AI Content Quality: The Argument Is Over

In February, Coca-Cola described a shift that sounded like a pricing decision and was really a production one.

Coverage of the company's leadership discussions in February 2026 described a new phase built on persuasion rather than pricing power, with digital platforms, AI, and in-store execution taking a larger role in building demand. The reporting originated with Mi-3 and was picked up across the marketing trade press.

Read it as an operations memo instead and it changes shape. A price increase is one decision, made once, by a handful of people in a room. Persuasion is thousands of assets, cut for every placement, localized for every market, refreshed before fatigue sets in, and tested against each other until something wins. Cheap to decide, expensive to be wrong about. Then the opposite, on both counts.

So the pivot only works if the production side can carry it. Two years ago it could not. The output was fine for a test and embarrassing on a billboard, and every senior creative in the industry knew it. I spent those years running commerce for Publicis across China and North Asia, watching brand teams greenlight AI pilots that died in legal review or died in the first round of client comments.

That changed, and it changed faster than the conversation about it did. We covered the mechanics of how in What AIGC Production Actually Is. This piece is about the consequence.

What shifted between 2024 and 2026

Two of those rows deserve their numbers spelled out.

The Duke University and Deloitte CMO Survey reports that AI and machine learning now power 24.2% of all marketing activities, up from 13.1% in 2024, with marketing leaders projecting 55.9% within three years.

Salesforce's State of Marketing 2026 puts generative AI use at 87% of marketers in at least one workflow, up from 51% in 2024.

Adoption curves are the least interesting part of this. Plenty of things get adopted and stay mediocre. A better question is whether the work got better, and there the evidence is unusually direct.

Cannes moved the line in June

Cannes Lions has spent seventy years deciding what the advertising business considers excellent. This June it changed the definition before a single entry was judged.

The festival introduced AI Craft subcategories across Design, Digital Craft, Film Craft, Industry Craft, and Creative Data. It also created the Creative Brand Lion, a new award for brands that have built the internal systems and culture making repeatable creative excellence possible. Google took the AI Craft Grand Prix for Project Genie.

Neither change is decorative. A craft category is an admission that there is craft to judge, and no industry concedes that about something it still thinks of as a toy. And the Creative Brand Lion points at the more useful insight: when generating and testing options gets cheap, execution stops being the constraint. What replaces it is judgment, and the discipline to kill the ideas that do not deserve to survive contact with a market.

Worth noting what won the Film Grand Prix in the same week. Mother London's spots for an AI company, made by people who understood tone, timing, and a competitor's weak spot. The tools were in the room. The decisions were not made by them.

For anyone briefing production work, that is the practical read. The craft question is settled enough to be judged on a stage in Cannes now. The question your agency roster now has to answer is who is making the calls, and whether they are any good at it.

Quality, measured in throughput

Brands furthest along stopped talking about pilots a while ago. They are reporting throughput. The numbers below are the ones we get asked to explain most often in procurement conversations, so read them carefully rather than quickly.

Unilever figures as reported for its digital twin content operation. Adidas figure as reported for personalized email creative. Nestlé figure as reported for its in-house AI-enabled content service. Smartly figures as reported for its AI Studio product.

The first Unilever line is the one to sit with. That jump is not a productivity gain, it is a different kind of company. And it showed up in audience numbers, not only in the cost line: a Dove launch running on that system drew 3.5 billion social impressions and pulled a 52% new customer base.

There is a cleaner quality proxy than any of these, though, because it is measured in money rather than volume.

DigitalApplied's Q1 2026 creative benchmark puts AI creative at full return on ad spend parity with human creative for products under $100 average order value. That threshold sat at $25 twelve months earlier.

A parity line that moves 4x in a year says something cost alone cannot. Below a certain price point, buyers stopped being able to tell the difference, and their spending shows it.

Where it still breaks, plainly

Now the part the case studies leave out.

Above that line, it does not hold. The same benchmark data shows human-designed creative keeping an 8% to 14% conversion advantage on products over $100. Considered purchases seem to need something the models have not learned to fake reliably. Anyone telling you otherwise is selling something.

Underneath the adoption numbers sits a second problem.

McKinsey's global survey finds 88% of organizations now using AI in at least one business function, up from 72% in 2024, while only about 6% qualify as high performers extracting meaningful bottom-line value.

Eighty-eight and six. Almost everyone has the tools. Almost nobody has the operation. Whatever closes that gap, it will not be a better model, because the people stuck in it already have access to the same ones as everyone else.

There is a disclosure layer sitting on top of all of this too. Provenance metadata now travels with generated assets by default on most major models, which turns "was this AI" from a detection question into a paperwork question your procurement team will ask. We wrote about what that means for brand-side contracts in Your AI content is about to introduce itself.

Skeptics have a case, and it deserves stating properly. The same Cannes week that created an AI craft category also featured people warning about AI slop from the main stage, and they were not wrong to. A parity number below $100 says buyers cannot tell on cheap goods. That is a lower bar than "the work is good." Both things are true at once.

Public failures have been useful here. Coca-Cola's fully generated holiday film drew real criticism for feeling emotionally flat even as it proved the technical point. Toys R Us took a public beating for an AI brand video. The tools performed exactly as specified in both cases. What was missing was someone deciding whether the output moved anyone, and whether that was the right job to hand over at all.

Inside the 6%

Pull apart the brands in that 6% and the pattern is boring, which is usually a sign it is real. Structured source material. A brand voice defined tightly enough that someone can check work against it. Review standards a reviewer can apply without calling a meeting, and a repeatable path from brief to approved asset.

None of that is a model. All of it is direction, written down somewhere a second person can check it.

Teams that skipped it got what you would expect. More variations, no better work, and in several documented cases more senior rewriting than before the tools showed up. We made this argument at the start of the wave in Without creatives, AIGC is nothing, and three years of brand results have not softened it.

This matters more, not less, as the format count climbs. A campaign that runs on Meta, TikTok, RedNote, and Douyin is not one campaign resized. Each of those platforms has its own read on what looks native, and an asset that lands in one can die in the next. Volume makes that harder to police, not easier.

How we run it at hubStudio

Which is the whole reason the studio is shaped the way it is.

We built it on that split. Senior creatives lead every concept, out of Paris and Hong Kong. AI carries the production volume underneath them, through hubs in China and the Philippines. Generation is the engine. Direction is the product.

The work lands in one of three modes, picked per job. We shoot, when a camera is the right call. We generate, when the piece lives entirely in AI. Or we shoot and then adapt, capturing a hero the traditional way and extending it into the fifty variations a real campaign needs. Different formats, different languages, and a read on what works in each market.

How you buy it is just as simple. Hand us a brief and take delivery. Run your own team on a platform we stand up for you. Or bring us in to direct while your people execute.

Figures from hubStudio engagement data. Across a recent campaign cycle for Mexicash, first-round approval moved from 22% to 78%, usable asset output rose roughly sevenfold, and production costs fell about 60%.

The approval number is the one that pays for itself. Seven times the output means nothing if four rounds of revisions sit behind every asset. Getting it right the first time is what makes volume usable. That comes from the direction rather than the generation. The AIGC adoption curve goes further into where the savings actually sit.

See it on your own brief

Only one judgment counts here, and it is yours, on your own work. Send us a live brief. We will walk you through real brand cases and produce a small set of test visuals against your actual requirements, usually inside a week, at no cost and no obligation. You see the direction and the output before anything is riding on it.