I spent this week reading about where AI is landing in marketing, and the more interesting story is where it isn't. The line people keep drawing turns out to be specific, and it falls almost exactly on the human parts of the work.
Gallup published a poll in August on how Americans feel about AI in advertising. Most people are fine with it for brainstorming and drafting, and just over half said they're comfortable with AI producing the final ad itself. Then the numbers turn. Nearly two in three said using AI to recreate a real person's face or voice crosses a line, even when the business discloses it. Among adults under 30, 73% reject AI-generated people and voices, and 66% view AI-made advertising negatively overall. That's the group brands most want to reach, and it's the least forgiving.
So it isn't AI in general that bothers people. It's AI standing in for a person.
The Washington Post hit the same line from the other side. AI search is shrinking the organic funnel, so the Post rebuilt its paid acquisition around its journalism rather than subscription offers. Journalist-led video delivered more than four times the click-through of its usual direct-response creative, and promoting articles earlier produced a 31% lift in conversion over organic alone. Those are one publisher's numbers, not a benchmark, so I'd read them as direction rather than proof. The direction is clear enough. When the feed fills with generated content, a real person with a name and a byline becomes the thing that earns attention.
Even the models are sorting for this. After the ChatGPT 5.6 update, the measurement firm Peec AI found listicles dropped to 7.8% of citations, down about half, and comparison pages fell to 6.17%. Those are the formats brands were producing at scale because they were cheap to generate. The cheap-to-generate work is the first to lose its value.
None of this means AI is overrated in marketing. The drafting, the first passes, the clustering and analysis, all of it is real and I use it daily. What's happening is a repricing. The work AI does well is getting cheaper, so the work it can't do is worth more by comparison. Judgment about what's worth saying, and a voice a reader actually trusts, hold their value precisely because a machine can't supply them.
That's a more useful frame than asking whether AI will replace this task or that one. The better question is which parts of your marketing earn their value from being human, and whether you're protecting those parts or quietly automating them away.
I'll get to test this against real practice soon. This fall I'm sitting down with marketing practitioners around Vancouver to hear how they're actually splitting the work between people and machines, and where they draw their own lines. I expect the honest answers will be less tidy than the poll numbers, and more interesting for it.
Sources
- Gallup poll on AI in advertising: news.gallup.com
- How AI forced the Washington Post to rethink paid acquisition (INMA): inma.org
- ChatGPT 5.6 citation shift, Peec AI data (via Lily Ray): linkedin.com