We’re having the wrong conversation about AI in advertising.
While the industry obsesses over algorithmic bias and data privacy-important issues, certainly-we’re ignoring a more insidious ethical crisis unfolding in real-time: the systematic dismantling of creative accountability.
Every major ad platform now uses AI to optimize, test, and even generate creative. At Sagum, we’ve spent over $2 million on TikTok alone in the past year, and I can tell you firsthand: AI isn’t just changing how we advertise-it’s fundamentally altering who is responsible when advertising causes harm.
And nobody seems to notice.
When Everyone and No One Is Responsible
Here’s the uncomfortable truth: When an AI system serves a predatory loan ad to a financially vulnerable teenager, optimizes diet pill creative to target users with eating disorder signals, or maximizes engagement by exploiting someone’s mental health crisis-who exactly is accountable?
- The advertiser? They didn’t write the targeting parameters; the AI learned them.
- The platform? They’re just providing tools; they don’t control how they’re used.
- The agency? We’re optimizing for the KPIs the client requested.
- The AI? It’s just math. It has no moral agency.
This isn’t hypothetical. This is happening millions of times daily across Meta, Google, TikTok, and Pinterest campaigns. The AI optimizes for conversion, engagement, or whatever metric feeds it-completely divorced from the human context of why someone is converting or engaging.
How Performance Optimization Became Ethical Anesthesia
The advertising industry has always had a complicated relationship with ethics. But historically, there was a human in the loop-a creative director, a strategist, a media buyer-who made decisions and could be held accountable for them.
AI has surgically removed that person.
Consider the standard workflow for a high-performing Meta campaign today:
- Upload multiple creative variations
- Define conversion objective
- Let algorithmic delivery optimize
- Winning creative gets 80% of budget automatically
- AI generates new variations based on top performers
- Rinse and repeat
At what point in this process does someone stop and ask: “Should we be doing this?” Not can we, not is it effective-but should we?
The problem is that “should” questions slow down the optimization flywheel. They introduce friction into systems designed for frictionless efficiency. And in a lean, data-first agency environment, that friction feels like waste.
But what if that friction is actually the most valuable thing we could be preserving?
Dynamic Creative: The Ethics of a Thousand Tiny Decisions
Let’s get specific. Dynamic Creative Optimization (DCO) is now standard across major platforms. The AI automatically tests hundreds or thousands of combinations of headlines, images, calls-to-action, and audience segments to find the optimal match.
Sounds efficient. Is efficient. And that’s precisely the problem.
When a human media buyer decides to target “new mothers interested in weight loss” with “Get your pre-baby body back” messaging, there’s an opportunity for reflection, pushback, or refinement. When an AI identifies that this combination has a 34% higher conversion rate and automatically scales it-that reflection point vanishes.
The ethical decision is atomized into micro-optimizations that no single person makes or even sees.
I’ve watched campaigns where the AI discovered-entirely on its own-that serving financial services ads at 2 AM on weekends dramatically improved conversion rates. Why? Because that’s when people with money problems can’t sleep and are most vulnerable to quick-fix solutions.
The algorithm found the exploitation point. It just didn’t know that’s what it was doing.
And here’s the thing: When I bring this up in client meetings or agency reviews, the response is almost always some version of “but the data shows it works.” As if effectiveness and ethics were the same thing. As if the point of advertising is simply to find the most efficient path to conversion, regardless of the human cost.
The Consent Illusion
We’ve built an entire regulatory framework around “informed consent”-GDPR, CCPA, iOS privacy changes. Users can theoretically control their data and how it’s used.
But here’s what those frameworks miss: Nobody can meaningfully consent to something they don’t understand, and AI advertising has become fundamentally incomprehensible.
Can you explain why you saw a particular ad? Really explain it? Not the surface-level “I searched for running shoes”-but the complex interplay of:
- Your behavioral patterns across platforms
- Lookalike modeling from converters with your profile
- Predictive models anticipating life changes
- Sentiment analysis of your social content
- Engagement patterns that signal psychological vulnerability
- Real-time auction dynamics and budget pacing
Neither can I. Neither can the advertisers. Increasingly, neither can the platforms.
We’ve created advertising systems so complex that even their operators can’t fully explain individual decisions. How can anyone consent to that? How can anyone be accountable for that?
Measuring Everything, Understanding Nothing
Here’s an uncomfortable question for my fellow performance marketers: What if our attribution models are hiding ethical harms by abstracting them into metrics?
When we report “2.3X ROAS” or “34% lower CPA,” we’re compressing an enormous amount of human experience into numbers that erase context. Those conversions represent real people making real decisions-sometimes good ones, sometimes harmful ones, often somewhere in between.
But our dashboards don’t show:
- The person who went into debt for a product they couldn’t afford
- The teenager who developed body image issues from repetitive targeting
- The parent who missed their kid’s event because a gamified app successfully reclaimed their attention
- The elderly person confused into a subscription they don’t understand
These outcomes don’t appear in our KPI reports. They’re externalities-costs borne by individuals and society that never show up in the campaign metrics.
We’ve optimized ourselves into ethical blindness.
When AI Makes Us Worse Marketers
Here’s the part that should terrify every creative professional: AI isn’t just raising ethical questions-it’s making us worse at navigating them.
When human creativity drove advertising, we developed intuitions, taste, and judgment about what resonated with people and why. The best creatives understood human psychology, cultural context, and emotional nuance.
But AI optimization trains us to think differently. It teaches us to:
- Trust the data over our instincts (even when our instincts include ethical guardrails)
- Value efficiency over appropriateness (because that’s what the algorithm rewards)
- Think in variants and variables rather than holistic human experiences
- Optimize for immediate response rather than long-term relationship or brand health
I’ve watched talented strategists deskill in real-time, becoming button-pushers who defer to the algorithm instead of collaborators who use it as one input among many.
The lean startup methodology-test quickly, iterate, optimize-is powerful for business efficiency. But it’s dangerous when applied to ethical questions. Some things shouldn’t be “tested” on real people. Some optimizations shouldn’t happen, even if they increase conversion by 40%.
Three Uncomfortable Proposals
If I’ve done my job, you’re feeling uncomfortable right now. Good. We should be uncomfortable. The alternative is sleepwalking into an advertising ecosystem that’s efficient, effective, and fundamentally inhuman.
So what do we do about it? Here are three proposals that go beyond the typical “create an AI ethics board” corporate theater:
1. Mandate Algorithmic Transparency at the Decision Level
Not just aggregate data about how AI works-specific transparency about why this person saw this ad at this moment. If platforms can’t explain it, they shouldn’t be able to serve it. This would force system designers to build accountability into the architecture rather than abstracting it away.
Would this slow down the optimization engines? Absolutely. That’s the point.
2. Establish “Ethical Friction” Requirements
Require human review and approval for:
- Ads served to algorithmically identified “vulnerable” audiences
- Creative variations that deviate significantly from approved concepts
- Targeting combinations that correlate with protected characteristics or life crises
- Budget increases beyond predetermined thresholds
Yes, this creates inefficiency. That inefficiency is the feature, not the bug. It reinserts human judgment at critical decision points.
3. Reimagine Attribution to Include Harm Metrics
What if campaign dashboards had to include:
- User complaints and negative feedback rates
- Post-conversion regret metrics (returns, cancellations, complaints)
- Vulnerable audience exposure rates
- Long-term brand health impacts
This would fundamentally change the optimization function. Instead of pure conversion maximization, AI would optimize for sustainable, ethical conversions-the kind that build brands rather than extract value.
The Agency Dilemma
I know what some of you are thinking: “Easy for you to say, but I have quarterly targets and competitive pressure. My clients demand results.”
You’re right. It’s easy to write about ethical advertising and much harder to practice it when you’re trying to scale Facebook campaigns profitably or prove your value against competitors who have no such scruples.
At Sagum, we’ve built our reputation on achieving results-on scaling profitable campaigns across Meta, Google, TikTok, and Pinterest. Our entire model is built on alignment with client goals and delivering measurable outcomes.
But here’s the thing: Short-term performance and long-term sustainability aren’t actually in conflict.
The same targeting practices that feel exploitative today create the consumer backlash, regulatory crackdown, and platform restrictions of tomorrow. The brands that build lasting value are the ones that treat their customers as humans rather than conversion targets.
When we limit our client roster and create deep alignment, it gives us the space to have harder conversations about what sustainable growth looks like. When we establish clear 30-60-90 day expectations, we can balance immediate performance with ethical considerations.
This isn’t about sacrificing results for principles. It’s about recognizing that principles are results-just on a timeframe that extends beyond the next campaign report.
The Choice We’re Making
Every time we launch a campaign with AI optimization, we’re making a choice about what kind of advertising industry we want to build. We’re choosing what gets measured, what gets optimized, and therefore what gets valued.
Right now, we’re choosing efficiency over accountability. Speed over scrutiny. Performance over people.
We can make different choices. But only if we recognize that we’re making choices at all.
The machines aren’t coming for our jobs. They’re coming for our judgment, our responsibility, and our humanity. And unlike privacy or bias-issues that can be addressed with better data governance or algorithmic fairness-this crisis of accountability threatens the very foundation of what it means to practice marketing ethically.
The question isn’t whether AI will change advertising. It already has.
The question is whether we’ll let it change us.
What ethical dilemmas have you encountered with AI-driven advertising? Where do you draw the line between optimization and exploitation? Share your thoughts and experiences in the comments below.
