Black Friday keeps breaking online sales records, but it is also getting blurrier: Adobe estimates roughly $10.8B in U.S. online Black Friday spend in 2024, up from about $9.8B in 2023, while promotions now start weeks early, run through Cyber Monday, and blend into endless "holiday offers" [1, 2]. Brands discount harder, ad auctions get more expensive, and it is not obvious who really wins beyond the platforms selling the clicks [3, 4].

At the same time, how people shop is changing. Google's holiday research finds that roughly three-quarters of shoppers now plan their holiday shopping ahead of time, and most purchases are researched rather than spontaneous, behavior that aligns with the shift from generic "deals" browsing to longer, more specific queries [5, 6, 7]. Many shoppers show up already knowing which brand or even which model they want; Black Friday is simply the moment they pull the trigger.

The next step is obvious: instead of humans hunting for deals, AI agents will do that work.

In this paper, I argue that once agents buy on our behalf, Black Friday shifts from a massive human shopping frenzy to an agentic one - the moment when AI systems run their biggest price checks and batch-buy from the wishlists they've been managing all year. In that world, the game for brands moves from bidding on keywords to influencing prompts, shaping agent rankings, and proving whether those interventions actually drive incremental sales.

When Software Starts Shopping for You

Now imagine that instead of doing all this work yourself, you delegate it to an AI agent.

You tell it what you care about: budget, preferred brands, styles you like, what has broken on you in the past, how much hassle you are willing to tolerate with returns. You connect it to your email, your purchase history, your loyalty accounts. You give it access to your calendar and your home address. Over time it learns that you travel frequently, that you dislike dealing with warranty claims, and that you are willing to pay a bit more to avoid unreliable sellers.

This agent can track far more than you ever could. You might track five brands; it can track five thousand. You might compare three retailers; it can check thirty. You might vaguely remember prices from last month; it can evaluate full price histories and detect whether a "40% off" sale is genuine or marketing theater.

Figure 1: U.S. Black Friday online sales (billion USD) and Meta CPMs indexed to 2015 (2015 = 100), 2015–2024.

Notes: Sales estimates based on Adobe Black Friday reporting [31, 29, 27, 25, 24, 23, 21, 17]. CPM benchmarks compiled from Meta/Facebook ad cost studies and benchmarks [8, 9, 10, 11, 12, 13, 14, 15, 16].

In that world, Black Friday is not your personal war room. It becomes one of the moments in the year when your agent expects substantial repricing and recalculates whether any items on your long-term wishlist have dipped below the thresholds you have set, explicitly or implicitly.

Instead of you doom-scrolling "today only" emails at midnight, your agent runs something like:

"Check all televisions on my list; if any high-trust retailer is offering at least a real 25% discount relative to the last six months, and delivery fits my schedule, go ahead and buy. If not, wait."

The surface behavior - orders clustering in late November - might look similar. But the underlying mechanism is different: not fear of missing out, but automated execution of a policy.

Winning the Prompt, Not the Keyword

If agents sit in the middle, the old search-ad game is over. Today, brands pour money into auctions around phrases like "Black Friday deals" or "best headphones 2025," fighting to appear when a user types a query. An agent collapses much of that into a single instruction - "Find me the best deal on this specific Samsung model," "Find shirts like AllSaints but under $100," "Upgrade my work wardrobe with three new outfits that fit these constraints." The human may never see a page of search results, only a curated shortlist or a decision the agent already made.

In that world, three things matter more than whether you can afford the "Black Friday TV deals" auction:

1. Users must explicitly name your brand in the prompt.

2. Your products must live inside the agent's evaluable universe: structured feeds, accurate stock data, clean attributes, and reliable shipping.

3. You must score well on the dimensions the agent actually optimizes for: price, reliability, reviews, return rates, sustainability, and user-specific learned preferences.

Marketing does not ask "How do I win the keyword auction?" anymore; it asks "How do I get into the prompt and into the agent's top recommendations?" Black Friday becomes the moment agents run massive comparison jobs, not a moment of heightened human impulse.

Figure 2: Human shopping path vs. agentic path; where brands must capture mindshare.

Measurement When Agents Click

In an agent-mediated world, measurement systems cannot stay the same. Deterministic, user-level attribution - "this impression led to this click led to this conversion" - was already fragile in a privacy-constrained, cross-device world. With agents mediating decisions, it becomes fiction. The agent consumes information from dozens of sources, updates internal models, and triggers purchases in ways that do not map to single-channel attribution paths.

Instead of chasing perfect user-level traces, brands have to rely on probabilistic tools: marketing mix models that treat agent-mediated traffic as a channel, and incrementality tests - geo experiments, audience holdouts - that answer questions such as: "If we invest more in top-of-funnel storytelling or improve product feed quality, do we see more agent-driven sales?" AI makes it easier than ever to run these models and harder than ever to pretend they are optional.

In an agent-mediated world, Black Friday gets bigger, not smaller. If agents maintain wishlists year-round and wait for high-signal deal windows, late November becomes a high-volume execution window. The winners shift: shallow brands, messy data and fake discounts lose; strong brands with clean feeds and authentic promotions gain.

A competitive version of this future brings "calendar wars" between ecosystems: Black Friday competing with Prime Day, Singles' Day, and agent-native events such as an "Agent Shopping Week," where assistants refresh rankings and move preferred partners. Consumers still experience a major event; brands face recurring checkpoints where both humans and agents re-evaluate them.

What Brands Should Do

Black Friday is becoming a human-plus-agent shopping event, with AI systems doing most of the comparison and an increasing share of the execution. That reality demands a different playbook.

First, you have to play for the prompt, and brand memory determines whether the user says "Get me AllSaints" or "Get me a shirt." Being named is the first filter, and it sets your starting position when an agent is tasked with finding deals.

Second, you must make yourself agent-friendly: Clean product data, honest pricing, reliable fulfillment, and strong post-purchase experiences are now marketing inputs, because agents optimize for them. A retailer with slightly lower discounts but far fewer returns is the rational choice for an agent acting in the long-term interest of the user.

Third, you need to redefine what Black Friday is for. If agents maintain wishlists and price histories all year, the Black Friday window is the moment they execute a large share of those decisions at once. That pushes the event up in volume, not down - but only for brands that are already present in people's prompts and rank highly in agents' internal lists. The work happens months earlier: building memory, cleaning feeds, and shaping pricing so that when the cultural moment arrives and everyone's agents are checking at once, you are the obvious choice.

Black Friday is not going away. As agents take over more of the comparison and execution, the event becomes bigger in absolute volume: the same cultural moment, but with software lining up wishlists and firing off purchases far more efficiently than humans ever could. From the outside it still looks like a major shopping day; under the surface it is driven less by last-minute browsing and more by systematic agent policies that balance what people value, what brands can profitably offer, and how the systems in the middle choose to rank options.

Brands that win this next phase are the ones that show up in prompts, dominate agent rankings, and prove - through rigorous measurement - that the investments they make to earn that position pay off.

References

[1] Adobe-based Black Friday 2023 online sales estimate. https://pacvue.com/blog/insights-from-black-friday-2023/

[2] Black Friday 2024 analysis: online sales, retail-media growth, CPC shifts. https://skai.io/blog/black-friday-2024-analysis/

[3] RetailDive: Holiday outlook and margins. https://www.retaildive.com/news/winners-losers-black-friday-2023/700647/

[4] Ipsos: Pull-forward and margin-risk analysis. https://www.ipsos.com/sites/default/files/ct/publication/documents/2018-11/black-friday-2018.pdf

[5] Cyber Week 2025 consumer study. https://impact.com/partnerships/black-friday-consumer-insights/

[6] Black Week / Black November overview. https://www.artefact.com/blog/last-minute-black-friday-black-week-or-better-yet-early-black-november-planning/

[7] Analyst commentary on "Black November." https://newschannel9.com/news/local/analyst-black-friday-evolves-into-black-november-as-early-deals-redefine-shopping

[8] ADCostly. The Latest Facebook Advertising Cost by Category [2021 Update]. 2021.

[9] Triple Whale. Black Friday Facebook Ads: Data, Strategies, and AI Agents for 2025. 2024.

[10] Tinuiti. Q4 2022 Digital Ads Benchmark Report. 2023.

[11] Andy Taylor and Mark Ballard. Cyber Five 2022: Black Friday and Cyber Monday Ad Trends Across Google, Amazon and Meta. 2022.

[12] Tinuiti. Cyber Five 2023: Black Friday and Cyber Monday Ad Trends Across Google, Amazon and Meta. 2023.

[13] KlientBoost. Key To Reducing Facebook Ad Cost. 2021.

[14] Starling Social. Facebook Ads in 2022: How to Lower Your Ad Costs. 2022.

[15] Business of Apps. Facebook Ads Cost (2025). 2025.

[16] AdStage. Facebook Ads Benchmark Report Q4 2017. 2018.

[17] Adobe. Cyber Monday Hits Record $13.3 Billion in Online Spending. Media alert, December 2024.

[18] Adobe. Holiday Shopping Season Drove a Record $241.4 Billion in Online Spend. Full-season recap, January 2025.

[19] Adobe. Adobe Forecasts Record $240.8 Billion U.S. Holiday Season Online. Holiday forecast, September 2024.

[20] Adobe. 2024 Holiday Season Set New Records for Ecommerce. Season wrap-up blog.

[21] Adobe. Media Alert: Adobe Cyber Monday Surges to $12.4 Billion in Online Spending, Breaking E-Commerce Record. Media alert, November 2023.

[22] Adobe. Media Alert: Adobe Holiday Shopping Season Drove a Record $222.1 Billion Online. Season recap, early 2024.

[23] Adobe. Adobe: Cyber Monday Drove $11.3 Billion in Online Spending, Breaking E-Commerce Record. Press release, November 2022.

[24] Adobe. A Record-Breaking Cyber Week 2020: Online Shopping and Key Trends. Cyber Week 2020 recap blog.

[25] Adobe. Adobe Data Shows Record Cyber Monday With $9.2 Billion in Online Sales. Cyber Monday 2019 data blog.

[26] Adobe. Retailers With Buy Online, Pick Up In-Store Fulfillment Option. Holiday 2019 BOPIS growth blog.

[27] Adobe. Adobe Digital Insights Holiday Recap 2018. Adobe Digital Insights slide deck.

[28] Adobe. Adobe Data Shows Cyber Monday as the Largest Online Sales Day in History. Experience Cloud press release on Cyber Monday 2017.

[29] Adobe. 2016 Holiday Shopping: Up-to-the-Minute Data From ADI. Holiday 2016 data blog.

[30] Adobe. ADI Holiday 2016 Recap Report. Holiday recap blog.

[31] Adobe. 2015 Holiday Shopping Predictions. Adobe Digital Index press release (PDF).

[32] Adobe. ADI Holiday 2015 Report. Holiday 2015 infographic from Adobe Digital Index.

[33] Adobe. Adobe Forecasts Record $207 Billion Holiday Season Online in the U.S. and $910 Billion Globally. Holiday forecast blog, October 2021.