AI AS A SHORTCUT OR A SPYGLASS? TURNING D&AD’S AI FEARS INTO A COMPETITIVE ADVANTAGE FOR MEDIA BUYERS

AI in advertising is being merchandised like a miracle gadget: push a button, get “more creative, faster.” Need 200 headline variations, 50 image crops, 12 cutdowns? Generative tools will happily flood your asset library in minutes. And that’s precisely why most brands are sleepwalking into the same trap: they’re using AI to make more ads, not better decisions.

While D&AD juries and creative directors publicly agonise over whether machines will replace human imagination, the real power shift is happening somewhere far less glamorous: inside the auction. The most sophisticated media buyers aren’t asking AI to come up with the next big idea; they’re using it as a spyglass trained on the marketplace itself, weaponising performance and pricing data to out-buy competitors long before anyone notices on a Cannes reel.

Look at what’s actually visible if you know where to point that spyglass. In social and programmatic auctions, the most valuable competitive signals are buried in media allocation choices, efficiency trends, and subtle shifts in placement and geography. A rival’s CPM quietly drops. Another suddenly leans into a new format. A third concentrates spend in one region instead of three. On their own, these are just curiosities. Interpreted together, they’re a live feed of your competitors’ strategy. As one analysis of social ad auctions explains, the real edge isn’t spotting who spends the most; it’s understanding why they’re winning before the rest of the market does, by decoding patterns in media efficiency, channel mix, and pacing.

That is where AI stops being a shortcut and starts becoming a competitive intelligence layer. Platforms like Polaris AI aren’t merely scraping ads; they’re contextualising live auction data across social and the open web, surfacing real-time indicators like CTR, CPM, share of voice, and spend efficiency, then tying them back to creative, placements, and audience hypotheses. When that kind of system flags that one insurer isn’t just outspending rivals but outbuying them with lower acquisition costs, the takeaway isn’t “we need cooler scripts.” It’s “they’ve built a more precise, diversified media machine—copy the system, not the slogan.”

At the same time, the auction itself is changing under your feet. As Google, Meta, and TikTok automate targeting through Performance Max, Advantage+ and automated expansion, audience knobs are disappearing. The algorithms decide who sees your ads; your main lever is the signal you feed them. That’s turning creative into a form of targeting in its own right. According to one analysis of AI-native campaigns, broad targeting now means your headlines, visuals, and CTAs are the strongest clues platforms use to decide who should see which ad, effectively making creative “the new targeting” signal.

In that world, AI-generated volume is table stakes. The real advantage comes from closing the loop between what you and your competitors put into the system (creative, bids, formats, budgets) and what the system gives back (who wins which impression, at what price, with what downstream performance). Leading advertisers are already using AI to run continuous creative optimisation loops—hundreds of variations launched, measured, and evolved automatically based on engagement and conversion data, not gut feel. As one overview of AI-native advertising notes, speed isn’t just efficiency; the ability to test, learn, and redeploy at scale becomes a structural edge.

Crucially, buyers themselves don’t want to hand over the wheel. Research into streaming TV shows that nearly half of media buyers see AI’s biggest upside in planning and optimisation, but only about one in five are comfortable with fully autonomous campaigns. They want AI that augments judgment—tools that surface signals, automate the grunt work, and make it easier to adjust strategy on the fly—rather than a black box that “does it all” while they hope for the best, as recent findings on AI’s role in TV buying and execution make clear.

So while the industry debates whether AI-made ads are “real” creativity, the sharpest media teams are using AI for something far more pragmatic: seeing the auction more clearly than anyone else, learning faster from every impression, and turning that intelligence into a compounding performance gap across search, social, programmatic, and streaming. The question isn’t whether AI will replace creatives. It’s whether your media buying team will be the ones quietly turning AI into a spyglass—while your competitors are still playing with the shiny shortcuts.

D&AD’s discomfort with AI is understandable, but it’s pointed at the wrong part of the problem.

The awards discourse treats AI as a creative shortcut: a way to crank out more concepts with less human effort. That lens makes sense if your world is scripts, storyboards, and key visuals. If AI is just a faster Photoshop, then yes, hand‑crafted work looks threatened and “cheating” becomes the core anxiety.

But that’s not where the real disruption is happening.

In media, AI isn’t just a content factory — it’s becoming a decision engine. Autonomous systems are learning to allocate budgets, pick placements, and prioritize audiences on the fly, based on live performance signals at a scale no human team can match. As illumin’s overview of AI in AdTech notes, marketers can now hand an AI a business goal (say, ROAS or incremental reach) and let it run thousands of micro‑optimizations in real time across bids, audiences, and inventory.

That’s not a shortcut; it’s a structural shift in how advantage is created.

The D&AD framing assumes that creativity is still the primary battlefield — your main risk is generic, samey work. For media buyers, the battlefield has moved. The biggest competitive gaps aren’t in ideas; they’re in how intelligently those ideas are distributed when attention is scarce and machines are intermediating every decision.

You can already see this in the auction data. Competitive intelligence platforms like Polaris AI, profiled in AdExchanger’s analysis of social ad auctions, don’t just scrape who’s running what ad. They infer strategy from a web of signals: falling CPMs, sudden shifts into specific placements, geographic concentration, and cross‑channel coordination. When Progressive shows lower acquisition costs across social and programmatic, Polaris’ AI doesn’t celebrate “more creative, faster”; it hypothesizes a deeper media system — tighter audience precision, diversified placements, and smarter allocation.

That’s the spyglass D&AD is missing. AI is not only making more assets; it’s exposing, in near real time, which brands are outbuying their category and why.

The same misalignment shows up in how the industry thinks consumers perceive AI. In TV and CTV, buyers and sellers are still skittish about machine‑made ads, assuming viewers will reject them as inauthentic. Yet research discussed in VideoWeek’s coverage of AI in TV advertising found that only about 10% of viewers say they dislike AI‑generated creative, while half of buyers and roughly 40% of sellers assume viewers do. In reality, audiences are far more open to personalized, multi‑version ads than the supply side believes.

That gap is profound. While awards juries quibble over whether an execution is “AI‑assisted,” actual viewers are quietly rewarding relevance and personalization — exactly the things AI is best at scaling. The fear of creative dilution is blinding many teams to the real opportunity: using AI to make the viewing experience better, not just the production process cheaper.

Meanwhile, the canvas itself is shrinking. AI‑driven interfaces from Google, Microsoft, and others are compressing discovery into a handful of options. As Search Engine Journal’s exploration of “agentic commerce” explains, AI Overviews and AI Modes are reducing available impressions, and the fastest‑growing “users” of the web are automated agents, not humans. These agents don’t browse or respond to clever art direction; they construct shortlists of three to five options and execute.

In that “shortlist economy,” the question isn’t whether your creative was human‑crafted enough to impress a jury. It’s whether your brand even makes it onto the agent’s list. You can already see this logic in Google’s AI performance insights in Merchant Center, which show share of voice on AI surfaces against similar brands — effectively a rank report for how often machines choose you.

Layer on the rise of agentic media buying, where autonomous systems continuously reallocate budgets and test creative variants without human prompts, and the D&AD narrative starts to look small. As MarTech’s guidance on AI‑native advertising stresses, execution is becoming a commodity; the differentiator shifts upstream to sharper strategy, clearer messaging, and smarter mapping of content to real‑time intent.

Which is precisely why treating AI as a creative shortcut is a competitive liability for media buyers.

If you see AI as a faster way to fill your asset library, you end up competing on volume in an environment where impressions are shrinking and agents filter ruthlessly. If you treat AI as a spyglass into auctions, intent signals, and competitive behavior, you can rewire how you plan, measure, and invest — and build an edge long before your rivals realize the battleground has moved.

Impressions used to be a volume game. If you could afford to carpet‑bomb search, social, and display, you’d probably get enough bites to make the spreadsheet work. That logic breaks in an environment where there are simply fewer places for your ads to show up, and more AI layers deciding what makes the cut before a human ever sees anything.

You can already see the floor dropping. Google’s AI Mode and AI Overviews have compressed the old SERP into a smaller, curated surface where organic results, shopping units, and ads all fight for a sliver of space. One performance team documented an 11% year‑over‑year decline in impressions as AI surfaces expanded, noting that “the canvas is smaller, and the bar is higher” for any ad that appears inside it, as Search Engine Journal reported. You’re not just bidding for an impression anymore; you’re bidding for a cameo in a machine‑curated answer box.

At the same time, the audience on the other side of that box is mutating. Microsoft’s framing of the web as coexisting eras — “help me find it,” “help me choose,” and “do it for me” — is already visible in the data. Automated, agent‑driven sessions are growing about eight times faster than human traffic, according to the same Search Engine Journal analysis. Those agents don’t scroll a category page or browse a carousel. They evaluate options, generate a shortlist of three to five brands, and execute.

Retail media strategist Roger Dunn’s phrase for this, the “shortlist economy,” is the real story underneath the impression squeeze. When a shopper asks Gemini, Copilot, or ChatGPT for “the best running shoes for flat feet” and gets four product suggestions, that list is now the entire consideration set. A December 2025 Semrush survey found that 43% of U.S. shoppers had discovered a new brand via AI, and nearly half say they notice AI‑mentioned brands “often” or “very often,” as summarized by Search Engine Journal. If you’re not on that shortlist, your beautiful film, your perfect brand platform, and your media plan never even enter the room.

For media buyers, this is where guesswork becomes fatal. In a world of abundant impressions, you could afford to be directionally right: broad match here, similar audiences there, a few creative variants, and the platform’s cheap reach would paper over your mistakes. But as AI consolidates exposure into fewer, higher‑intent moments, the cost of being wrong is multiplied. If your signals are fuzzy — mismatched creative, generic landing pages, undifferentiated offers — you’re not just underperforming; you’re systematically training AI agents to ignore you.

Meanwhile, the buying machinery on the other side is getting sharper. Modern AdTech systems don’t wait for your team’s weekly optimization meeting. They evaluate millions of impressions in real time, predicting which users — or agents — are most likely to convert and adjusting bids in milliseconds, as illumin describes in its overview of AI‑powered campaign management. These systems are already practicing “agentic advertising,” setting bids, pacing budgets, and reallocating spend based on probabilistic models you never see. The platforms are no longer passive pipes; they’re active negotiators deciding whether your ad merits a place in the next micro‑shortlist of the auction.

Overlay that with another structural shift: as Google, Meta, and TikTok push broader, AI‑driven targeting modes, the audience filters you used to lean on are dissolving. Performance Max, Advantage+ Shopping, and automated audience expansion are designed to expose your ads to wide swaths of users and let machine learning refine who converts. In that context, creative becomes the primary signal that tells the algorithm who an ad is “for,” a point that MarTech captures in arguing that “creative is the new targeting.” Headlines, hooks, visuals, and offers aren’t just persuasion anymore; they’re data labels for the auction.

Put these dynamics together and the old habit of “launch, watch, and hope” turns into a liability. You can’t spray 50 AI‑generated variants into the ether and wait for a winner if each visible impression is now a rarified, AI‑filtered event and each auction is a battle between autonomous bidding agents. Media buyers who still treat AI as a production shortcut — a way to crank out more copy or more cutdowns — are optimising the wrong bottleneck.

The competitive edge goes to the teams who accept the impression squeeze as a given and design for the shortlist economy on purpose: feeding AI systems clean, intentional signals; building creatives that qualify the right audience; and measuring not just “did we get seen?” but “did we get selected by the agent doing the choosing?” In that world, every guess you leave untested is an invitation for someone else’s model to push you off the list.

Inside the auction, AI is not a mood board generator – it’s a pattern engine. Every time your ad enters a social, search, or open‑web auction, it leaves a faint strategic fingerprint: what you were willing to pay, where you chose to show up, which audiences you leaned into, how quickly you pulled back when performance dipped. Multiply that across thousands of auctions a day – plus your competitors’ – and you have a live, continuously updating map of the category’s real strategy, not the version that shows up on stage at an awards show.

The catch is that those signals are almost impossible to read with human eyes alone. A rival’s CPM dropping 18% on Meta while their impression share climbs on YouTube is just noise in a spreadsheet. At scale, though, those shifts add up to what one competitive intelligence platform calls “the most valuable signals in modern advertising” – the media allocation decisions, efficiency trends, and placement patterns that only appear first in the auction, not in case studies or QBRs, as described in an analysis of competitive signals hiding inside social ad auctions.

This is where AI stops being a shortcut and starts acting like a spyglass.

Modern media‑side systems work much more like the “agentic” optimizers that AdTech analysts have been tracking: models that scan millions of impressions, predict the likelihood of a conversion, and decide in milliseconds whether to bid – and how much – against a specific user, placement, and moment in time. Instead of working off if/then rules, these systems ingest contextual signals, performance history, and auction dynamics to make thousands of micro‑decisions per day, as outlined in illumin’s overview of.

When you point that same kind of intelligence at competitive data, the auction turns into a living storyboard of everyone else’s strategy.

A few concrete examples:

• A competitor’s CPM falls, but their average position and click‑through rate improve. AI can correlate that with creative shifts, new formats, and audience segments, then surface a likely explanation: they’ve found a more efficient placement mix or unlocked a new look‑a‑like model. According to the breakdown of how platforms like Polaris AI work, this is less about counting ads and more about turning those auction deltas into hypotheses about audience precision and diversified placement.

• Another brand abruptly redirects budget into a new geography and a specific inventory type – say, vertical video in connected TV apps that support programmatic. AI can spot the timing (right after a product launch), the environment (premium AV), and the efficiency trend (CPM rising, but ROAS holding) and flag that they’re deliberately paying a premium to dominate a strategic context.
• A third player starts losing auctions they used to win, without cutting spend. By comparing impression share, cost curves, and platform changes, an AI system can infer that they’re being outbid by a new entrant or squeezed by platform‑side changes like “AI Mode” and conversational surfaces that reduce overall impression supply.

None of these stories show up in a press release. They’re only visible if you treat auctions as a source of truth for what brands are actually doing – and give an AI system permission to read between the lines.

For media buyers, the payoff is twofold.

First, you gain a real‑time view of the competitive field that isn’t constrained by channel silos. When the same signals – a rival’s improved CPM efficiency, a surge in short‑form placements, a pullback from broad match – appear simultaneously across social, search, and programmatic, AI can connect them into a single narrative about how that brand is preparing for the shortlist economy. That’s exactly the pattern the Polaris analysis surfaced when it concluded that Progressive’s advantage was less about raw budget and more about a coordinated media system operating across channels.

Second, you can turn those insights into concrete actions in your own buying stack. The same agentic engines that optimize your bids can ingest competitive signals as constraints and goals: “Avoid overbidding where Brand X is clearly overspending for vanity placement,” or “Test into the mid‑funnel placements Brand Y just scaled, but cap CPMs at the level where their efficiency started to decay.” In AI‑native environments where the “recommendation itself becomes the ad,” as one analysis of conversational AI advertising put it, that kind of adaptive strategy is what keeps you visible when agents are compiling their three‑to‑five‑option shortlists.

The D&AD fear is that AI lets everyone copy faster. Inside the auction, the reality is harsher and more useful: AI lets you see who is actually winning, why they’re winning, and how to respond before that advantage shows up in your quarterly numbers – or in someone else’s acceptance speech.

Award‑winning campaigns are terrible teachers if you only study the output. The film case study, the three‑minute director’s cut, the sizzle reel at D&AD — those are the epilogue. The strategic story that beat you lives upstream, inside bidding patterns, audience choices, creative rotation, and retreat points.

This is where AI stops being a shortcut and becomes a spyglass. Instead of trying to out‑guess “the big idea,” you out‑iterate the system that delivered it.

Modern auctions already run this way. AI is taking over the tactical work of media buying — evaluating millions of impressions, predicting which ones are likely to convert, and adjusting bids in real time to hit a goal ROAS or CPA, as platforms like illumin describe. You’re not going to out‑click a machine at millisecond bid decisions. But you can absolutely out‑learn the brands you’re competing with if you wire your own AI to watch what the auction is telling you.

Think of every impression as a data point in a live‑action strategy lab. Competitive intelligence tools like Polaris AI already show how this works at scale: instead of scraping ads for headlines and colors, they track shifts in CPMs, placement mix, geography, and spend efficiency across social and the open web, then translate that into hypotheses about why a player is winning. When Progressive’s acquisition costs come in consistently lower even as spend rises, the system doesn’t just applaud the creative — it infers a structure underneath it: tighter audience precision, diversified placements, and disciplined reallocation when marginal returns soften.

That is the mindset you need to beat award‑winners: stop treating creative as a talisman and start treating it as evidence.

Concretely, that means pairing AI with “ad spying” in three layers:

1. Strategic fingerprinting. Use AI to mine auction‑level signals the way Polaris AI does: who is suddenly paying more for a previously cheap audience, who just consolidated into Reels or CTV, who pulled back from a geography after two weeks. You’re not copying the ad — you’re mapping the bet.

2. Hypothesis generation. When your tool flags that a competitor’s CPM drops 18% the same week they start running modular, product‑led creative, that’s a test brief, not a trivia fact. Your AI should translate raw signals into “why” statements: “They likely unlocked cheaper reach with broader interest targeting plus tighter creative relevance” — then turn those into experiments queued in your media plan.

3. Rapid test‑and‑tilt. In an environment where AI is already compressing inventory and pushing us into a shortlist economy, the winner isn’t the brand with the bravest single idea; it’s the one that can cycle through 50 small ideas while everyone else is polishing case‑study number one. You define the sandbox (guardrails on brand, offer, and risk), then let AI propose and prioritize micro‑tests — creative variants, audience pivots, frequency caps — and scale only what clears your performance bar.

You also have more freedom to do this than most buyers think. While marketers are comfortable using AI for “safe” tasks like content generation and data analysis, they remain cautious about handing over full control of media buying, as a Digiday survey reported. That hesitation is a competitive opening. You don’t need to flip a switch to autonomous spend; you need to be the team that uses assistive AI to run ten times as many controlled experiments as everyone else — with human approval on what actually rolls out.

And audiences are far less allergic to AI‑shaped work than the industry fears. Research discussed in a recent VideoWeek interview found only about 10% of viewers actively dislike AI‑generated creatives, while roughly half of buyers assume viewers do. That gap is pure margin for the media buyer who uses AI to generate and optimize multi‑version ads without getting squeamish. If the viewer is happy to see a more relevant spot, and the auction rewards you with better engagement and lower CPMs, there’s no virtue in leaving that efficiency on the table because the creative process didn’t feel artisanal enough.

Out‑iterating doesn’t mean you stop caring about the idea. It means you treat the idea as the opening bid, not the finished sculpture. You launch faster with “good enough,” wire your stack so AI can read competitive moves in near‑real time, and then you learn in public — with 100 controlled micro‑bets instead of one immaculate, fragile campaign.

Awards recognize what worked last year. In an AI‑driven auction, your advantage comes from something no jury can see: how many disciplined iterations you were able to squeeze in before anyone else realized the game had changed.

AI isn’t here to take the media buyer’s job. It’s here to take the parts of the job that never should have been yours in the first place.

Bid tweaks in 15‑minute blocks. Spreadsheet gymnastics to reallocate 3% of budget from one ad set to another. Late‑night Slack debates about whether to raise a cap by $0.10. Those were artifacts of a world where automation could only follow rules you pre‑wrote.

In an agentic era, that work is simply better done by machines.

Modern buying systems already act more like autonomous co‑pilots than glorified macros. You set a goal — ROAS target, marginal CAC ceiling, share‑of‑voice threshold — and the platform optimizes toward it, learning from performance and adjusting thousands of times a day in real time, as described in illumin’s overview of AI‑powered campaign management. Algorithms now evaluate millions of impressions, predict the lowest viable bid, and reflow spend across placements and audiences faster than any human team could. The emerging “agentic advertising economy,” as McKinsey and others frame it, is one where AI systems execute the bulk of optimization autonomously.

That doesn’t make the media buyer obsolete. It makes the old version of the role — the button‑pusher — obsolete.

The new leverage point is what you ask the agents to do, where you unleash them, and how you validate that their actions are actually aligned with business reality. As one activation lead notes in AdExchanger’s coverage of Meta’s AI tools, humans increasingly have to be “decision makers and managers,” sanity‑checking agent output and interrogating whether a spike is a genuine win or a measurement glitch. The machine can test 10,000 micro‑moves; only a strategist can decide which of those moves fit the brand, the margins, and the board deck.

In other words, the media buyer becomes a market architect.

Instead of living inside Ads Manager, you’re designing the system the agents operate within:

• Defining the objective hierarchy. Is the real constraint margin, cash flow, LTV/CAC, or offline capacity? Agentic platforms are very good at optimizing the metric you feed them — even if it’s the wrong one. Your job is to translate messy business reality into clean optimization targets that won’t create perverse incentives.
• Designing inventory and supply strategy. On the sell side, publishers are using AI to re‑package, classify, and price supply more intelligently, focusing less on “agent‑to‑agent” automation and more on maximizing inventory value, according to VideoWeek’s analysis of TV sellers. On the buy side, the mirror image of that work is deciding which supply paths, contexts, and environments your agents are even allowed to touch — and how hard they’re allowed to lean into each.

• Orchestrating channels around the shortlist. As AI agents compress consumer consideration into three to five options, we move into what one strategist calls the “shortlist economy”. Impression volume matters less than being one of the few brands surfaced by assistants and AI Overviews. Media buyers shift from “how many impressions can we buy?” to “which surfaces most influence the agents that build the shortlist, and how do we earn disproportionate presence there?”
• Instrumenting competitive intelligence. Platforms like Polaris AI already read auction‑level signals — CPM shifts, placement mixes, geographic concentration — and turn them into hypotheses about why a competitor is outbuying the market. The architect’s role is to wire those insights directly into your own agents’ playbooks: where to relax constraints, where to tighten, where to test a counter‑move before the rest of the category reacts.

Underneath all this sits a more uncomfortable truth: AI creative and optimization will routinely go off the rails if left unchecked. Meta’s much‑publicized “slop” incidents — distorted products, mismatched casting, surreal details — are a preview of what happens when platforms push generative outputs at scale and then remind brands that, per their own terms of service, “AI can make mistakes and it is the advertiser’s responsibility to review the AI outputs.” The agent will happily ship something that technically hits response goals while quietly corroding brand trust.

So the architect doesn’t just set goals and budgets. They set guardrails:

• Brand and safety constraints the agents can’t violate, even in the name of performance.
• Feedback loops that flag when auction‑side wins are creating downstream problems — returns, complaints, channel conflict — that the algorithm can’t see.
• Governance over which AI creative tools are allowed to run live, on which accounts, under what approval workflows.

As AI takes over optimization, the value of the media buyer migrates upstream and outward: from tactical controls to market design, from platform wizardry to cross‑channel architecture, from “What bid should I set?” to “What game are we even playing — and how do we rig the rules in our favor?”