Milena Traikovich is a strategist who understands that the “warning labels” often placed around artificial intelligence in marketing are missing a much more exciting reality. As an expert in demand generation and performance optimization, she views AI not as a threat to creativity, but as a sophisticated lens that is finally unlocking inventory and content opportunities that were once invisible to the naked eye. In this conversation, we explore the shift from rigid, keyword-based safety tools to multimodal systems that can actually understand tone and intent. We also dive into new data regarding creator-led content, examining why traditional brand-first messaging is failing and how a “payoff-first” structure is revolutionizing engagement rates across platforms like TikTok and Instagram.
The current landscape of digital advertising is undergoing a quiet but massive recalibration, moving away from “set it and forget it” tactics toward a more nuanced, data-driven approach to human emotion. This interview covers the hidden costs of outdated exclusion lists, the psychological impact of product demonstrations versus brand claims, and the specific emotional triggers that drive success in different industries. By analyzing recent findings on audience attention during high-stakes periods like election cycles, Milena provides a roadmap for brands to remain present and authentic without sacrificing safety.
Traditional keyword-based safety tools often flag safe content, causing advertisers to miss out on significant inventory. How does this impact modern campaign reach and the overall efficiency of media spend?
It is a massive, often invisible blind spot that keeps brands from reaching their most engaged audiences. Recent data reveals that a staggering 54% of URLs were blocked based on keywords alone, even when the underlying content was perfectly appropriate once you looked at the full context. This means more than half of the potential web is effectively hidden because we are using blunt instruments that can’t tell the difference between a dangerous situation and a high-quality news report. For years, this has led to an artificial scarcity of inventory, driving up costs for the remaining “safe” spots while leaving premium content untouched. It feels like a burden to change these settings, but the math shows that teams refusing to evolve are simply leaving money on the table.
With multimodal AI now capable of analyzing video, audio, and images together, what should media buyers be doing differently to move beyond simple transcript scanning?
We have to move toward a holistic read of intent rather than just hunting for trigger words. Multimodal AI is a game-changer because it builds a comprehensive understanding of a video’s tone, which keyword lists were never built to capture in the first place. My advice to teams is to pull up their exclusion lists this week and ask exactly when they were last reviewed; for many, the answer is “far too long ago.” You need to challenge your verification partners on how they handle edge cases and ensure they are using tools that see the nuance in a scene. It isn’t a “set it and forget it” upgrade, but a necessary recalibration that still requires a human to check the model’s work during live, fast-moving cycles.
High-stakes periods like election cycles often see brands pulling back entirely from news inventory. Why is this strategy counterproductive for brands looking to maintain audience attention?
The instinct to retreat is understandable, but the data suggests that pulling back entirely means your brand is absent precisely when consumers are most engaged with media. Election cycles are when news consumption peaks and audience attention is at its highest, creating a prime environment for brands that know how to navigate it with precision. Instead of abandoning the space, marketers should use content-level evaluation to distinguish between a factual report on voter turnout and high-risk partisan commentary. These two types of content have completely different risk profiles, and modern tools can finally tell them apart. If you aren’t present during these peaks, you’re missing the moments when your audience is most focused on the world around them.
For publishers who find their content unfairly blocked, what structural changes can they make to ensure AI-powered verification systems correctly identify their videos as safe?
Publishers have to start making their video content easier for these new verification and evaluation systems to interpret from the jump. This starts with a commitment to clean, clear metadata and providing high-quality transcripts for every single piece of content. When each video is assessed on its own merits rather than being lumped into broad, blocked categories, the monetization opportunities open up significantly. It’s about doing the legwork that allows contextual AI to see the value in what you’ve created instead of letting it default to a “safe” but profitable-less exclusion. By cleaning up these internal signals, publishers are essentially inviting the AI to see the nuance and quality of their work.
New research indicates that leading with brand messaging in the first few seconds of a video can actually hurt performance. What does this mean for the way we brief creator partnerships?
This finding should lead every content marketer to immediately rewrite their creative briefs. When assets lead with a product, a specific benefit, or brand-heavy messaging in those opening seconds, we see view rates drop by 44% and brand favorability take a 12% hit. Even worse, consideration can drop by as much as 41% compared to content that builds a compelling hook first. Creators have known this instinctively for years, but now we have the data to prove that the “pitch” is the enemy of the “view.” We have to let the brand arrive as the payoff rather than the initial hook if we want to keep the audience from scrolling past our work.
Authenticity is a common goal, but the data suggests that “showing” is much more effective than “telling” when it comes to product claims. How does proof-based content shift the needle on favorability?
There is a visceral difference between a brand claiming a product is “amazing” and a creator demonstrating it in their own real-world environment. Content built around authentic demonstration, showing the product in actual use with a before-and-after, outperformed declarative messaging by 33% in brand favorability. This works because it allows the viewer to visualize the product in their own life, building a level of confidence that a polished ad simply cannot replicate. When a creator like Laura Adlington explains the reasoning behind a styling choice while showing it on her own body, it builds a foundation of trust. It moves the needle on consideration by 15% because the audience feels they are getting an honest assessment rather than a corporate script.
You’ve highlighted that different emotional registers work for different verticals. How should content strategists decide which emotions to prioritize for their specific industry?
You have to realize that there is no universal “best” emotion; what works for a beauty brand might be a disaster for a fashion retailer. For instance, anxiety actually lifts performance in beauty and food content, but it can stifle results in entertainment or retail categories. On the other hand, gratitude is a powerful lift for retail and fashion, but it tends to suppress engagement in the beauty and food spaces. A one-size-fits-all brand voice is essentially leaving performance on the table by design because it ignores these category-specific emotional logics. You must map your content calendar to the specific emotional signals that resonate with your particular audience’s needs and expectations.
The “Peak-End Rule” suggests that audiences remember the emotional high point and the conclusion of a video. How does a satisfying payoff impact organic reach on platforms like TikTok?
If your content builds to a satisfying resolution, you are tapping into the way the human brain naturally encodes memories. Content that successfully hits that emotional peak and then provides a clear payoff sees organic view rates rise by 110% across all platforms. On TikTok specifically, the impact is even more dramatic, with view rates soaring by a massive 318% and engagement jumping by 83%. Most brands front-load their messaging and then let the ending trail off, which means they are optimizing for the part of the video that viewers are most likely to forget. By focusing on a strong ending and an earned resolution, you ensure the most memorable part of the video is the one that drives the most value.
What is your forecast for the evolution of brand safety and content performance as AI becomes more integrated into our workflows?
I believe we are entering an era where the tools are finally getting better at telling the difference between genuinely good content and content that merely looks compliant on the surface. We will see a shift where raw, emotionally authentic reactions—even those that feel a bit awkward—will drive a 25% lift in organic view rates over safe, polished alternatives. The “set it and forget it” mentality will die out, replaced by a need for constant recalibration and human oversight to handle the nuances of live media. Brands that move their product mentions to the “payoff” position and embrace category-specific emotional logic will see their consideration rise by 22%. Ultimately, AI is going to reward specificity and nuance, which is, for once, very good news for creators and marketers alike.
