful data point in the debate about AI and copywriting. It shows that audiences have started auditing ads for machine authorship. And it points to the argument this article makes: when every brand can generate competent copy at almost no cost, competent stops being enough, and the work that sounds like nobody else becomes the scarce thing worth paying for.

AI production works. That is the problem.

Start with what is true about AI in advertising, because the case for human copy is not that AI is bad.

Colgate-Palmolive ran a generative AI pilot, described in Think with Google, in which its team produced two consumer-ready video ads four to six times faster than usual, at a significantly lower cost per concept, with the same or better consideration lift than business-as-usual creative. Its Hello brand uses AI to find consumer insights a close-up team would miss, the approach covered in how Hello uses AI to find what humans miss.

The same company also shows the other side. In April 2026, a Colgate promo for a new tropical toothpaste was picked apart online for garbled packaging text that viewers attributed to AI, as Creative Bloq reported, with commenters telling the brand to hire new designers. Earlier in the year, people on social media were already asking whether new Colgate ads were AI-made.

Put those together and the pattern is clear. AI makes production fast and cheap, and when it is used well it performs. But fast and cheap is available to every competitor at once, and the cost of being caught looking machine-made is paid in trust.

It also matters how the pilot was run. It was a structured test, with a team setting the brief, reviewing the output and measuring results against business-as-usual creative. Whatever happened behind the tropical promo, the reaction to it shows what viewers do when review looks thin: they notice, they screenshot it, and they say so in public.

When everyone generates, everyone converges

By design, a language model writes the most likely next words for the prompt it is given. Give a hundred brands in the same category similar briefs, and they get similar sentences: the same benefits, the same structure, the same safe adjectives, all fluent and all forgettable.

That is the convergence problem. AI raises the floor of copywriting dramatically. A small business can produce ads, emails and landing pages that are clean and grammatical. But it raises the floor for everyone at the same time, and an ad that sounds like the category is an ad people skim past and fail to attribute to anyone in particular.

The tells are familiar to anyone who reads a lot of marketing: the opening question, the list of three, the promise to “discover” or “transform,” the closing line that restates the headline. Ad platforms add to the effect. Meta and Google both build generative tools into their ad products that write headlines and produce image variations from an advertiser’s existing assets, so even the variations meant to make an account distinctive come from the same kind of model as everyone else’s.

WPP’s Open Pro platform, which offers small businesses AI-planned and AI-generated campaigns without an agency, is a sign of where this goes, as discussed in what WPP’s AI platform means for small businesses and agencies. When competent campaigns are available to everyone on demand, competence is no longer an advantage. It is the entry fee.

Audiences are reading for the machine

The second force is trust. Two pieces of research point the same way.

The Nuremberg Institute for Market Decisions showed people identical ads, some labeled as AI-generated and some as human-made. Those who saw the AI label rated the ads as less natural and less useful, and were less inclined to click or engage, even though the content was the same.

A study led by Washington State University, published in the Journal of Hospitality Marketing and Management, ran six experiments and found that mentioning artificial intelligence in product descriptions reduced purchase intentions, because it lowered emotional trust. The effect was stronger for higher-risk purchases.

Neither study says AI-assisted work fails. They say that when people believe a machine made something, they trust it less. The searches about Colgate show people actively trying to find out. Copy that reads as generic carries a second cost: it can read as generated.

That changes what a weak ad costs. A generic ad used to be a wasted impression. An ad that reads as generated can lower trust in the brand behind it, and the WSU findings suggest the damage is greatest where the purchase carries risk: financial services, healthcare, B2B contracts, anything expensive. Those are the categories where trust was hardest to build in the first place.

What human copy actually contributes

Human copy wins back ground for reasons that have nothing to do with romance about craft. It does specific things a model drawing on the average cannot.

It brings observed detail. The objection a customer raised on a sales call last week, the phrase people actually use for their problem, the small true thing about how the product is used. Specifics are what make copy feel written for someone. A model can write “easy to use.” A person who sat in on onboarding calls can write about the third screen where every new customer used to get stuck, and why that no longer happens.

It takes a position. A model trained to be agreeable avoids claims someone could disagree with. Persuasion often needs exactly those claims: what this product is not for, what the category gets wrong, why the obvious choice is a mistake. A bank that says its app is not built for day traders has told the right customers something useful, and said something a competitor cannot copy without changing its product.

It knows when to break the pattern. Distinctive advertising works by not looking like the category. A writer can deliberately go against the convention; a model produces it by default.

And it carries accountability. A person who wrote the claim can explain it, defend it and fix it. That matters more as audiences grow suspicious.

The division of labor that works

None of this means writing everything by hand. The efficient split is the one in the commercial reality of positionless marketing: specialists do the zero-to-one work, and AI helps with the one-to-many work.

Use AI for research, synthesis, first drafts of routine material, variations of a proven idea, resizing and adapting for placements, and testing at volume. Use people for the insight, the core argument, the claim, the voice and the final edit. Then check everything that goes out for the tells that cost trust: garbled text, impossible details, testimonials no one actually gave.

In practice, a campaign built this way runs in a clear sequence:

1. A person writes the brief: who the customer is, what they already believe and the one argument the campaign will make.
2. AI gathers and summarizes the raw material: reviews, support tickets, sales call notes and competitor ads.
3. A writer drafts the core message and two or three genuinely different directions.
4. AI produces variations of each direction and adapts them to every format and placement.
5. A person edits the final set and checks every claim before launch.
6. Results decide which direction gets more budget, and the cycle starts again with what was learned.

That split also changes the economics of advertising. Ad platforms reward creative that earns attention and response, so distinctive work tends to cost less to run than work that looks like everything else, one of the cost levers in how to lower ad costs. Running a creative program built on that division, with people setting the argument and AI scaling it, is the core of performance marketing that treats creative as the main lever.

What to pay for when words are cheap

If producing words costs almost nothing, paying for volume makes no sense, whether you pay a team, a freelancer or an agency. Pay for judgment instead: understanding the customer, choosing the argument, writing the line only your brand could run, and knowing what to cut. That is the same principle behind deciding whether to hire an agency or build in-house, and it is what content marketing built around arguments that sell should mean.

Measure copy the same way. Fewer metrics about how much was shipped, more about what it did: conversion rate, cost per result, growth in people searching for your brand by name.

The same thinking gives you better questions to ask anyone who writes for you. Who writes the core message, and what do they read about your customers before they start? Which parts of production use AI, and who reviews the output? How is copy judged after launch, and what happens to a direction that underperforms? A partner who can answer those plainly is selling judgment. One who answers with the number of assets per month is selling volume.

The efficiency pivot moves the money from producing more to standing out, with AI still doing much of the producing. When everyone can make the average, the average stops working, and the brands that sound like themselves get remembered.

Frequently Asked Questions

Is the new Colgate advert AI generated?

Colgate-Palmolive has used generative AI in its advertising. Think with Google reported a pilot in which its team produced video ads, including one for Hill’s Pet Nutrition, with Google’s Veo tool. In April 2026, a Colgate promo for a tropical toothpaste was criticized online for garbled packaging text that viewers attributed to AI. Whether any specific ad used AI is for Colgate to confirm, but the brand clearly uses AI in parts of its production.

Does Google penalize AI-generated content?

Not for being AI-generated. Google’s guidance on AI content says it rewards helpful, high-quality content however it is produced. Its spam policies target content made mainly to manipulate rankings, including mass-produced pages with little value, whether written by people or machines. For advertising programs such as AdSense, the same principle applies: original, useful pages are fine, and low-value auto-generated sites are not.

Will AI replace copywriters?

AI is replacing routine copy production: simple product descriptions, variations, first drafts and adaptations. It is much less able to replace the work of finding the insight, choosing the argument and writing in a voice that is recognizably one brand’s. As AI makes competent copy universal, that second kind of work becomes more valuable, so the role shifts from producing volume to directing what gets said.

Should brands tell customers when ads use AI?

Research suggests that labeling content as AI-generated can lower how natural and trustworthy people find it, even when the content is identical. That is not a reason to hide AI use, especially where platform rules or local regulations require disclosure, and never a reason to fabricate testimonials or results. It is a reason to keep people responsible for the claims, the voice and the final check on anything that goes out.

How should a marketing team use AI for copy?

Use AI for research, summarizing customer feedback, first drafts of routine material, variations of proven ideas and adapting copy to different placements. Keep people responsible for the core insight, the main argument, brand voice and final edits. Review everything for signs that cost trust, such as garbled text or invented details, and measure copy by results like conversion rate and cost per result rather than output volume.

What is the efficiency pivot in marketing?

It is the shift in where marketing budgets create an advantage once AI makes content production cheap. When every competitor can generate competent copy and creative at low cost, spending on more output stops differentiating a brand. The advantage moves to the work AI averages out: customer insight, a clear position, a distinctive voice and careful editing. AI keeps doing much of the production, while people decide what is worth saying.