If AI Can Produce More Ads Than Ever, Why Are Your Margins Still So Thin?

Here's the paradox most affiliates are living inside without naming it. You can produce more ads, faster and cheaper, than at any point in the history of the job. And your margins are under more pressure than they've been in years.

Both are true. They're happening in different parts of the business, and once you separate the parts, the AI question stops being a mood and becomes a diagnosis. One buyer, mid-mood: "Are we evolving… or being replaced?" The honest answer is neither, and both, depending on which part you're looking at.

Part one: the ads got cheap

This is where AI is unambiguously on your side. Hooks, scripts, landing page variants, translations, voiceovers: the work that used to need a team or a retainer now takes a morning. "Write 10 hooks in the morning, have all the videos ready by lunch," as one buyer put it. Another: "now i just build what i need in like 20 minutes and move on with my day." If your constraint was creative volume, that constraint is gone.

The catch is that it's gone for everyone. Cheaper production doesn't give you an advantage; it removes a disadvantage. The buyer who started last month has the same tools. What you do with the extra capacity is the advantage, and "more of the same" isn't it.

Part two: the traffic got harder, but not the way you think

This is where the picture depends on your model, and where most AI commentary blurs things together into one scary sentence.

If you're an organic affiliate publisher, the story is AI answers in search: where an AI summary appears above the results, clicks to the page below it fall, and a large share of searches now end without a click at all. That's a distribution problem, and it's real. One publisher, on the content side of it: "AI content has made low-quality content almost worthless."

If you buy paid social, the story is automation. Meta has said it wants campaigns to run themselves, creative and targeting included, and the automated campaign types already make most of the decisions that used to be your craft. That's a commoditisation problem. The buying skill isn't worthless, but the part of it that lived in the settings is being handed to everyone with a card.

If you're in search arbitrage, both apply at once, which is why that corner feels the way it does.

Different mechanisms, different fixes. Lumping them into "AI is killing affiliate" is how people end up scared of the wrong thing and fixing the part that wasn't broken.

Part three: the lever nobody's pulled

This is the part no model touches, and it's where the margin actually lives.

What you're paid per conversion. Whether the cap is yours. How fast you're paid, and whether a "new affiliate hold" of up to fourteen days, as one network's terms put it, is quietly financing their cash flow with yours. Whether the offer fits your traffic at all. A program cutting its commission, as one buyer described getting in a single lunchtime email, from 25% to 12%, does more to a year's margin than any change in creative cost ever will. And it's decided by people, in conversations, not by a model.

Here's the diagnostic. Take your best campaign and ask three questions.

Has my production cost fallen? Almost certainly yes. Has my acquisition got harder? Depends on your model, but probably. Have my terms improved? For most buyers, honestly, no. They're the same terms the network board offered the day you found the offer, and you've been optimising everything around them while they sat still.

That's the whole paradox in three answers. Production is cheaper, acquisition is tougher, and the one lever that could offset both hasn't been pulled, because pulling it needs a person, not a tool.

What pulling the lever looks like

If that diagnosis fits, the useful next step isn't another creative tool. It's a conversation with the people who set terms, and a look at whether better-structured offers exist for what you run.

That conversation has a shape. Your volume, honestly. Your traffic type and geos. The one term you'd change first and the number you'd ask for. And the thing the cheaper production gives you to offer: velocity, geo adaptation, format range. It's the pitch a one-person studio can now make, and it turns "can I have a bump" into a proposal.

Where the people who set terms are

Affiliate Grand Slam is in Rome, 2 to 5 November, at Fiera Roma. Two examples of who's confirmed, with what the terms conversation is with each. TerraLeads owns and runs its own consumer offers and pays partners per sale; it sets the numbers in part three, and a conversation there is with someone who can move them. ClickBank is a marketplace where offers are listed with their owners' terms; a lap of it shows what better-structured offers exist in your vertical and which owners are looking for direct partners. TikTok, Meta and Google are exhibiting with staff, which is where the part-two question about what the automation is optimising for gets asked of a person.

A confirmed name doesn't tell you who attends or whether they'll accept a request. It tells you who's worth requesting.

The AGS Rome 2026 Welcome Pack lists the confirmed companies by vertical with what each does, and carries the terms-conversation page and the five-people worksheet. If you promote offers or work with affiliate networks, apply for the complimentary Affiliate Pass with proof of activity; review takes up to 24 hours. Once approved, the Match App opens and you request the specific conversations you've identified. They accept or decline.

The pack doesn't decide which side of the AI story you land on. Your terms do. It just makes it easier to find the people who can change them.