How AI Decides What's for Dinner

The Coming Revolution in How We Shop, Order, and Discover Food

Twenty years from now, we might look back at food bloggers who wrote three thousand words about their childhood before sharing a brownie recipe as the good old days. Not because we'll miss the narrative bloat, but because those human quirks and inefficiencies were the last time food discovery felt genuinely messy, surprising, and human. AI is beginning to change how we search for everything online—food included—and the implications run deeper than we realize.

That ChatGPT conversation about what to do with leftover chickpeas isn't just convenient—it's training you to expect instant, frictionless answers to food questions you haven't even fully formed yet. Once you taste that efficiency, going back feels impossible.

The Delivery Apps Already Know You

Online food buying is where AI's impact is being felt first. DoorDash knows you order Thai on Fridays and Chinese on Sundays. Instacart has figured out that you buy oat milk every 11 days and only spring for name-brand cereal on sale. Uber Eats can predict that rain makes you crave ramen and you'll abandon your cart at $47 but complete it at $45. These apps know your patterns better than you do—every click, every scroll, every hesitation before checkout. The recommendations feel helpful until you realize you haven't surprised yourself with a food choice in months. The question becomes whether we actually want our choices predicted all the time.

Recipe discovery hasn't fully transformed yet, but LLMs are about to upend how we search for what to cook. Think about the current nightmare. You Google "chicken dinner recipes," click through twelve blogs, scroll past someone's life story about their semester in Florence, only to find a recipe that requires sumac and a mandoline you don't own.

Now imagine just asking "I have chicken thighs, wilted spinach, and half a lemon. My kid's allergic to nuts and I need dinner in forty minutes." The AI instantly gives you exactly what you need—no life stories, no pop-up ads. It adjusts the recipe for three people instead of four and swaps butter for oil because it knows you're dairy-free.

You can interrupt mid-recipe with "What the hell is deglazing?" and get instant answers. The difference between hunting through Google results and having a conversation with something that actually understands your context isn't just convenience—it's the difference between searching for recipes and actually cooking.

Grocery Gets Smart

COVID forced every grocery chain to accelerate their online shopping timeline by about five years. Since then, apps like Instacart, Amazon Fresh, and retailers' own platforms have been quietly accumulating massive amounts of data about our buying habits. For now, we mostly see basic recommendations—"Because you bought pasta, you might want pasta sauce"—the kind of rudimentary matching that's now ubiquitous in any e-commerce experience.

But grocers have been swimming in oceans of consumer data for decades, and enterprise AI finally gives them the tools to do something interesting with it at scale. What happens when Walmart embeds ChatGPT-level intelligence directly into their shopping platform? Suddenly you're not just reordering groceries—you're conversing with something that knows your entire purchase history and can think creatively about it.

"I'm hosting Thanksgiving for eight people, including my gluten-free sister" becomes a complete shopping list that remembers you always forget aluminum foil for leftovers and that your family prefers jellied cranberry sauce over whole berry. The leap from primitive recommendations to genuinely intelligent shopping assistance isn't a technical problem anymore—it's just a matter of implementation.

CPG Brands Face the Algorithm

CPG brands are scrambling to understand their place in this evolving ecosystem. Physical shelf space—that precious real estate that Big Food pays millions to secure—might matter less when AI gets really good at recommending products and if meaningful numbers of shoppers stop going to stores altogether. For the growing segment buying groceries through apps and chatbots, the real battle shifts to algorithmic visibility. Getting properly indexed in Instacart's recommendation engine or surfacing when someone asks their AI for "healthy kids snacks" becomes the new endcap.

For now, ChatGPT and other LLMs don't slip sponsored products into recipe suggestions, but how long before that changes? The shift from transparent advertising to conversational commerce raises uncomfortable questions. There's something deeply unsettling about the difference between seeing "Sponsored" links in Google versus having a conversational AI casually recommend Kraft Mac & Cheese mid-conversation.

It's the Truman Show problem—Laura Linney pivoting mid-argument to pitch Mococoa—or that Black Mirror episode where the woman with the ad-supported brain implant randomly blurts out product placements. Once your trusted AI assistant starts selling to you, does it break the integrity of the relationship someone has built with their LLM?

Big CPG companies that once dominated through marketing muscle might watch smaller brands punch above their weight by targeting micro-segments the algorithms identify. Small-batch producers who could never afford Super Bowl ads might suddenly find their artisanal pasta sauce recommended to exactly the right customers, assuming they can afford whatever "algorithmic trade spend" emerges.

But where's the room for stumbling upon something unexpected when AI pre-filters everything? That weird Japanese candy you grabbed on impulse, the farmer's market honey that became your obsession—these moments of chaotic discovery might become casualties of optimization.

Restaurants Split in Two

The restaurant dining experience faces an entirely different set of AI challenges than grocery shopping. McDonald's already uses AI-based decision technology through its Dynamic Yield platform. Cheesecake Factory's technology stack manages their encyclopedic menu with precision. These chains have the resources and centralized control to make AI work. Your neighborhood bistro? That's another story.

For most independent restaurants, keeping menu data current and AI-ready would be a nightmare. Consider what actually happens in a working kitchen. The morning produce order arrives wrong, forcing today's salad swap. A key ingredient runs out mid-service. The daily catch changes based on what's actually fresh and available.

This isn't an edge case, it's a Tuesday night in restaurants. Chain restaurants can standardize their supply chains and menu offerings enough to make AI integration seamless. But independent restaurants run on improvisation and morning market finds. That beautiful unpredictability that makes dining out worthwhile is exactly what makes AI integration difficult.

Perfect AI recommendations might be exactly what you want for your weekday lunch salad order at the office—quick, optimized, no surprises. Date night at that new place downtown? Different story. You don't want an algorithm steering you toward the "safe" choice based on your history. Part of the experience is discovery, the server's recommendation, maybe trying something you can't pronounce.

The technology creates a bifurcated dining future. For transactional meals—breakfast sandwiches, desk lunches, quick dinners between activities—AI recommendations could remove friction we never wanted anyway. But for experiential dining, where unpredictability and human curation are features, not bugs, algorithmic optimization becomes a threat to what makes restaurants matter.

The Native Generation

The most fascinating shift might come from a generation that never knew a world without ChatGPT. Kids born after 2022 won't find it strange that an AI helps mom plan dinner or that the grocery list writes itself. For them, asking an algorithm for recipe suggestions will feel as natural as we find scrolling through Instagram for meal inspiration. They might look at our nostalgia for browsing farmers markets and discovering random products the way we look at our grandparents' insistence on clipping paper coupons by hand—quaint, inefficient, and unnecessarily complicated.

This native AI generation might not even understand the concept of "not knowing what you want to eat." Their food decisions could be so seamlessly augmented that the distinction between their preferences and algorithmic suggestions dissolves entirely. They might develop entirely new relationships with food discovery—perhaps valuing algorithmic serendipity over human unpredictability, or creating new rituals around deliberately choosing chaos over optimization.

Or they could rebel entirely. Much like millennials rejected the ultra processed TV dinners and Wonder Bread their parents and grandparents celebrated as modern miracles, Gen Alpha might see AI food assistance as something deeply uncool—a crutch their parents relied on because they couldn't trust their own taste.

We might see a radical return to analog food discovery, where knowing how to shop without algorithms becomes a status symbol, like knowing how to drive stick or bake bread from scratch. Think of how Gen Z discovered vinyl records—hunting for something tactile and real in an increasingly digital world. The farmers market could become their record store, the handwritten recipe card their vintage album, the spontaneous restaurant discovery their underground band.

Generations aren't monocultures, and food is too contextual for any single approach to dominate completely. The same person who uses AI to streamline their Tuesday night grocery run might spend Saturday mornings wandering the farmers market, touching every tomato and talking to vendors about their chickens. They might let an algorithm plan their meal prep but refuse to let it anywhere near their dinner party menu.

This analog-digital divide won't be a battle with winners and losers—it'll be a permanent coexistence, with people moving fluidly between modes depending on their mood, their time, their company, and what kind of experience they're after. Food is too personal, too cultural, too emotional to ever be fully automated or fully manual. We'll use AI when we want efficiency and ignore it when we want adventure, and that might be the most human outcome of all.

The Choice Ahead

The real transformation isn't in any single application but in the cumulative effect of AI touching every digital food interaction. Order takeout, AI suggests options. Search for recipes, AI customizes them. Buy groceries online, AI predicts your needs. Each interaction seems helpful in isolation, but together they create a food life increasingly mediated by machines.

Physical grocery stores might remain surprisingly unchanged on the surface. The produce section will still let you squeeze avocados. But online ordering—whether for delivery, pickup, or meal kits—becomes an entirely AI-mediated experience. Two parallel food systems emerge—analog in-store shopping for those who want it, and increasingly sophisticated AI assistance for everything else.

The question isn't whether AI will change how we shop for food—it already has. That viral video featuring a hot sauce you saw on TikTok that led you to order a bottle was served up to you by an algorithm. Is that considered discovery or being sold to? Every recommendation, every subscription, every predictive cart addition shifts us slightly toward a more automated food future. Will we maintain spaces for inefficient human choice, or will the convenience of AI optimization eventually feel so natural that we forget we ever did things differently?

Perhaps the most profound question isn't about technology at all, but about desire itself. When AI can predict what we want to eat better than we can, when it can satisfy our cravings before we fully feel them, what happens to the human experience of wanting? Do we lose something essential when we never have to be disappointed by a meal choice, never have to compromise, never have to discover we actually hate what we thought we'd love?

For those of us caught between worlds—old enough to remember true analog food experiences but young enough to embrace the convenience of AI-assisted food buying—we face a choice with every meal. We're training AI to understand our food preferences while still preserving our right to surprise ourselves.

How long that balance lasts might depend not on the technology, but on how much we value the messy, inefficient, gloriously human act of not knowing what we want until we find it.

—

Because you enjoyed this article, you may also enjoy…

• My Book - Mise: On the Future of Food

• My Podcast - The Tomorrow Today Show

• My Instagram - The Book of Mise

• My Other Instagram - Mike Lee

• My Consultancy - Mise Futures