AI content creation: what actually works, and what still needs a person

Summary

Almost every content team uses AI now: 93% of content marketers use it to help make content, and 75% of creators call it integrated or essential to how they work.

Quality has not followed. 88% of marketers say AI increased how much work they produce, but only 45% say it improved quality significantly, and 57% of creators say the output still needs moderate or extensive editing before it goes out.

Singapore's firm data shows the same shape. 28.5% of firms have adopted AI, and only 3.8% have built it into how they work. The skill in short supply is not prompting. It is judgement.

Almost everyone making content uses AI now

Adobe's 2026 Creators' Toolkit Report surveyed more than 16,000 creators across eight countries. Among those who use creative AI, 87% say it has sped up the growth of their business or audience, and 75% call it integrated or essential to how they work.

The pattern holds everywhere. Among designers, 91% use AI for design tasks at least weekly, up from 54% a year earlier, and the average set of tools has grown from three to seven. Among video marketers, 63% have used AI tools to create or edit marketing videos, up from 51% a year ago.

Turning up with AI is no longer the interesting part. It is the baseline. The question is what you do after the tool has produced something.

More content is not better content

When WARC and TikTok asked 400 marketing leaders in July 2026, 90% said AI had become a core creative tool and 88% said it had increased their creative volume. Only 45% said it had significantly improved quality. That gap is the story of the year: much more content, and not much more of it that is good.

The growth is fastest at the bottom. In Ahrefs' 2026 survey of content marketers, 69% publish a bigger share of AI content than a year ago. 7% say more than three-quarters of what they publish is pure AI, with no editing or extra input, up from 4% the year before.

There is a cost to it. Adobe found 53% of creators point to sheer volume when they say standing out is harder than a year ago, and 42% say AI-generated content is making it harder for distinctive voices to be heard.

The part the tools still get wrong

Faster to draft is not the same as ready to publish. In Adobe's survey, 57% of creators said their AI output typically needs moderate or extensive editing before it is ready to share.

The AI in Design Report 2026, an industry survey of designers working in tech, reaches the same place from another direction. Output quality is at once the main reason designers stay with a tool and the biggest weakness in the tools they use. Reliability, not capability, is the complaint.

Ahrefs' numbers hint at how people are adjusting. 40% of marketers say none of what they publish is pure AI, down from 50% a year earlier. Fewer are willing to publish the raw output.

What actually works: keep the human signals

The clearest evidence about what to do comes from a study of real advertising. Taboola worked with researchers at Columbia, Harvard, the Technical University of Munich and Carnegie Mellon. They looked at more than 500 million impressions and 3 million clicks in live campaigns, comparing AI-made ads with human-made ones for the same advertiser, on the same day.

The AI ads performed just as well: an average click-through rate of 0.76%, against 0.65% for the human-made ones, and they held up under the tightest statistical controls the researchers could apply.

Which AI ads won is the part worth keeping. It was the ones that did not look like AI. The strongest signal of trust in the study was a large, clear human face in the picture. So the lesson is not that AI creative matches human creative. It is that people respond to human signals, and AI-made work can carry them or leave them out.

What to learn first

Three things come up again and again in this evidence, and none of them is prompting.

The first is judgement about what to keep. If 57% of the output needs serious editing, the editor is the scarce job, and it is one you can practise on your own material this week.

The second is knowing which jobs suit the tools. Images, variations and short video give the clearest gains, and video is moving fastest: 63% of video marketers now use AI video tools.

The third is taste, the least teachable and the most valuable. Adobe's reading of its own survey is that point of view is what separates one creator from another now that AI use is ordinary.

Singapore's firm data agrees from the employer's side. The Ministry of Manpower found the two biggest barriers were cost (44.9%) and a lack of expertise in house (42.4%). For smaller firms, the most common first step was training staff (46.6%).

If you would rather learn this with your hands, the two-day Hands-On AI Content Course covers AI images, AI video, UGC-style content and building your own AI twin. The next run is 22-23 October 2026, with another on 17 and 24 October. Details at elzatian.com/courses/ai-content-course