AI in content marketing: what a month of human-only content showed

Aziz says it came from the numbers. The team had noticed reach, impressions and engagement falling sharply for clients' competitors. So it exported six months of content from every client and every competitor and analysed it. One thing ran through everything that performed: a person. A photo, a video, a podcast, a short clip, but always a human at the centre. The question was simple: if we stopped everything else and did only human marketing, what would happen?

The partner was chosen from Cush's clients: one that saw itself as part of the experiment, would give it time, and accepted that it might not work. They ran around five events across X Spaces, a Facebook Live, a webinar and a round table.

The event itself, Aziz explains, was never the point, and nor was the live audience. The point was what came out of it. Most people think a round table is an hour and then it is over. In human marketing, one round table produces about sixty pieces of content: ten short clips, twenty quotes, ten blog posts and a white paper. Enough for a month, and every piece has a person in it.

The first surprise was X Spaces in Saudi Arabia. Run properly, he says, the engagement is very high, and a 30-second audio clip from one, posted on LinkedIn, drew more engagement than video content.

The second was Facebook. Most people have forgotten it, Aziz included. But for people in their forties and fifties, many of them decision-makers, it is still the favourite channel, and marketers trying to reach them forget it.

In the month of the experiment, against the months before, Aziz says impressions and reach rose 400% on LinkedIn, 200% on X, 500% on Instagram and 50% on Facebook. Do people prefer human content? The numbers said yes.

More evidence came the same month. For Saudi National Day the team made two kinds of content for clients: one from the material they had, and a video shot by some clients' own staff. The second, full of people, drew at least 300% more impressions. Aziz had expected the experiment to need two or three months. The answer was clear by the end of the first.

Three years ago, Aziz says, a single Instagram post took the team two or three days: shooting, editing, design, copy, hashtags. Today AI can make that post a hundred times in three hours, at the same quality or better. And anything anyone can produce without effort loses its value. AI is always polished and people are not. If AI imitated him, his Arabic would come out perfect rather than broken. Audiences want to feel the person talking to them, small flaws included.

He expects the platforms to move this way, and points to YouTube. On 15 July 2025 YouTube renamed its "repetitious content" policy "inauthentic content", making clear it covers mass-produced and repetitive content, which cannot earn money on the platform. A platform wants you to stay, Aziz argues, and if it filled up with machine-made content people would leave.

Aziz is not against AI. From this very episode, the team will ask AI for ten quotes from each speaker and where they sit in the video, so the editor knows where to cut. Research that took him weeks now takes a day, and he then goes back over it to think.

"AI makes the clever cleverer and the fast faster, but it also makes the dull duller."

The condition is foundational knowledge: without it you ask the wrong question and get an even wronger answer. Cush's history with these tools goes back to around 2021, when tools built on GPT-3, Writesonic among them, cut small errors in its written Arabic from about 4% of posts to zero. Aziz recalls ChatGPT reaching 100 million users in its first week; the verified figure is 100 million monthly active users two months after launch, according to a UBS study reported by Reuters. Watching it spread, he expected people to want more human contact, not less.

Asked what brands should stop doing immediately, Aziz answers: publishing anything no human has checked. AI makes things up, and he has seen large companies publish content that worked against their own brand. One responsible person who looks and approves is enough, because AI cannot read the small details of human behaviour, or empathy.

If he had to pick one country for human marketing, it would be Saudi Arabia, where you reach a company through a person: acquaintances, friends, family. That is why conferences work better there than anywhere else. He contrasts it with England, where he lived eight years in his last house and took four of them to learn his neighbour's name. To see the future of marketing, he says, look at marketing in Saudi Arabia now, the market we work in from Riyadh.

His advice: start. Pick a channel, an X Space if your audience is there, talk with people about something you care about, and develop the idea as you go. Mistakes are human, and that is why it works. This month's episodes are all in Arabic, he says, and he knows his Arabic is broken, but the information matters. To see how we build this for companies, this is the marketing department we run for clients. For the wider question of jobs, there is the episode will AI replace jobs?