There is a line buried in this month’s platform research that deserves to be read twice.
“The creative is the targeting.”
It sounds like a pithy conference slogan. It is actually a precise description of a structural shift in how digital advertising now works. Meta’s Andromeda engine and TikTok’s Symphony algorithm no longer distribute ads based on the audience boxes a marketer ticks. They read the content of the creative itself and decide who sees it. The ad finds its audience based on what it says, not based on who was pre-selected to receive it.
This single change ripples through every other decision in a marketing operation: how much creative to produce, how often to publish, how long to run a campaign, and what skills actually matter in a media buying team. Understanding it is now table stakes for anyone managing paid social.
Here is what this shift means, and what else changed in the past 30 days alongside it.
1. Manual Targeting Is Becoming Obsolete. The Algorithm Reads the Ad, Not the Audience Brief.
Meta’s Andromeda and TikTok’s Symphony represent the most significant shift in paid social mechanics since the introduction of lookalike audiences. Both engines have moved away from granular manual audience filtering toward creative-performance-based distribution. The algorithm evaluates hook rates, hold rates, and engagement signals within the creative, then determines who sees it based on those signals, not based on demographic or interest categories set by a human.
The practical consequence is counter-intuitive. A marketer who spends significant time refining their audience targeting parameters is now doing work that has decreasing marginal value. The algorithm is better at finding the right audience than a human-built segment because it has access to real-time behavioral signals at a scale no targeting brief can replicate.
What matters now is the creative itself. The implication for content volume is demanding. Platforms now recommend 60 to 80 new, conceptually distinct creatives per month to feed the algorithm adequately. Not 60 variations of the same concept. Sixty to 80 genuinely different approaches, different hooks, different frames, different emotional registers, different proof points.
There is also a compounding cost to creative fatigue that most marketers underestimate. Every time a campaign is paused or a winning creative is turned off, the algorithm resets its learning phase. That reset increases costs. The brands that stop and restart campaigns constantly are paying a “learning phase tax” that compounds across the year. Continuous, high-volume creative production is now a performance efficiency decision, not just a brand consistency one.
2. Mukuru Cut Cost Per Install by 89% by Killing Its Targeting. The Data Is Impossible to Ignore.
This case study from African fintech Mukuru is the most instructive proof point for the “creative is the targeting” argument, and it deserves to be read carefully.
Mukuru abandoned hyper-targeted audience filters entirely and let the algorithm run broad. They produced three to four variations of every creator video, changing hooks, frames, and calls to action across each variant. The result: an 89.5% reduction in cost per install and a 290% increase in app installs.
The brand did not improve its product. It did not increase its budget. It stopped over-constraining the algorithm with manual targeting and gave it more creative surface area to work with. The algorithm, freed from narrow audience boxes and fed a diverse creative set, found a dramatically larger and more responsive audience than the human-curated targeting had ever reached.
This is not an isolated experiment. It is a replicable result that is appearing across verticals and platforms as brands move from managed targeting to creative-led distribution. The brands still treating audience segmentation as the primary lever for performance are optimizing the wrong variable.
3. DoorDash’s Chatbot Is Ordering Food Its Customers Have Never Tried. That Is the Agentic Commerce Story in One Statistic.
DoorDash is testing “Ask DoorDash,” a conversational AI ordering chatbot. The early results contain a finding that is startling in its implications: approximately 50% of orders placed through the chatbot were from restaurants the customer had never previously ordered from. Grocery baskets ordered through the bot were more than 35% larger than non-bot orders.
This is the agentic commerce story in miniature. When an AI mediates the discovery and selection process, it does not reproduce the customer’s existing habits. It expands them, based on a richer reading of context, preference signals, and real-time availability than a customer manually browsing a menu would access on their own.
The 35% larger grocery basket is the commercially significant number. Customers spending more per transaction, in a category they were already actively purchasing in, simply because an AI intermediary surfaced options they would not have found themselves. The discovery function that was previously driven by search, browse, and social recommendation is being partially taken over by conversational AI, and the early evidence suggests that AI-mediated discovery actually converts at higher values than human-driven browsing.
For any brand operating in a category with high SKU count, broad catalog depth, or complex product selection, the agentic commerce opportunity is significant. The question is whether the product data, pricing signals, and descriptive content are structured in a way that AI agents can read and recommend effectively.
4. Walmart’s First-Party Data Is Now Targeting YouTube Viewers. Retail Media Just Got a Lot More Interesting.
The integration of Walmart Connect with Google’s Display and Video 360 platform closes a loop that media planners have been trying to close for years. Brands can now use Walmart’s granular first-party purchase data, built from actual transaction history across millions of shoppers, to target those buyers through YouTube campaigns.
The strategic significance is direct. A brand that sells through Walmart can now reach YouTube viewers who have a documented history of purchasing in their product category, from their specific retail context, with video advertising that is directly attributable to in-store and online sales outcomes. That is a closed measurement loop that most digital advertising has historically been unable to provide.
“The closed-loop between social content and physical retail purchase is tightening, making high-fidelity first-party data the ultimate competitive moat.”
This integration is also a signal of direction. The most valuable advertising inventory over the next several years will not be inventory that offers the most reach. It will be inventory that is connected to the most precise purchase intent signals. Walmart’s first-party data is one of the richest purchase intent signals in existence. The fact that it is now available for YouTube targeting changes the value equation for connected TV and online video advertising in ways that most media plans have not yet absorbed.
5. 20% of Consumer Conversations With AI Already Have Direct Shopping Intent. The Discovery Layer Is Shifting.
Research from Mizuho Securities quantifies something that marketers have been sensing anecdotally: approximately 20% of consumer conversations with AI assistants now demonstrate direct shopping intent across categories including beauty, electronics, and travel. Zalando and ASOS are already piloting ChatGPT-driven shopping journeys to meet customers with personalized inspiration at the moment they are actively planning a purchase.
The implication is structural, not tactical. For two decades, the brand that won discovery won the category. Discovery happened in search results, social feeds, and editorial content. AI is now inserting itself into discovery as the primary mediator, and the brands that are visible and credible within AI-generated recommendations are capturing intent before it reaches a search bar or a product page.
GenAI referral traffic currently represents only 0.37% of total web traffic, but forecasts suggest it will reach 0.50% within two months. That rate of acceleration, doubling its share in 60 days, is the signal worth paying attention to. The absolute numbers are still small. The trajectory is steep.
The brands investing now in ensuring their content is machine-discoverable, structurally clear, authoritative, and citation-worthy within large language model responses are buying a position in a channel that will be crowded and expensive to enter once its scale becomes undeniable.
6. Finance Apps Grew Mobile MAUs 43% While Web Traffic Declined 10%. The App Habit Is a Moat.
Here is a finding that cuts against the prevailing narrative about AI disintermediation of brand-owned channels. RBC Capital Market Research data shows that while web traffic for credit monitoring and personal finance has declined 10% over the past three years, mobile app monthly active users for the same category have surged 43% in the same period.
The reason is important. Consumers trust authenticated, personalized, proprietary data access in a way they do not trust AI-synthesized answers. FICO and Experian users open their apps because they want their specific score, their actual credit report, their personal financial data. An AI assistant cannot provide that with the same authority and specificity. The app is the moat.
This generalizes beyond financial services. In any category where the value proposition is built on personal data, specialized expertise, or proprietary information, a well-designed mobile app experience can resist AI disintermediation better than an equivalent web experience. The brands that have invested in genuine app utility, not just mobile-optimized web interfaces, are building a channel that AI answer engines cannot easily replace.
7. Short Drama Apps Are in the Worldwide Top 40. Brands Have Not Found the Format Yet.
Apps like DramaBox and ReelShort have reached worldwide Top 40 download rankings on the strength of serialized, high-emotion, cliffhanger-driven short drama content. These are episodic micro-dramas, each episode running two to five minutes, designed to be consumed in sessions, not in isolated singletons.
The engagement mechanics are well understood in long-form television and publishing. The cliffhanger creates investment. The episode structure creates habit. The serialized format creates loyalty. What is new is the delivery: mobile-native, short enough for commutes and waiting rooms, and organized around emotional intensity rather than production quality.
TikTok’s own “Minis” ecosystem, including micro-dramas and in-app mini-games, is moving in the same direction. Platform-native episodic content is creating session depth and ad inventory that pure short-form content does not.
For brands, the format represents an emerging opportunity that has not yet been competitively crowded. Sponsoring episodic mini-dramas, integrating into serialized creator content, or developing brand-adjacent micro-drama content of their own are all early-stage tactics with limited competition and high audience engagement. The window for first-mover advantage in this format is narrowing as more brands recognize what the download rankings are signaling.
The Pattern Across All of It
The thread running through this month’s intelligence is a single, clarifying shift in how attention is allocated and how brands earn a claim on it.
Attention has become the most expensive marketing currency. And the methods for earning it are changing at every level simultaneously: algorithms now read creative content rather than audience briefs; AI agents mediate discovery before consumers reach a search bar; apps hold users in authenticated environments that resist AI disintermediation; and serialized short drama content is capturing session-level engagement that isolated content pieces cannot.
The brands adapting fastest are not the ones with the largest budgets. They are the ones that have accepted that the campaign mindset, discrete bursts of spend followed by measurement and reset, is being replaced by an interaction management mindset, where AI handles the continuous evaluation of behavioral signals and the brand’s job is to maintain a rich, diverse, machine-readable creative and content ecosystem at all times.
– Manpreet Jassal
Leave a Reply