Positioning Is the Last Moat When AI Recommends Everything

It's a Tuesday night and someone is lying on the couch, phone held above their face, talking to an assistant instead of typing into a search bar. "I want a greens powder, but I can't stand the taste of most of them, and I want the brand to actually care about what's in it." Three seconds later there's one product on screen. Not ten. Not a grid. One, with a little line underneath explaining why it's the pick for this exact person.

That moment is closer than most founders are planning for. And it quietly changes the question every ecommerce brand has to answer.

For fifteen years the game was "how do I get found". Rank for the keyword, win the auction, show up in the feed. The game that's coming is "when the machine already knows this person better than they know themselves, why does it pick me". That's a different question. And I don't think more spend answers it.

Why discovery is moving from the shelf to the assistant

Search used to hand you a wall of options and make you choose. The next layer of discovery doesn't do that. It narrows. It looks at what a person has bought, returned, lingered on, complained about, and it serves a recommendation shaped to them specifically.

The important word there is "specifically". AI won't just recommend the best product. It'll recommend the best product for that one human, based on what they care about and who they think they are. So the brand that wins isn't the cheapest or the loudest. It's the one that maps cleanly onto an identity the machine can recognise.

Which means the thing you've maybe treated as fluffy - what you actually stand for - becomes the most practical asset you own.

Stand for something, or be a rounding error

Here's my take, and I'll be blunt about it. Most ecommerce brands stand for nothing. If you asked the founder "give me one logical reason someone should buy from you instead of the three near-identical stores next to you", they couldn't. They have a product, a Shopify theme, and a media buyer. That's not a brand. That's arbitrage on a platform, and platforms close arbitrage windows.

The brands that compound do the opposite. They pick an enemy. Not a competitor - an enemy. A thing they refuse to be.

The water brand whose enemy is plastic and boring "wellness" branding. The greens brand whose enemy is the powder itself, the mess, the taste, so it refuses to ever sell a powder. The coffee brand whose enemy is what veterans carry home and have to live with. None of those brands are for everyone. That's the point. They're violently for someone.

To put this in perspective with invented but realistic numbers: imagine two skincare brands both spending ~$40k/month on Meta. One sells "clean, effective skincare for everyone". The other sells "skincare for people in their 40s who are done being talked to like teenagers". Same products, roughly. The second one has a narrower door, but the people who walk through it convert at a far higher rate, return less, and tell their friends. When an assistant is matching a 44-year-old to a brand, which line do you think it picks up on?

A clear position doesn't shrink your market. It tells the algorithm, and soon the AI, exactly who to hand you to.

Storytelling is the moat AI can't copy

Now to the part I find genuinely interesting. We're about to drown in informative content. Anyone can ask a model for "five tips" or a "structured explainer" and get something passable in seconds. That stuff is going to zero in value because supply is going to infinity.

The thing a model can't manufacture is the true reason a founder started. The actual mission, the real story, the late nights and the bet that nearly broke the business. That's not content you can prompt into existence. It happened, or it didn't.

And story is how humans decide. A purchase is an emotional decision first and a logical justification second, every time. Storytelling is just the vehicle that carries the emotion. So a founder who can tell their story well isn't doing marketing fluff - they're using the one channel a competitor's GPT output can't fake.

There's a simple structure underneath every good story, and it's worth stealing: intent, obstacle, resolution. A character wants something. Something gets in the way. How they handle it is who they are. Brands have this too. Your mission is the long-running show. Every obstacle you hit - a supplier who cut corners, a cheaper knock-off flooding the category, the temptation to dilute the product to widen the margin - is an episode where your values get tested in public. Handle it in line with what you stand for, and everyone who shares those values pulls in closer.

I've watched a version of this work without a single sale being asked for. Picture a small brand, two years in, that emails its list during a genuinely scary moment - not a discount, not a launch, just the honest story of why they started and what they're trying to do. No link. Money comes in anyway, because the connection had been built story by story for two years. You cannot buy that with a CPM. You earn it, slowly, by being a brand that means something.

What I'd let AI do, and what I'd never hand over

We're not anti-AI at Pigeon. Far from it. We use it daily across client work, and I think founders who refuse to are choosing to lose. The trick is being deliberate about the line.

Here's roughly how I split it:

• Hand to AI: the volume work. Generating creative variations, cutting fifty hooks down to the ten worth testing, drafting first-pass scripts, resizing and reformatting, sorting data, surfacing patterns in the account we'd otherwise miss. This is where speed actually helps.
• Keep human: the core narrative. The enemy you've chosen, the mission, the values, the founder's real story, the single sentence that says who you're for and who you're not. The taste call on which angle is true to the brand and which one is just clever.

Put plainly: let the machine produce the permutations, but only a human decides what the brand believes. AI can make your story travel further and faster than ever. It can't decide what the story is. The moment you let it, you sound like everyone else who's prompting the same model with the same request, and you're back to being a rounding error.

Why this actually lowers your cost to acquire

This isn't a branding lecture dressed up as strategy. It shows up in the numbers. When your positioning is sharp and your story is real, your creative resonates harder with the right person, which means a higher click-to-buy rate, which the algorithm rewards with cheaper delivery. Tighter narrative, lower CAC. They're the same lever.

The brands that are going to struggle in an AI-recommended world are the ones still treating brand as a logo and a colour palette while competing entirely on price and ad spend. The brands that win will be the ones the machine can describe in one sentence - because they spent the time deciding what that sentence is.

So the question I'd sit with this week isn't "how do I out-spend the category". It's quieter than that. If an AI assistant had to recommend your brand in one line, to one specific person, what would it say - and is it true?

If you can't answer that cleanly, that's the gap. A Signal/Noise Audit is one way to get an outside read on it: we look at your account, your creative history, and your unit economics, and we'll tell you plainly whether your positioning is doing any work or whether you're paying the platform to paper over the fact that there isn't one. No pressure either way. But that one sentence is worth getting right before the assistants start writing it for you.