AI-generated fake videos are spreading rapidly, misleading vulnerable viewers like the elderly. These videos evade traditional filters, causing ads to appear on harmful content and wasting ad spend. A healthier ad ecosystem requires better content detection and clearer insight into where ads are placed.

In today's digital landscape, video content has become an indispensable part of our daily lives, projected to make up 80-90% of global internet traffic by 2027. Yet, lurking in the shadows of this vast digital space is a new and unsettling threat: harmful or deceptive content generated by artificial intelligence (AI) technology. Recent reports of "AI fake documentaries" circulating on online platforms, with outlandish titles like "Man Who Impregnated 32 Women" or "50-Something Korean Man Impregnated Three Mongolian Beauties," reveal a shocking reality.

Source: YouTube

These videos, crafted by combining AI image generators, synthesized narration, and provocative captions and thumbnails, are racking up millions of views despite their preposterous narratives. The alarming issue is that some elderly viewers are mistaking these fabricated videos for real events. Comments such as "Lucky man, I'm envious" and "Solution to population decline" underscore the severity of this misunderstanding. This phenomenon cunningly exploits the low digital literacy among the elderly and their vulnerability to information. Furthermore, the societal impact of AI-generated content cannot be ignored, as these videos are even consumed in public spaces, causing discomfort to those nearby. The spectrum of harm these videos inflict is vast, presenting themselves as "life wisdom" or "life lessons," while also incorporating racist tropes, excessive sexual objectification, and distorted fantasies about age gaps.

The Proliferation of AI-Generated Content and the Advertiser's Dilemma

The unchecked proliferation of such AI content presents advertisers with a significant, often unseen dilemma. In the digital advertising market, the "context" in which an ad appears remains a challenging issue for many advertisers. Traditional brand safety measures have primarily relied on keywords or metadata, but AI-generated content cleverly sidesteps these filters. This is because even sensational content can be disguised with ordinary titles and descriptions, or by subtly rephrasing specific words to bypass filtering mechanisms. Moreover, certain AI audio novels, which have minimal visual harmfulness, are difficult to detect with existing visual-centric filtering systems. 

This opaque and imprecise video inventory environment directly leads to financial losses for advertisers. According to an analysis by Pyler utilizing a Video Understanding AI model on approximately 10,000 campaigns and over 100 million ad placements, it's estimated that a typical brand spends 20% to 30% of its ad budget on video ads placed within unsafe or unsuitable content. This exposure to harmful videos directly translates into wasted ad spending and brand safety concerns for advertisers. Beyond mere budgetary waste, sponsoring creators of harmful videos can draw criticism regarding a company's corporate social responsibility (CSR). Ultimately, this can damage brand value and negatively impact long-term profitability.

The advancement of AI technology is a double-edged sword. While it offers groundbreaking opportunities, it also serves as a tool for mass-producing harmful or deceptive content. Now, more than ever, we must clearly recognize what to trust, how to protect ourselves in the "fake world" created by AI, and precisely where our valuable advertising budgets are being spent. Acknowledging and understanding these complex, multi-layered issues is the crucial first step toward building a healthy and responsible digital advertising environment.

PYLER supercharges video ad safety with NIVDIA's DGX B200 and AI Blueprint for smarter, safer placements.
Discover how PYLER's real-time video understanding helps brands like Samsung, BYD, and Hana Financial thrive.

At PYLER, we leverage our platform, video understanding AI, which deploys visual AI agents to ensure brand safety by verifying that ads are never displayed alongside inappropriate videos and placing them in more contextually relevant settings to protect brand value and enhance campaign effectiveness.

With the explosive growth of generative AI, the sheer volume of content uploaded to video platforms has skyrocketed. In such an environment, displaying ads next to content that’s not brand aligned increases marketing costs to target the correct audiences, but also presents a risk of negative brand impressions and potential reputational damage.

Video Understanding AI is crucial for effectively managing and validating content in an era of massive, high-speed uploads. When properly implemented, it becomes a key asset in safeguarding and enhancing a brand’s overall value.

Video content processing is significantly more complex than image or text processing, demanding substantial computational resources from preprocessing to inference. To handle this real-time influx of video content effectively, a high-performance pipeline utilizing optimized computing resources is essential.

In this post, we’ll discuss how PYLER addresses these challenges by adopting NVIDIA AI Blueprint for video search and summarization (VSS) in conjunction with NVIDIA DGX B200 systems. We’ll explore how we pair NVIDIA software and accelerated computing, with PYLER’s proprietary Video Understanding AI, to help ensure brand safety at scale.

PYLER Boosts Content Safety with NVIDIA AI Blueprint for VSS and NVIDIA AI Enterprise

PYLER automatically filters, classifies, and analyzes large-scale video data to help advertisers run their campaigns safely and efficiently. To accomplish this, PYLER selected NVIDIA AI Blueprint for VSS, built on the NVIDIA AI Enterprise software platform, offering enterprise-grade support, advanced security measures, and continuous, reliable updates.

NVIDIA AI Enterprise delivers innovative blueprints and commercial stability, enabling PYLER to deploy secure, scalable and continuously maintain AI workflows.

Key advantages provided by VSS Blueprint include:

1. High-Quality Video Embeddings

   • State-of-the-art embedding models are essential for capturing rich contextual and semantic information from videos. PYLER integrates the NVIDIA NeMo Retriever  nv-embed-qa-1b-v2 model for minimal information loss during the embedding and captioning process.

2. Scalable Infrastructure for training custom VLMs and Large-Scale Video Processing

   • NVIDIA DGX B200 accelerates custom VLM training and delivers up to 30x faster performance and best-in-class energy efficiency compared to its predecessor. 

   • With the immense volume of video content uploaded daily, GPU-based parallel processing is critical for scalability. PYLER utilizes VSS blueprint’s chunk-based parallel processing and GPU node expansion to dynamically distribute workloads and process video data in real-time. We look forward to enabling this on NVIDIA DGX B200 for inference.

3. Flexible Verification and Search via RAG

   • PYLER rapidly searches and verifies vast amounts of video embeddings, leveraging Context-Aware Retrieval-Augmented Generation (RAG). This supports automated summarization, indexing, Q&A, and content validation tasks, significantly improving inference speed, accuracy, and context-awareness.

     
     

Expected Outcomes for PYLER

With these integrations, PYLER’s Video Understanding AI software can provide:

1. Maximized Brand Safety

   • Precisely filter, classify, and recommend safe ad placements, ensuring ads never appear next to inappropriate or harmful content.

2. Enhanced Brand Trust

   • Provide advertisers with robust and reliable AI models, fostering stronger viewer trust. As brands gain consumer confidence, PYLER’s credibility as an industry leader grows in parallel.

3. Improved Data Pipeline Efficiency

   • Deploy automated, GPU-accelerated workflows optimized at a scale far beyond traditional MLOps pipelines, significantly increasing operational efficiency and throughput.

4. Faster Service Launch

   • Leverage NVIDIA’s cutting-edge technology stack – including VSS blueprint, NVIDIA Metropolis, NVIDIA Dynamo-Triton, formerly Triton Inference Server, , NVIDIA NeMo Retriever and DGX B200 systems, – enabling innovative enterprises to quickly launch and scale AI-powered video services with fewer resource constraints.

     
     

Customer Success Stories

1. Samsung Electronics: PYLER AiD Redefines Brand Safety Standards for Product Branding Campaigns

• As a leading global brand, Samsung Electronics faced challenges protecting its brand from appearing alongside misaligned content on digital platforms like YouTube
  By integrating PYLER AiD’s real-time AI video analysis, Samsung Electronics automatically blocked ads from harmful content categories based on global standards.
  By integrating the AiD solution into Samsung Electronics' ad campaigns, ads were automatically prevented from running on content categories threatening brand values, such as violence, explicit content, hate speech, and fake news. The AiD dashboard ensured brand safety and trust with transparent real-time monitoring and performance tracking.

• Results:

  • Significant Reduction in Harmful Content Exposure: PYLER AiD reduced ad exposure to sensitive or inappropriate content that could threaten brand value by 77%, setting a new standard for brand protection.

  • Optimized Ad Budget Spend: Preventing ads from appearing on unsuitable content allowed Samsung to focus its marketing budget on safe, high-value, brand aligned placements, greatly increasing advertising effectiveness.

  • Enhanced Brand Trust and Value: Proactively eliminating potential brand crises and consistently delivering a safe and trustworthy brand image to consumers successfully defended and strengthened brand reputation and customer trust.

2. BYD: PYLER is Ensuring Successful Korean Market Entry Through Brand Protection and High Intent Moment Targeting.

• As a global leader in electric vehicles, BYD faced the critical task of successfully launching its first model in Korea, the 'Atto 3'. It was essential to create a positive first impression while proactively managing the risk of ads appearing alongside potentially negative content (e.g., critical reviews, defect discussions) that could harm the new brand's image upon market entry.

• PYLER's AI-Powered Contextual Targeting and Brand Suitability Management technology precisely targeted BYD ads for positive or neutral YouTube content about EVs and small SUVs such as vehicle reviews, test drives, and competitor comparisons. Simultaneously, the AI excluded negative videos by analyzing content tone and context. This ensured BYD ads appeared only in positive, relevant high intent moments aligned with potential customer interest, effectively achieving both contextual targeting and brand suitability simultaneously.

• Results:

  • Maximized Exposure in Brand-Friendly Contexts: Ad placements within safe, positive, and relevant EV/SUV video contexts increased exponentially—more than 100x— compared to previous methods—delivering high relevance and brand suitability to potential customers.

  • Significantly Boosted Potential Customer Interest: By effectively filtering out negative noise and concentrating ads on highly relevant content, BYD captured genuine potential customer interest, driving click-through rates (CTR) nearly 4x higher than baseline campaigns.

  • Successful Market Entry & Enhanced Brand Image: PYLER's integrated AI solution effectively managed the risks associated with new market entry, built a positive brand image, and played a key role in BYD's successful initial launch in the Korean market, receiving over 1,000 pre-order applications in the first week.

3. Hana Financial Group: PYLER AiM Optimizes Performance Across Diverse Campaigns with Automatic Video Placement Recommendation and Targeting

Hana Financial Group, one of South Korea’s largest financial holding companies manages a diverse portfolio including banking, credit cards, securities, and asset management. Facing the challenge of running multiple YouTube campaigns with distinct goals-such as attracting public pension accounts, providing retirement pension information, and promoting a public service announcement against illegal gambling-the group needed to reach targeted audiences within relevant content contexts while achieving varied KPIs like conversion rates, ad efficiency, and user engagement.

To address this, Hana Financial Group implemented PYLER AiM, an AI-powered contextual targeting solution. PYLER’s AI video analysis enabled precise targeting of specific YouTube content moments aligned with each campaign’s objective. For example, the Public Pension campaign focused on retirement planning and celebrity ambassador content, while the Retirement Pension campaign targeted videos about post-retirement finance. The Anti-Gambling PSA featured esports star Faker and targeted gaming and esports content popular with younger viewers, maximizing engagement and campaign effectiveness.

• Results:

  • Optimized Performance for Each Campaign Goal:

    • (Public Pension) Achieved high average conversion rates exceeding 1.5%, demonstrating effective customer acquisition, with specific target groups surpassing an outstanding 3.6% conversion rate.

    • (Retirement Pension) Showcased dramatic improvements in ad efficiency compared to general targeting, with click-through rates (CTR) increasing more than eightfold and cost per click (CPC) reduced by over 70%.

    • (Public Service Ad) Successfully connected the featured celebrity 'Faker' with fan interests, resulting not only in a noticeable uplift in CTR but also proving high user immersion, with average view duration per impression and 100% video completion rates doubling compared to general campaigns. Furthermore, it reached more unique users efficiently, with average frequency halved.

  • Accurate Target Channel Reach: Across all campaigns, PYLER's contextual targeting concentrated ads within highly relevant channels, demonstrating significantly better reach within key target channels compared to general targeting. This confirmed the effective allocation of the ad budget to the most suitable potential customers.

    
    

Conclusion and Future Outlook

By adopting agentic AI through NVIDIA AI Blueprint for video search and summarization and DGX B200 systems, PYLER aims to strike a balance between protecting brand value and maximizing advertising efficiency in a market flooded with video content.

With large-scale GPU parallel processing, efficient embeddings, and Context-Aware RAG for search and verification, we will be able to detect negative content almost in real time and reliably adhere to brand guidelines.

Going forward, PYLER plans to continue actively harnessing NVIDIA’s latest AI technology to deliver fast, high-quality services that connect video content safely and appropriately across various domains.

Korean YouTube’s “storytime” boom hides risky content beneath viral narratives.
Learn why brands must watch where their ads appear — before trust turns into backlash.

One notable trend on Korean YouTube is the immense popularity of what Koreans call "ssul punda" (썰 푼다), which translates to creators sharing "storytime" videos, or personal anecdotes, in a casual and engaging way. If you're familiar with Korean YouTube content, you’ve likely already come across this phenomenon. 

These videos feature individuals sharing a wide range of tales, from humorous dating mishaps and strange encounters with colleagues to behind-the-scenes celebrity gossip and shocking accounts of specific events. Whether based on true experiences or creative fiction, these emotionally charged stories often perform well with the YouTube algorithm. However, this type of content can also harbor harmful elements such as fake news, sensationalized narratives, hate speech, explicit sexual experiences, and even stories about suicide attempts.

This raises the question: why is this "storytelling" format so prevalent and popular specifically on Korean YouTube, and what do advertisers need to keep in mind?

What is "Storytime" Content?

From the perspective of PYLER's brand safety solution, 'AiD', "storytime content refers to videos where the primary purpose is to convey true or fabricated stories and share gossip. Think of it as online storytelling. This style of sharing narratives has deep roots in Korean online culture, with many popular internet communities fostering a tradition of sharing personal accounts. On YouTube, common examples of content range from dating scandals, family conflicts, and workplace rumors to reviews of sex work establishments and accounts of failed suicide attempts – often prioritizing sensationalism over public interest.

Harmful "Storytime" Content: A Greater Risk in Korea

We searched for the same keyword—"grandmother"—on YouTube in four different languages: Korean (할머니), English (grandmother), Chinese (奶奶), and Japanese (おばあちゃん). The results revealed a stark contrast: when searched in Korean, the platform predominantly surfaced provocative "storytime" videos. In contrast, the English, Chinese, and Japanese search results mainly featured wholesome family content, informational vlogs, or cultural introductions.

This discrepancy goes beyond mere algorithmic variation—it highlights the overproduction of sensational, story-driven content in the Korean YouTube ecosystem. Narratives often revolve around sexual experiences, personal conflicts, or violent incidents, with creators prioritizing shock value and algorithmic optimization over ethical storytelling or content quality.

As a result, brands advertising in Korea face a significantly higher risk of having their ads placed alongside harmful storytime content compared to other markets. Over time, this misalignment can pose a serious threat to brand image and consumer trust.

Why is "Storytime" Content So Dominant in Korea?

While similar formats like "Storytime" or "Personal Confession" content trends exist internationally, they haven't achieved the same level of mainstream popularity and genre establishment as this "storytime" content in Korea. Korean "stories" have become a distinct genre due to a combination of factors: rapid audience engagement, immersive narrative structures, and narratives often based on anonymity.

1. A Community Culture Familiar with Sharing Stories Online: For a long time, Korean online communities like DC Inside, Nate Pann, Blind, and Today's Humor have been active spaces for anonymous "story sharing." This established method of narrative delivery naturally extended to YouTube, making the consumption of video-based "storytime" content feel familiar and accessible.
   
   

2. Influence of Broadcast Formats: Popular Korean talk shows like "Knee Drop Guru" and "Radio Star" frequently feature guests sharing personal anecdotes and experiences. This long-standing exposure to spoken storytelling in mainstream media has created a viewing culture that readily accepts "stories from good talkers" as engaging content.
   
   

3. Anonymity and Third-Party Storytelling: Unlike direct first-person confessions, Korean "storytime" content that relays or reports others' stories in the third person is also widely consumed. Reconstructing events, narrating urban legends, and summarizing incidents can generate high levels of immersion, making it easy to diversify content formats.
   
   

4. Low-Cost, High-Engagement Format Combined with Algorithms: "Storytime" content can often be produced simply with text-to-speech narration, without the need for cameras or extensive editing. Its emotionally engaging narrative structure and clickbait-style titles align very well with the YouTube algorithm. This allows creators to achieve high view counts with relatively low production costs.

   
   

What About Overseas?

While "Storytime" content exists in English-speaking countries like the United States, its nature and consumption patterns differ.

Why Should Brands Pay Attention to This Phenomenon?

This "storytime" content presents a high-risk environment for advertising. However, many brands are currently running ads without fully understanding these risks.

Key Risks:

• Risk of Context Distortion: When a brand's advertisement appears within content dealing with personal misfortunes or social conflicts, it can lead to negative associations between the brand and those issues.
  
  

• Emotional Mismatch: Exposure to content featuring topics like sex work, suicide, or hate speech can severely damage a brand's image and reputation.
  
  

• Decreased Credibility: The truthfulness of "storytime" content is often questionable, making it highly susceptible to fake news or defamation. Appearing alongside such videos can erode trust in the brand itself.

In fact, PYLER's analysis of advertising campaigns has revealed cases where up to 35% of the brand’s ad budget was spent on content inappropriate for the brand. This highlights the reality that even branding-focused advertisements are being placed in unsuitable environments.

An Advertising Environment that Considers Content Context is Necessary

This "storytime" content is not just a passing trend; it's a sensational narrative structure that is highly optimized for the YouTube algorithm. However, with harmful content like depictions of sex work, suicide, violence, hate speech, and fake news being circulated under this "story" umbrella, it has become crucial for brands to prioritize a safe advertising environment that considers content context as much as advertising efficiency.

To this end, PYLER monitors real-time ad placements and blocks ads from appearing on inappropriate content, providing a safe and high-performing advertising environment. PYLER is committed to helping brands connect with consumers in a safe and healthy content ecosystem.

PYLER CEO Oh Jaeho named to Forbes Asia’s ‘30 Under 30’ for reshaping video advertising with AI.
Recognized for leading PYLER’s rise in brand safety innovation across global markets.

Recognized as a Young Leader Shaping the Future of AdTech

In May 2025, Oh Jaeho, Co-Founder and CEO of PYLER, was named to the Forbes Asia “30 Under 30” list in the Marketing & Advertising category. The list celebrates young leaders across Asia who are transforming their industries, and Oh’s selection marks a significant milestone for Korea’s emerging AdTech ecosystem.

Revolutionizing Ad Placement in Video: PYLER’s AiD

Founded in 2021, PYLER set out to solve a long-standing issue in the video advertising industry: brand safety. The company’s flagship solution, AiD, is an AI-powered platform that ensures ads appear in brand-safe video environments—particularly on platforms like YouTube.

AiD analyzes the content of videos in real time, evaluating tone, context, and messaging to determine whether an ad should be placed. By automating what previously required manual review by agencies and marketers, AiD not only protects brand image but also enhances ad effectiveness at scale.

Today, PYLER works with global brands like Samsung, Bulgari, and BYD, as well as leading advertising agencies including Cheil Worldwide and Innocean.

Proven Growth and Technology with $24M in Total Funding

To date, PYLER has raised a cumulative 34 billion KRW (~$24 million) in funding. In July 2024, the company secured a 22 billion KRW Series investment from top-tier investors such as KDB Bank, KT Investment, Stonebridge Ventures, and SV Investment—validating its technical excellence and market potential.

Beyond the Forbes List: A Signal of What’s to Come

Being named to the Forbes “30 Under 30” list is more than an accolade—it's a recognition of the transformative work being done in a rapidly evolving digital ad landscape. PYLER continues to lead this transformation by building AI solutions that make advertising smarter and safer.

Oh Jaeho’s inclusion is a strong signal of the global impact PYLER aims to achieve in the years ahead. As the company moves forward, it remains committed to setting new standards in video-based advertising and delivering AI solutions that marketers and brands can trust.

You can view the full list of honorees in the Marketing & Advertising category of Forbes 30 Under 30 – Asia 2025 at the link below:
Forbes 30 Under 30 – Asia 2025

To learn more about one of this year’s honorees, PYLER CEO Jaeho Oh, visit his Forbes profile here:
Jaeho Oh on Forbes

PYLER becomes the first in Korea to adopt NVIDIA’s DGX B200, redefining AI infrastructure for AdTech.
With 30x faster performance, PYLER accelerates brand safety, contextual targeting, and global AI leadership.

Pioneering AI Infrastructure with Next-Gen NVIDIA Hardware

On February 27, 2025, PYLER became the first company in Korea to deploy NVIDIA’s latest AI accelerator, the DGX B200, marking a major leap in the company’s AI infrastructure. As NVIDIA’s newest generation AI system, the DGX B200 has drawn significant global attention since its release. PYLER's adoption—preceding even major institutions both locally and abroad—highlights the company's position at the forefront of AI innovation.

A ceremony was held to commemorate the milestone, symbolizing PYLER’s renewed commitment to redefining advertising technology through AI.

DGX B200: A New Benchmark in AI Performance

Equipped with NVIDIA’s next-generation Blackwell GPU architecture, the DGX B200 offers up to 30x improved computational performance compared to its predecessor. It delivers industry-leading energy efficiency and is purpose-built for training and inference of large-scale, video-centric AI models.

For PYLER—whose core business involves real-time video understanding across vast volumes of online content—the DGX B200 is a game-changing addition that will significantly boost its technical capabilities.

Powering the Next Generation of AI AdTech

The DGX B200 will supercharge PYLER’s core AI solutions across the board, especially in three key areas:

• Brand Safety: Faster and more accurate detection of harmful content and ad placement control

• High-intent Moment Targeting: Enhanced precision in real-time targeting based on video context

• Contextual Ad Optimization: Smarter prediction of user responses and ad relevance

PYLER’s flagship solution, AiD, has already expanded its footprint both domestically and globally, collaborating with major advertisers like Samsung Electronics, Nongshim and KT Corporation. This infrastructure upgrade is expected to greatly enhance customer experience and maximize advertising performance.

Building World-Class Video Understanding for Advertising

Jihun Kim, CTO of PYLER, stated:
“With the DGX B200—the first of its kind in Korea—we’ve laid the foundation for building world-class video understanding capabilities tailored to the advertising domain. We will continue to push the boundaries of AI innovation to ensure that brand messages are delivered in the safest and most contextually relevant environments possible.”

Under its strategic partnership with NVIDIA, PYLER is focused on leading the future of AI-powered advertising—from content moderation to contextual AdTech. The adoption of DGX B200 is more than just a hardware upgrade—it marks a pivotal step in realizing PYLER’s vision of raising both the quality and trust of digital advertising through cutting-edge AI.

PYLER raises KRW 22 billion to scale its AI-powered brand safety solution, AiD.
Trusted by top brands, PYLER aims to lead globally in video understanding and ad transparency.

Recognized for Its Cutting-Edge AI Brand Safety Technology

PYLER, a leading provider of AI-powered brand safety solutions, has successfully secured KRW 22 billion in its latest series funding round. Key investors include Stonebridge Ventures, Korea Development Bank, SV Investment, and KT Investment—underscoring strong confidence in PYLER’s technology and growth potential.

‘AiD’: Helping Brands Regain Control Over Ad Placement

PYLER’s flagship solution, AiD, leverages AI to analyze the context of YouTube videos where brand ads are placed, blocking exposure to harmful or inappropriate content. The solution protects brands from being associated with adult content, hate speech, fake news, and fringe religious material that could damage brand reputation.

With millions of new videos uploaded to YouTube every day, manual review is no longer feasible. In this context, investors recognized AiD’s value in empowering advertisers with greater control and visibility over where their ads appear.

Brand Safety Directly Impacts Consumer Trust and Behavior

According to a joint report published in January by PYLER and the Korea Broadcast Advertising Corporation (KOBACO), 89.5% of consumers said brand safety is important, while 96% stated they would not purchase from advertisers who neglect it. These figures clearly show that brand safety plays a critical role in shaping both consumer trust and purchasing behavior.

On Track to Become a Global Leader in Video Understanding

“Our solution has already been tested and trusted by major Korean brands like Samsung, Hyundai Motor Company, Cheil Worldwide, and Innocean,” said Jaeho Oh, CEO of PYLER. “With this new funding, we aim to establish ourselves as one of the most competitive players in video understanding on a global scale.”

Powered by advanced AI and real-time content analysis, PYLER remains committed to building a trustworthy advertising environment—for both brands and consumers.