From 4af12a620af8461d1ad908325f8f7be6442d77b6 Mon Sep 17 00:00:00 2001 From: Gray Date: Wed, 28 Jan 2026 16:09:11 +0800 Subject: [PATCH] Update The Rising Strategic Value of Creative in Digital Advertising --- .../README.md | 24 +++++++++---------- 1 file changed, 11 insertions(+), 13 deletions(-) diff --git a/articles/The Rising Strategic Value of Creative in Digital Advertising/README.md b/articles/The Rising Strategic Value of Creative in Digital Advertising/README.md index 1716a97..7feec67 100644 --- a/articles/The Rising Strategic Value of Creative in Digital Advertising/README.md +++ b/articles/The Rising Strategic Value of Creative in Digital Advertising/README.md @@ -13,27 +13,27 @@ thumb_h: "./thumb_h.png" intro: "Creative has emerged as a key driver influencing success in performance marketing, and digital advertising companies are increasingly asked to facilitate the creation of advertisements. How did we get here, and what comes next?" --- -The digital advertising landscape has recently undergone a significant shift in the wake of short-form video, privacy initiatives, and generative AI. As a result, the industry has moved from championing targeting above everything towards creative as the most critical axis upon which demand-side platforms (DSPs) need to compete. In this article, we’ll recount the history of digital advertising as it pertains to the importance of creative and articulate the current landscape of relevant players. The takeaway is clear: DSPs in 2025 don’t have a choice; they have to facilitate the process of helping firms create, test, and launch creative in order to stay competitive. +The digital advertising landscape has recently undergone a significant shift in the wake of short-form video, privacy initiatives, and generative AI. As a result, the industry has moved toward recognizing creative iteration as a primary performance lever. In this article, we’ll recount the history of digital advertising as it pertains to the importance of creative and articulate the current landscape of relevant players. The takeaway is clear: modern advertising platforms are facing mounting pressure to embed creative iteration directly into the performance advertising workflow. ## Part 1: Know Your History ![](./chart%201.png) -*(Dates are approximate as technological evolutions are a gradual process)* +_(Dates are approximate as technological evolutions are a gradual process)_ The first time that exchanges, data, and bidding were stitched together in a true demand-side platform as we understand them today was around 2007. Before firms like MediaMath, advertisers that wanted to purchase space across the web encountered a scattered, unintuitive, and manual landscape; the revolutionary idea around this time was that software could join all the ad auctions happening across the internet and decide, in real time, which impressions to buy and for how much. Therefore, early DSPs found it especially worthwhile to compete on two major fronts: how much inventory they could access and how advanced their targeting algorithms were. As time went on, the industry consolidated as a few key winners proved they could deliver upon these axes. During the late 2000s and throughout the 2010s, creative was something that advertisers would manage independently and bring to DSPs. A DSP would only think about when and where to show creative and wouldn’t concern themselves with the process of crafting compelling advertisements on behalf of their advertisers. Over time, however, mobile and app-install advertising increased the complexity of formats, ushering in video and playable, interactive advertisements that wouldn’t be possible or relevant without the iPhone and the App Store. As the app economy exploded, mobile-first DSPs built sustainable businesses by merging user acquisition algorithms with format-specific expertise. When performance platforms like AppLovin, Unity, and Liftoff rose in prominence concurrently, DSPs had to get more literate about creative because they noticed how big of a difference elite creative had on conversions and ROAS. DSPs started building internal creative studios to facilitate the process, but they were still mostly service arms as opposed to embedded self-serve tools. -These evolutions set the stage for 2020, which was arguably the most historic year in digital advertising history because Apple broke the existing model of targeting as a differentiator by introducing App-Tracking Transparency (ATT), TikTok became a category-defining consumer app, and COVID forced people to meaningfully engage with the digital landscape far more than they ever had before. Because of iOS 14.5 and ATT, targeting got meaningfully weaker since much of digital advertising’s infrastructure was contingent on being able to follow users all over the web and target advertisements with the data. At the same time, Facebook was forced to reconcile with TikTok’s emergence and their dominance as a true competitor, leading them to launch Reels that August. Aggregating content from all over the internet (as opposed to just from your social network) convinced the masses that short-form video was the medium of the future. The masses were right, and the hours a day that the average DAU spent on TikTok and Reels led to an exponential spike in ad inventory. Yet to work in this new environment, advertisements needed to compete with user-generated content, and walled gardens became acutely aware of the need for compelling creative. +These evolutions set the stage for 2020, which was arguably the most historic year in digital advertising history because of a series of concurrent developments. Most importantly, Apple made the industry's dependence on deterministic user tracking untenable by announcing App-Tracking Transparency (ATT) at WWDC. Because of iOS 14.5 and ATT, targeting got meaningfully weaker since much of digital advertising’s infrastructure was contingent on being able to follow users all over the web and target advertisements with the data. Concurrently, TikTok continued its ascension as a category-defining consumer app, and COVID forced people to meaningfully engage with the digital landscape far more than they ever had before. Facebook was forced to reconcile with TikTok’s emergence as a true competitor, leading them to launch Reels that August. Aggregating content from all over the internet (as opposed to just from your social network) led many analysts to believe that short-form video was the medium of the future. They were ultimately proven right, and the hours a day that the average DAU spent on TikTok and Reels led to an exponential spike in ad inventory. To work in this new environment, advertisements needed to compete with user-generated content, and walled gardens became acutely aware of the need for compelling creative. -While DSPs weren’t directly involved with the rise of short-form video, they were forced to reconcile with the new precedent that creative was a critical axis upon which they needed to compete, especially in the aftermath of ATT. Thankfully, it was only a few years later that generative AI demonstrated the ability to create videos that were essentially indistinguishable from real-life. In 2025, there’s a new battleground that DSPs are forced to compete on: generative AI for ad creative to maximize return on ad spend (ROAS). With this in mind, it’s worth internalizing a deep understanding of the modern landscape when it comes to generative AI for video, generative video for ad creative, vertically integrated walled gardens, and, finally, DSPs and the tools and products they offer (or should offer) in today’s world. +While DSPs weren’t directly involved with the rise of short-form video, they were forced to reconcile with the new precedent that creative was a critical axis upon which they needed to compete, especially in the aftermath of ATT. Thankfully, it was only a few years later that generative AI demonstrated the ability to create videos that were essentially indistinguishable from real-life. In 2026, there’s a new battleground that DSPs will need to compete on: generative AI for ad creative to maximize return on ad spend (ROAS). With this in mind, it’s worth internalizing a deep understanding of the modern landscape when it comes to generative AI for video, generative video for ad creative, vertically integrated walled gardens, and, finally, DSPs and the tools and products they offer (or should offer) in today’s world. ## Part 2: The Modern Landscape ![](./chart%203.png) -Despite not being built explicitly for performance marketing, AI-generated video has redefined what’s possible on the bleeding-edge of technology. OpenAI’s Sora and Google’s Veo 3 have rightfully dominated headlines in 2025 as generative video has become virtually indistinguishable from real life, and the industry has accelerated in relevance and capability behind Big Tech’s flourishing ecosystem. China’s largest players have demonstrated that chip export controls haven’t slowed down innovation as leading Chinese content community and social platform Kuaishou released arguably the world’s best video generation model, Kling 2.5 Turbo 1080p, under their Kling AI brand in September. Minimax has also done exceptionally well as one of the country’s premier foundation model companies, and Alibaba and ByteDance have released multiple models in 2025 that have been met with positive reception within the industry. Startups have brought forth innovative product capabilities that complement pure research as well; Pika Labs and HeyGen have shipped text-to-video products, and Synthesia has built an enduring business around quality AI avatars. Runway is arguably the most successful startup in this space and has already worked alongside major companies to assist with commercials and other creative ventures, offering proprietary models like Gen-3 Alpha and Gen-4 in addition to products like creator tools and professional workflows. Demand right now for products in this space are significant; all four startups raised between 60 and 300+ million dollars each within the last 36 months. Venture capital firms are essentially tripping over themselves to write these startups blank checks. +Despite not being built explicitly for performance marketing, AI-generated video has redefined what’s possible on the bleeding-edge of technology. OpenAI’s Sora and Google’s Veo 3 have rightfully dominated headlines in 2025 as generative video has become virtually indistinguishable from real life, and the industry has accelerated in relevance and capability behind Big Tech’s flourishing ecosystem. China’s largest players have demonstrated that chip export controls haven’t slowed down innovation as leading Chinese content community and social platform Kuaishou released arguably the world’s best video generation model, Kling 2.5 Turbo 1080p, under their Kling AI brand in September. Minimax has also done exceptionally well as one of the country’s premier foundation model companies, and Alibaba and ByteDance have released multiple models in 2025 that have been met with positive reception within the industry. Startups have brought forth innovative product capabilities that complement pure research as well; Pika Labs and HeyGen have shipped text-to-video products, and Synthesia has built an enduring business around quality AI avatars. Runway is arguably the most successful startup in this space and has already worked alongside major companies to assist with commercials and other creative ventures, offering proprietary models like Gen-3 Alpha and Gen-4 in addition to products like creator tools and professional workflows. Demand right now for products in this space are significant; all four startups raised between 60 and 300+ million dollars each within the last 36 months. Venture capital firms are essentially tripping over themselves to write these startups blank checks. In the open-source ecosystem, products like CogVideoX, Mochi-1, Hunyuan, Allegro, and LTX Video have emerged to varying degrees of success. However, the open-source ecosystem has yet to meaningfully compete with closed-source alternatives; a December 2024 arXiv paper published by Tencent’s Hunyuan team articulates, “In contrast to the image generative model community, a significant gap has emerged between open-source and closed-source video generation models. Closed-source models tend to overshadow publicly available open-source alternatives, severely limiting the potential for algorithmic innovation from the public community”. There’s a lot of reasons as to why this is the case. Since token counts in video generation models are massively larger than in image generation models, they’re significantly more computationally expensive to train. The Hunyuan team invested significantly in reducing computational resource requirements for this reason, and for labs that aren’t backed by cash-printing machines like Tencent, it can be prohibitively expensive. HuggingFace wrote that, “Costs arise from dataset collection, hardware requirements, extensive training iterations and experimentation. These costs make it hard to justify producing open-source and freely available models.” Another constraint that open-source teams run into is data; one 2025 arXiv paper affiliated with ByteDance pointed out how leading open datasets have significant issues, writing, “WebVid-10M contains low-quality videos with watermarks and Panda-70M contains many flickering (or still) and blurry videos along with imprecise captions”. Building models with primarily publicly available data puts small teams at an inherent disadvantage because the highest-quality video data is typically proprietary to big players or legally risky to scrape and train with. Researchers aren’t willing to take the legal risks that firms like OpenAI would because the upside doesn’t justify the liability. With all of that being said, though, I still predict that the open-source community will take a big step forward in 2026 – it’ll just have to come from a large Chinese or American firm that has the revenue to stomach the costs and the requisite data it takes to get a model off the ground. @@ -41,30 +41,28 @@ The broad recognition that video generation capabilities are real and here to st Of course, Big Tech isn’t oblivious to these shifts either. In 2024 Ben Thompson detailed how Meta has pushed this product category forward, writing: ->“Advertisers have long understood the importance of giving platforms like Meta multiple pieces of creative for ads; Meta’s platform will test different pieces of creative with different audiences and quickly hone in on what works, putting more money behind the best arrow. Generative AI puts this process on steroids: advertisers can provide Meta with broad parameters and brand guidelines, and let the black box not just test out a few pieces of creative, but an effectively unlimited amount. Critically, this generative AI application has a verification function: did the generated ad generate more revenue or less? That feedback function, meanwhile, is data in its own right, and can be leveraged to better target individuals in the future.” +> “Advertisers have long understood the importance of giving platforms like Meta multiple pieces of creative for ads; Meta’s platform will test different pieces of creative with different audiences and quickly hone in on what works, putting more money behind the best arrow. Generative AI puts this process on steroids: advertisers can provide Meta with broad parameters and brand guidelines, and let the black box not just test out a few pieces of creative, but an effectively unlimited amount. Critically, this generative AI application has a verification function: did the generated ad generate more revenue or less? That feedback function, meanwhile, is data in its own right, and can be leveraged to better target individuals in the future.” -Thompson’s takes are supported by real-world evidence; Meta’s Advantage+ Creative and Ads Manager products have already incorporated generative AI tools corresponding to background generation, image expansion, full image variations, and now tools that can turn static product photos into multi-scene video ads with music, overlays, and text. Millions of advertisers are finding these tools useful; Meta’s engineering blog commented in December 2024, “We estimate that businesses using image generation are seeing a +7% increase in conversions”. A year is a lifetime in the world of AI, though, and the Wall Street Journal has reported that Meta aims to allow brands to completely automate the process of creating advertisements by the end of 2026. +Thompson’s takes are supported by real-world evidence; Meta’s Advantage+ Creative and Ads Manager products have already incorporated generative AI tools corresponding to background generation, image expansion, full image variations, and now tools that can turn static product photos into multi-scene video ads with music, overlays, and text. Millions of advertisers are finding these tools useful; Meta’s engineering blog commented in December 2024, “We estimate that businesses using image generation are seeing a +7% increase in conversions”. A year is a lifetime in the world of AI, though, and the Wall Street Journal has reported that Meta aims to allow brands to completely automate the process of creating advertisements by the end of 2026. Google / YouTube are marching in the same direction. Amazon is too – we’ll come back to that in a second. - ## Part 3: Comparing and Contrasting Modern DSPs ![](./chart%202.png) Everyone understands the tectonic shift that generative AI represents for maximizing return on ad spend, but it’s been interesting to observe the different approaches firms take as they seek to realize this vision. One head-to-head worth spotlighting is Amazon vs. The Trade Desk. Amazon has chosen to build a first-party, vertically integrated creative stack directly inside Amazon Ads and Amazon DSP. By integrating image, video, and audio generation natively inside their ad console, Amazon has effectively removed any and all creative friction for advertisers looking to turn product assets into advertisements. It’s been successful too; their website writes that on average, “advertisers saw a +10.3% higher return on ROAS on Sponsored Brands campaigns that used AI-generated images compared to Sponsored Brands campaigns without AI-generated images”. Big Tech seems to have reached a consensus on the importance of generative AI for asset creation moving forward, and their product capabilities have reflected that as a result. -The Trade Desk (TTD), on the other hand, has elected to partner with external vendors to offer similar services. Setting aside that their offering isn’t nearly as comprehensive or relevant as Amazon’s, TTD’s not owning the IP that their platform integrates with exposes themselves to strategic vulnerabilities. Most notably, TTD doesn’t own the data flywheel for their product, effectively robbing themselves of the opportunity to refine their models over time as they collect data on what kind of creative works, for which product, for which audience, on which surface, and at which time. Over time, I predict that the advantage in the market will amass towards vertically integrated DSPs like Amazon in addition to similarly structured walled gardens like Meta and YouTube where the data feedback loop becomes a larger and larger moat with every passing year. +The Trade Desk (TTD), on the other hand, has elected to partner with external vendors to offer similar services. Setting aside that their offering isn’t nearly as comprehensive or relevant as Amazon’s, TTD’s not owning the IP that their platform integrates with exposes themselves to strategic vulnerabilities. Most notably, TTD doesn’t own the data flywheel for their product, effectively robbing themselves of the opportunity to refine their models over time as they collect data on what kind of creative works, for which product, for which audience, on which surface, and at which time. Over time, I predict that the advantage in the market will amass towards vertically integrated DSPs like Amazon in addition to similarly structured walled gardens like Meta and YouTube where the data feedback loop becomes a larger and larger moat with every passing year. AppLovin’s Axon platform has also leaned into generative AI tools for adjusting creative, and CEO Adam Foroughi spent this November’s Q3 earnings call articulating plans to expand this vision. He explained: “We'll be testing generative AI-based ad creatives. Over time, if we can move to mostly automated creative generation, we believe user response rates to more customized ads on our platform will materially improve”. As a vertically integrated DSP tuned for the mobile gaming and app-install universe, it should catch your attention that these developments are just a few weeks old. Smaller firms with products that behave as DSPs like Appier have also moved into this space seeking to gain a competitive advantage, having acquired AdCreative.ai for 38.7 million dollars back in February. Together, all of these data points form a cohesive narrative: helping advertisers craft exceptional creative with generative video is where this industry is headed, and AdTech firms will be under increased pressure to adapt to these shifts in the coming quarters. ## Conclusion -It’s been interesting to watch how digital advertising has evolved in the wake of generative AI. The old paradigm where DSPs only have to focus on targeting and can ignore creative is quickly breaking down. Privacy changes and short-form video elevated creative from supporting asset to main performance lever, and generative AI has made it technically and economically feasible to generate and adapt huge volumes of creative. In a 2024 interview, Eric Seufert commented, “I don’t want to overstate it, but I don’t think I can; advertising is going to be transformed by AI across every single input. Generative ad creative production, targeting, personalization, measurement; everything is going to be transformed by AI.” At 57Blocks, we’re acutely aware of the generative video and AdTech landscape and have thought deeply about their convergence. - -If today’s DSPs don’t help advertisers create, test, and launch creative, someone else will! +It’s been interesting to watch how digital advertising has evolved in the wake of generative AI. The old paradigm where DSPs only have to focus on targeting and can ignore creative has quickly broken down. Privacy changes and short-form video elevated creative from supporting asset to main performance lever, and generative AI has made it technically and economically feasible to generate and adapt huge volumes of creative. In a 2024 interview, Eric Seufert commented, “I don’t want to overstate it, but I don’t think I can; advertising is going to be transformed by AI across every single input. Generative ad creative production, targeting, personalization, measurement; everything is going to be transformed by AI.” At 57Blocks, we’re acutely aware of the generative video and AdTech landscape and have thought deeply about their convergence. If today’s advertising platforms don’t help advertisers create, test, and launch creative, someone else will! ## Sources + - [AppLovin Corporation (APP) Q3 2025 Earnings Call. “AppLovin Corp (APP) Q3 2025 Earnings Call Highlights: AI …” Yahoo Finance, 5 Nov. 2025.](https://finance.yahoo.com/quote/APP/earnings/APP-Q3-2025-earnings_call-371548.html?utm_source=chatgpt.com) - [Hunyuan Foundation Model Team. “HunyuanVideo: A Systematic Framework for Large Video Generative Models.” arXiv, 2024, arXiv:2412.03603v1.](https://arxiv.org/html/2412.03603v1) - [Zhou, Penghao, et al. “OpenVid-1M: A Large-scale High-quality Dataset for Text-to-Video Generation.” arXiv, 2024, arXiv:2407.02371v3.](https://arxiv.org/html/2407.02371v3) @@ -73,4 +71,4 @@ If today’s DSPs don’t help advertisers create, test, and launch creative, so - [Amazon Ads. “Creative Solutions – Optimize your creative with AI-powered solutions.” Amazon Advertising.](https://advertising.amazon.com/generative-ai-ad-solutions?utm_source=chatgpt.com#vid) - [Meta to Let Brands Create, Target AI Ads Fully by End of 2026, Report Says.” Investopedia, 2024.](https://www.investopedia.com/meta-to-let-brands-create-target-ai-fully-by-end-of-2026-report-says-11746215) - [Sun, Chonglin, Nancy Yu, Haiyu Lu, Liang Wang, Yunchen Pu, Gaoxiang Liu, Gian-Paolo (GP) Musumeci, and Neeraj Bhatia. “Meta Andromeda: Supercharging Advantage+ automation with the next-gen personalized ads retrieval engine.” Engineering at Meta (Facebook Engineering), 2 Dec. 2024.](https://engineering.fb.com/2024/12/02/production-engineering/meta-andromeda-advantage-automation-next-gen-personalized-ads-retrieval-engine/?utm_source=chatgpt.com) -- [Seufert, Eric. “An Interview with Eric Seufert about the current state of digital advertising.” Stratechery, 2024.](https://stratechery.com/2024/an-interview-with-eric-seufert-about-the-current-state-of-digital-advertising/) \ No newline at end of file +- [Seufert, Eric. “An Interview with Eric Seufert about the current state of digital advertising.” Stratechery, 2024.](https://stratechery.com/2024/an-interview-with-eric-seufert-about-the-current-state-of-digital-advertising/)