A client recently came to us with an urgent brand safety request: they wanted to avoid advertising on any content concerning the LA Fires. They did not want to be seen as trying to sell something in a time of great tragedy. Over 80% of advertisers consider brand safety a major concern and this was a particularly difficult situation. They needed to manage their ad placement against a time-sensitive topic, with new articles being published every minute, adding complexity above and beyond the typical evergreen brand safety topics brands want to avoid like racism or drugs.

The client asked partners to block content using the standard keyword implementation used by many advertisers today. Cognitiv can implement keyword blocking, but we offer a much more powerful and complete solution called ContextGPT, thanks to our use of large language models and advanced deep learning AI.

Using the client’s keyword block list,1500 active URLs were detected and blocked. We then prompted ContextGPT, asking it to find all the URLs that had anything to do with the LA Fires, and found over 7500 active URLs that represented millions of impressions in possible exposure, an increase of 5X over the keyword approach. Cognitiv was also able to scan the additional URLs its tool found, and suggest additional keywords for the client to give to their partners that only had keyword brand safety tools available.

ContextGPT not only finds more URLs with sensitive content, it updates frequently, while traditional keyword solutions are also slow to update - daily at best. ContextGPT is able to update its standard and custom segments (this was a custom “LA Fire” segment) every 15 minutes, ensuring that ads were never accidentally served to new articles covering the fires.

ContextGPT found 6,000 more unsafe pages that this client could proactively avoid that would have still been allowed under the old keyword blocking approach.

Why ContextGPT Delivers Deeper Brand Safety

Advanced AI models, like those powered by large language models, are transforming how our industry approaches contextual advertising. These models do not just scan a webpage—they actually read and understand its content across multiple languages and formats, including images and multimedia. In contrast, traditional brand safety methods, such as keyword blocklists, lack this deeper level of understanding. They simply flag a page if a certain word appears without any regard for context.

Keywords with Multiple Meanings:

Consider the word “attack.” This word is used to describe political unrest and violence, but means something different when talking about sports, problem solving or heart health. A simple keyword blocklist does not differentiate between these various meanings. It could unfairly flag content that could actually align with the brand's safety requirements.

Insufficient Nuance:

Keyword blocking also struggles to capture the subtlety that often defines high-quality content. A journalist may write about an act of violence using words like “incident.” A keyword filter could easily let the article slip through, missing the nuanced context that might actually make it inappropriate for the brand.

By truly understanding the content, ContextGPT is able to differentiate between these nuanced scenarios and ensure that brand safety is maintained - finding many more brand unsafe placements while at the same time not mistakenly blocking a perfectly safe placement.

ContextGPT Blocks 5X More Unsafe Content Than Keywords Alone

Through our analysis, we were also able to determine that we had already served ~1M impressions to those 7500 URLs flight-to-date. Consider how many brands are delivering millions of impressions on unsuitable inventory if they are relying solely on keyword solutions.

Brand safety is a cornerstone of any effective media strategy, and brands place immense trust in their partners to protect their reputation by ensuring their messaging never appears alongside content that could damage their brand or alienate their audience. With ContextGPT, we elevate that trust to a new level, offering not just an enhanced layer of protection but a smarter, more precise approach to brand safety.

‍