By the week before Black Friday, group chats stop behaving like group chats and start behaving like project management software with better memes. One friend is live-blogging lightning deals from her desk. Another is rage-texting screenshots of “doorbusters” that look suspiciously like last week’s prices wearing more exclamation points. A cousin has built a color-coded Google Sheet for everyone’s wish lists, with tabs for “kids,” “parents,” and “troublemakers.” Someone’s aunt is essentially running a one-woman affiliate network, firing off links at all hours with the zeal of a QVC host.
And threaded through all of it is a new character. It doesn’t sleep, doesn’t shop for itself, and has zero emotional attachment to those pesky, free-shipping thresholds: the AI assistant.
The assistant remembers that you’ve searched “TV deals under $600” three times this week. It knows that your mom hates clutter, your dad hoards loyalty points like a dragon, and your sister will absolutely buy that one viral pan if it drops under $80. It can read more product pages in 30 seconds than you can in a weekend. And it has no idea who’s mad at whom this year, which is probably an advantage.
2025 marks the first holiday season where that AI presence feels less like a parlor trick and more like it's part of the family shopping psychosis.
Bain says that for some retailers, up to a quarter of their referral traffic is now coming in through AI agents and answer engines — still well under 1% of total traffic, but with unusually high conversion. The consulting firm is warning retailers that third-party agents such as ChatGPT, Gemini, and Perplexity are on track to sit between them and their customers unless they build agents of their own and make their sites legible to everyone else. Retailers are rewriting product pages in AI-friendly formats so that the tools can parse them more easily; and some retailers are taking it a step further, building “invisible” pages meant more for AI crawlers than humans.
McKinsey pegs the long-term “agentic commerce” prize in the hundreds of billions of dollars as agents orchestrate more of that spend. According to the consulting firm, agents are being built to “anticipate consumer needs, navigate shopping options, negotiate deals, and execute transactions,” swapping human-driven browsing for autonomous, multi-step shopping workflows. In Adobe’s 2025 consumer survey, 38% of U.S. shoppers say they’ve already used generative AI for online shopping and 52% say they plan to use it this year, mostly for the stuff people hate doing by hand — researching products, hunting deals, building shopping lists, and figuring out what to get the cousin they only see twice a year.
Across Deloitte, PwC, and other holiday trackers, the same picture keeps showing up: Tight budgets are forcing shoppers to adapt, and their solution is treating the holidays like a tactical operation. According to a recent RetailMeNot survey, 65% of shoppers plan to hit Black Friday and Cyber Monday sales. Fifty-eight percent are leaning on coupons and promo codes. About half are playing the cash-back and rewards game, squeezing points out of every swipe. A quarter are flirting with buy-now-pay-later. People are trimming gift lists, downgrading trips, and quietly swapping “big experience” gifts for “small but nice” ones.
And now, layered on top of all the usual scheming, about half of them are bringing AI into the mess.
In RetailMeNot’s survey, 52% of shoppers say they’re using AI somewhere in their holiday shopping. A quarter say they fully trust it to find deals and gift ideas. Another quarter use it as a starting point and then double-check. The rest are still side-eyeing the whole thing from the couch, which is more than fair. But the generational split is unmistakable. Gen Z and millennials are the ones handing the grunt work to the robot and saying, “Go figure it out while I’m at work.”
“AI is now that starting point versus the finish line,” Stephanie Carls, RetailMeNot’s retail insights expert, told me. “It is inspiring that buy, but shoppers are still leaning on those trusted savings platforms to then feel confident once they are reaching the checkout line.” A few years ago, she said, the hunt started with Google. Type the thing, click the ads, drown in tabs. Now, the search starts not with, well, a search but with a prompt. “We can say the AI has officially entered the group chat for holiday shopping,” she said. People are asking it what to buy, where to buy it, and how not to get fleeced on the way to the checkout screen.
When the “pink, sparkly world” she likes to imagine — the one where you screenshot a couch, and it appears in your house for less than you expected — starts showing up in your living room, it stops feeling theoretical. She knows how crazy that sounds. She also just used AI-plus-early-Black-Friday sales to redo her entire living room.
The AI boom has been “coming for retail” on earnings calls and analyst slides for years. But now, the agent in your pocket has quietly wormed its way into the most chaotic, emotionally loaded, logistically cursed shopping ritual in America and started taking over the worst bits — the tab-hopping, the spec-checking, the doorbuster detective work you used to do half asleep on a Thanksgiving hangover. Which raises a simple question:
On Black Friday, why would you stand in a strip-mall parking lot at 4 a.m., in the dark and the kind of cold that bites through three layers of fleece, when a chatbot can wait in line for you?
Black Friday shopping gets an intern
Holiday shopping used to be a scavenger hunt you ran by hand. You saw a candle on TikTok, a mug on Instagram, a sweater in a newsletter, and dutifully saved every link into a tiny digital junk drawer called “gift ideas.” You left 27 tabs open. You told yourself this counted as a plan.
Now, an AI assistant sits in the middle of that chaos like an overqualified intern.
You still scroll. You still screenshot. You still have an entire folder of “things my sister would like if she weren’t so picky.” The difference is that when something catches your eye, you can shove it at an AI and say, “Tell me if this is actually a good deal,” or, “Find this in a size medium that will actually arrive before February,” or “OK, but where’s a version of this that doesn’t cost my entire rent.”
Carls says she’s watched that shift happen in real time. Shoppers still arrive through social feeds and email blasts. The first stop after that has changed. Her favorite example right now is TVs, because nothing exposes Black Friday nonsense faster than a wall of almost identical TVs with almost identical prices. One has a better panel. One is last year’s model in disguise. One quietly dropped a couple of ports you actually need. In the old system, you stared at spec sheets until your eyes crossed. In the new, AI-friendly one, you hand the problem to a model that can compare refresh rates across five retailers while you answer your aunt’s text about what you’re bringing to dinner this year.
The human jobs haven’t vanished. Somebody still has to decide whether Uncle Mark is really “a projector guy” or just aspirational. Somebody still has to know that your mother will never touch a smart speaker. But the grunt work — the spec-checking, the price-tracking, the fake “markdowns” — is suddenly something you can outsource. That’s the entry point. The question is who gets to own what happens after the bot says, “Buy this one.”
Behind the scenes, it looks more like a Black Friday staff meeting. OpenAI has spent the fall turning ChatGPT into less of a homework buddy and more of a virtual mall. Instant Checkout arrived in late September with Etsy and Shopify as launch partners, so U.S. users could buy from indie sellers and long-tail merchants directly in chat. Then Walmart showed up, with a deal that lets you plan meals, restock basics, and check out inside ChatGPT while Walmart handles the fulfillment. A few weeks later, Target joined with its own app inside ChatGPT, where you can build multi-item baskets — including fresh food — and send them to Drive Up, pickup, or shipping through your Target account.
Together, that’s everything from the handmade mug to the console bundle to the toilet paper run, all running through the same agent.
For OpenAI, the holiday timing is the point. Instant Checkout, the Etsy and Shopify rollout, the Walmart and Target integrations, the grocery and delivery experiments — all of it hit as shoppers were turning their group chats into battle plans and their browsers into minefields. If ChatGPT can get through this season as the savviest shopper in the chat who knows every spec sheet and every promo code, it graduates from cute helper to default starting point.
The battle for the ‘buy' button
Now, companies such as PayPal and Stripe are staring at this same chaos and seeing something more structural. PayPal plugged its wallet into the same world and struck its own deal with OpenAI to launch inside ChatGPT — powering “Instant Buy” flows for OpenAI’s personal shopper, where you can do the whole journey, from “What should I get my brother” to “Your order is on the way,” without ever touching a traditional checkout page. Forrester is describing a “race to agentic payments,” with card networks and wallets quietly retooling so that AI agents can initiate and authorize purchases safely — even as only about 24% of U.S. online adults say they actually trust AI to make routine purchases on their behalf.
Prakhar Mehrotra, PayPal’s senior vice president of AI, breaks commerce into four jobs: finding the thing, knowing what exists, moving the money, and fixing it when something goes wrong.
In the tab-hopping version of the internet, those jobs lived in completely different places. Google handled the finding. Retailers handled the catalog and the cart. Payment processors tried not to trip over each other at the checkout page. If something arrived broken, you hunted down a support email and hoped. In an agentic world, all of that collapses into one conversation. You tell an assistant, “Find me a Kitchen-Aid mixer in the perfect color.” The assistant has to understand the request, look across catalogs, sort out which offers are real deals, figure out which payment method makes the most sense, and keep enough of a paper trail so that, if the mixing bowl shows up cracked, someone can fix it.
“The real estate is very small,” Mehrotra told me. When you search on a retailer’s site, you get a grid of options. When you ask an agent, you get a single answer or, at best, a shortlist. That answer is suddenly the most valuable shelf space in retail. Whoever controls the pipes supplying that answer — the catalog format, the payment rails, the post-purchase guarantees — has a lot more leverage than whoever is screaming “50% OFF” in a banner ad.
That’s why Mehrotra says the idea of “agentic payments” at PayPal came from staring at everyone else’s protocols and realizing they’re just the roads the agents will drive on. He describes PayPal’s role now as a kind of conduit that plugs into “every network, every issuer, whatever protocols they have,” so that as agents get more autonomous, the money still moves cleanly, and the buyer is still protected. That’s where his world of “protocols” comes in — standards that tell agents how to behave so they don’t burn the whole thing down. Commerce needs the same thing social media eventually had to invent for logins and payments. Without it, your chatbot is just a very confident guesser with your credit card on file.
None of this answers whether agentic AI will kill the thrill. Mehrotra, when asked if he could imagine a future where an assistant sees your calendar, spots a friend’s wedding in September, and simply tells you what to wear — an intrusion that, for anyone who lives for the thrill of the perfect-dress hunt, feels like sacrilege — says he sees that as a design choice. Some shoppers will want full delegation. Others will cap AI at the boring stuff: the matcha reorder, the toilet paper, the annual panic-buy of charging cables.
Still, a few years ago, the idea of a bot quietly digesting your family drama, price sensitivities, and shipping phobias would have sounded like sci-fi. Now it’s another Thursday: You forward a link, your aunt drops a referral code, your AI flags that the “deal” is a rerun of last weekend’s price, and somewhere in the background, an entire industry of retailers, card networks, and platforms is jockeying to be the thing that actually executes your “Yeah, that one.” The emotional work stays analog. The optimization has gone fully synthetic. Aspects of the pink, sparkly Jetsons world are technically plausible. The question is how much of it people actually want.
For now, the Black Friday big-box parking lots will still fill up before dawn. Someone will still camp out for a TV they can now find online in three taps. But somewhere, in a cozy bedroom under piles of blankets, a shopper will roll over, refresh a thread, and watch an assistant report back on prices, stock levels, and stacked rewards in real time.
You’ll still argue about who’s being cheap and who’s being reckless, who deserves the nice headphones, and who is getting socks. But the actual act of holiday shopping is already migrating to the one participant who never sleeps, never blows the budget on a whim, and never has to stand in the cold. The question that’s left is whether you’re still running the holiday operation or just approving what an AI agent already put in the cart.
