By Darya Nikolaeva — marketing lead, ecommerce and social across the US, Canada and UK
Every second brand meeting I sit in this year opens with the same question: "Should we go AI or keep it human?"
It's the wrong question. And it's costing people money.
I've spent the last two years running paid social, TikTok Shop, Amazon and DTC for pet and CPG brands across three English-speaking markets. I've shipped AI creative that outperformed a $12K studio shoot. I've also watched a beautiful AI-generated campaign get buried under 200 comments saying "why does this look so fake" — and tank a product launch in a week.
The pattern behind both isn't "AI good" or "AI bad." It's something much more useful, and almost nobody is saying it out loud.
Let's go through it properly.
First: everyone has the tools now. That's the whole problem.
AI assistant apps: monthly active users, 2026 - US / Canada / UK — directional, from our own accounts
Look at that chart and let it sink in. In the US alone, we're talking about a nine-figure monthly audience across AI assistants. In the UK and Canada, penetration per capita is right behind. Your customer is not "someone who might one day encounter AI." Your customer opened an AI app this morning, before they opened Instagram.
This changes two things at once, and most marketers only account for the first:
1. Production cost collapsed. A concept-to-creative cycle that took my team nine days in 2023 takes about four hours now.
2. Because of #1, output stopped being a differentiator. When a tool is universal, the thing it produces stops being scarce. And nothing that isn't scarce commands attention.
This is the trap. Brands rushed in to save money on production and quietly destroyed the only thing that made their creative work: it looked like it came from someone. Your competitor is using the same model, the same prompt structure, probably the same three reference aesthetics. The feed has flattened into what people now openly call slop — and audiences developed an eye for it faster than any of us predicted.
Your customer can spot AI creative in about 0.4 seconds. They just can't always articulate why. And they don't need to — they've already scrolled.
What people actually click
These are directional numbers from our own accounts, not a study - read them as the shape of the pattern, not as a benchmark to plan against.
Paid social CTR by creative type, indexed against brand-average creative — directional, from our own accounts
Read this carefully, because the interesting part is the middle bar.
Fully AI-generated creative — static or video, no real human, no real product footage — underperforms by 20–30%. Predictable.
But AI-assisted edits of real footage land above average. Same models. Same tools. Completely different result.
And at the top, by a mile: a real person, holding the real product, shot on a phone. Founder-on-camera content sits just behind it, and for early-stage brands it often beats creator content outright, because there's no ambiguity about who's talking.
So here's the actual rule, and it took me about $40K in ad spend to learn it:
AI loses when it's the subject. AI wins when it's the crew.
Generate the entire scene — you lose. Use AI to cut, caption, localise, upscale, variant-test, translate, storyboard, script, and pump one real clip into forty variants — you win, and you win at a cost structure your competitors can't touch.
The brands crushing it right now aren't choosing between AI and human. They're using AI to make more human content, faster. That's the whole game.
Different platform, different physics
This is where most strategies fall apart. Teams build one creative philosophy and push it everywhere. But these channels reward genuinely opposite things.
Conversion rate index by creative type and channel — directional, from our own accounts
TikTok Shop — raw wins, and it isn't close. The platform's economics run on trust transfer from a creator to a product. Polish reads as advertising, and advertising reads as risk. My best-performing TikTok Shop assets are shaky, badly lit, one-take, no music bed, someone talking directly at the lens. AI-generated visuals are the worst-performing category I've ever run there — plus you're carrying disclosure requirements and a real chance of reduced distribution if content reads as synthetic. On TikTok, AI creative isn't just weak. It's a liability.
Instagram / Meta Ads — the most forgiving, and the most misread. Meta's algorithm is a volume machine: it wants many variants fast so it can find the pocket of people who convert. This is where AI genuinely earns its keep — not generating the hero asset, but multiplying it. One good creator video becomes 30 hooks, 12 aspect ratios, 6 caption treatments, 4 languages. That's an AI job. But the source still has to be real. AI-generated visuals still sit meaningfully below average here, they just don't get punished as brutally as on TikTok.
Amazon — the exception nobody talks about. Amazon isn't a feed. It's a decision engine. The shopper is already in buying mode and what they need is clarity, scale, texture, and the sense that this is the actual object that will arrive at their door. Clean studio product photography beats everything. Raw creator content actively underperforms on the PDP — it belongs in Posts and video modules, not in your main image stack. And AI-generated product imagery is the single most dangerous thing you can put on Amazon: it inflates expectations, drives returns, spikes "not as described," and quietly poisons your review velocity. That damage takes months to undo. I've watched it happen.
One brand. Three channels. Three completely incompatible creative strategies. If you're running a single content pipeline across all of them, you're leaving 30–40% of your performance on the table right now.
The US backlash is real, and it's a business risk
Here's the part I keep pushing clients on, because it's not showing up in dashboards yet — it shows up in comment sections first.
US social users' reaction when brand content is visibly AI-generated — directional, from our own accounts
American social users have moved from curiosity to open hostility toward AI-generated brand content, faster than any other market I work in. Not toward AI as technology — plenty of them use ChatGPT daily. Toward brands using AI on them.
The reading is simple: if you couldn't be bothered to make this, why should I be bothered to buy it?
AI content has become a proxy signal for corner-cutting. And once a customer decides you cut corners on your creative, they start wondering where else you cut them. For a supplement, a food product, or anything that goes into a body — human or animal — that is an extremely expensive assumption to let form.
Meanwhile the same audience is rewarding the opposite end of the spectrum: unedited, unglamorous, obviously-real content from people with 4,000 followers and no ring light. Not because it's better. Because it's verifiable.
Market nuance from running all three: the UK sits a step behind the US on backlash intensity but converts hard on dry, understated, credible content — overclaiming kills you there faster than AI does. Canada is the most tolerant of the three and the most price-led, which makes it a genuinely useful test market: if creative works in Canada but dies in the US, you've usually got an authenticity problem, not a product problem.
What I actually do now
No philosophy. Here's the operating model I run: AI backstage. Humans onstage.
— Research, analysis, planning, scripting, hooks, variant generation, localisation, editing, captions, SEO, listing copy, reporting — AI. All of it. Aggressively. This is where the margin lives.
— The face, the hands, the voice, the product, the proof — real. Always. No exceptions.
— Amazon PDP — real product photography, no AI imagery in the main image stack, ever.
— TikTok — raw, native, creator-first. If it looks produced, it's already lost.
— Meta — one real source asset, multiplied into as many AI-assisted variants as the algorithm will eat.
— Every asset — passes a single test before it ships: does this prove a human made a decision here?
That last one is the whole framework compressed into one line.
The expert take
The AI-vs-human debate is a distraction, and it's a comfortable one, because it lets marketers argue about tools instead of doing the harder work.
Here's what's actually happened: AI has become infrastructure, not strategy. Nobody wins market share for using electricity. Nobody's going to win it for using AI either — within eighteen months, everyone will, and the ones who don't will simply be gone.
Which means the competitive question flips completely. It's no longer "how do we produce more with AI?" Everyone will. It's:
"What do we have that AI cannot manufacture?"
A founder who actually knows why the formula changed. A customer whose dog stopped scratching. A factory floor. A mistake you owned publicly. A face that's been on the account for three years. A point of view someone could disagree with.
Those are the scarce assets of 2026. Everything else is a commodity with a rendering cost attached.
So use AI everywhere — and I mean everywhere. Just never let it be the thing your customer is actually looking at.
The brands that win from here aren't the ones producing the most content. They're the ones producing the most proof.
Working on ecommerce or social across the US, Canada or UK and want a second opinion on your creative mix? Get in touch: 1door.cc
Comments
No comments yet. Be the first.