There's a new buyer, and it never walks the aisle
Designing packaging for the AI that shops for your customer
We do a lot of our thinking in grocery stores. Nicole and I came up on the buying side at The Fresh Market, so we still read a shelf like buyers: what gets picked up, what gets skipped, whether a loyal shopper can even find you.
But lately, more of that first decision is happening somewhere we can't stand and stare. A chat window.
Someone asks ChatGPT for "a clean, high-protein snack for my kid's lunch," or tells Amazon's Alexa to "reorder my pantry staples," and something picks the brands before a person ever lays eyes on a shelf. This isn't a someday thing. More than a quarter of shoppers already use AI to help make buying decisions (FMI, 2026). On Amazon alone, 300 million people used Rufus last year, and the ones who use it are 60% more likely to finish the purchase.
So we've been digging into what that actually means for packaging. Here's where we've landed so far.
Your brand now competes on three shelves
There's the physical shelf: retail placement and the package itself. You know this one cold.
There's the digital shelf: your Amazon listing, your DTC product pages, your reviews. Most brands are at least paying attention here.
And now there's the third shelf: the layer where an AI assistant decides which brands to put in front of someone. It's the one almost nobody is designing for yet, and it's the one quietly growing the fastest.
The good news, and the reason we think this is a packaging story and not a tech story: the same clarity that wins a distracted shopper in a real aisle is exactly what that machine is reading now.
How the AI actually "sees" your product
Here's the part that got our attention. That new buyer doesn't see your packaging the way a person does. It reads your product as words and as a picture: the name, the claims, the ingredients, the dietary tags, the front-of-pack photo. Then it matches all of that against what someone asked for.
Around 70% of AI product suggestions lean on structured details like ingredients, dietary tags, and origin. So if one of those is blank or buried, you don't rank a little lower. You just don't come up. It's less like losing a beauty contest and more like not being on the list at all.
And the AI reads your package image too. A claim that only lands at full size and disappears at thumbnail gets missed the same way a rushed shopper misses it. Legibility was always a shelf problem. Now it's a discovery problem.
Being asked for by name
The brands that survive this are the ones people request by name. There are flavored seltzers, and then there's Topo Chico. An algorithm can't quietly substitute you out of "add Topo Chico to my cart."
A framework we picked up from the folks at Brand Now calls this earning the ask: giving people (and the assistants shopping for them) a reason to want you specifically, not the closest match. Commodity products get swapped silently. The ones asked for by name don't. Every brand should take an honest look at which one it is right now.
The second moment only packaging can win
Here's the piece we keep coming back to, because it's literally our work.
When an AI does the buying, the brand that "won" never actually met its customer. The product just shows up at the door. That makes the package the one guaranteed moment your brand meets the person who bought it. The unboxing, the in-pantry glance, the thing on the counter.
That's a design opportunity hiding in plain sight. A QR code on the pack doesn't have to be a dead link to your homepage. It can be a real value exchange (a community signup, an offer, something genuinely useful) that turns an AI-delivered stranger into a customer you actually know. When the machine owns discovery, owning your own customer matters more, not less.
What you can do this week (no budget required)
You don't need to overhaul anything to start. A few things we'd actually try:
Ask the robot about yourself. Open ChatGPT (or Rufus, or Google's AI answers) and ask it the way a shopper would: "best [your category] for [your customer]." See if you come up, and see what it says about you when you do. It's a humbling five minutes.
Read your title out loud. Say your product page title and first line to someone who doesn't know your brand. If they can't tell you what it is, the AI can't either.
Make your claims consistent. Most brands have slightly different descriptions, ingredients, and tags across every retailer. That inconsistency is exactly what gates you out of AI results. Pick the true version and make it match everywhere.
None of this is a "hair on fire, redo everything" situation. It's the opposite. It's the same work we've always done, making the few things that matter unmistakable, aimed at a new reader. And it compounds. The brands moving early are already pulling ahead.
That's the shift as we see it. A new buyer at the table, reading your package in a new way, and a design opportunity most of the category hasn't noticed yet.
If you want to see what an assistant actually says when it meets your product, that's something we've started doing for brands. Just reach out.
Andy and Nicole Kurtts run Buttermilk Creative, a food and beverage brand and packaging design firm in North Carolina.
