Media in Canada June 18, 2026
A Toronto Holocaust Museum campaign shows how contextual technology can be used to map harmful content environments and intervene in algorithmic feeds.
By Andrea Hernandez
A Toronto Holocaust Museum campaign shows how contextual technology can be used tomap harmful content environments and intervene in algorithmic feeds.
Brand safety has traditionally been built on avoidance, using blocklists, keyword filters and adjacency controls to keep ads away from harmful or controversial content.
But as platforms like TikTok, Instagram and YouTube have become fully algorithmic, that approach is starting to break down. Content is no longer consumed in fixed pages or predictable environments; instead, it moves through fast, personalized feeds where misinformation, entertainment and mainstream media can appear one after another within seconds. For media planners, this loss of stable context is forcing a rethink of how targeting, adjacency and measurement actually work.
A recent campaign from the Toronto Holocaust Museum, developed with Diamond and funded by Heritage Canada, sits within this shift. The objective, according to Diamond VP of strategy Erin Banting, was to address the spread of misinformation and hate online by reaching Canadians aged 18 to 24 on the platforms they already use, and encouraging them to think more critically about the content they see in algorithm-driven feeds.
From a media perspective, the approach went beyond awareness. It was built as an in-feed intervention designed to interrupt consumption at the moment content is served, rather than simply sit alongside it.
Launched in May under the anonymous Hatetags profile, the campaign was designed to operate like a native user on social platforms, without institutional branding, so it blended into the feed rather than appearing as a traditional campaign, Banting says. It used familiar platform behaviours and engagement mechanics seen in the spread of misinformation, but redirected them to surface and challenge the same types of content.
At the centre of the execution was Silverpush, deployed through Epitaph Group. The core shift was the reversal of how contextual tools are typically used: instead of filtering out unsafe environments, the system was used to detect them as they emerged. It analysed video and audio signals, captions, comments and hashtags to flag content linked to hate speech and misinformation, and those signals were then used to inform media placement.
Paid placements were served directly within those environments, reaching users at the point where they were already engaging with or being exposed to that type of content. Targeting decisions were driven by predictive signals, keyword patterns and platform-level behaviour indicators.
The Brief:
- Brand safety is shifting from avoiding content to understanding and mapping it for
targeting. - Media is moving into feeds, with ads designed to interrupt attention inside algorithm-driven platforms.
- Context matters more than audiences, as planning focuses on real-time content environments.
- AI powers targeting, using signals like video, audio and text to identify relevant contexts.
Success is behavioural, measured by attention and “interceptions,” not just reach or impressions.
“We needed to stop the mindless scroll,” Banting says. “The rabbit hole effect that pulls people in and in five videos you are watching something you never sought out.
Our videos popped up to intercept this moment and have the viewers take pause and encouraged critical thinking.” The campaign also extended into comment ecosystems on Instagram and TikTok, engaging with posts to maintain visibility and respond to platform dynamics in real time.
In practice, Banting says adjacency was treated less as a risk and more as a signal for relevance.
Brand safety systems typically used to exclude harmful content were instead used to map and identify it, allowing media to be placed with greater precision based on actual content context rather than only audience definitions.
Platform choice reinforced this approach. TikTok, Instagram and YouTube were selected because they are driven by recommendation algorithms, not user intent.
On these platforms, users move rapidly through sequences of content, creating continuous “content cascades” shaped by platform logic. The campaign was built to enter those flows mid-stream, appearing during transitions between recommended videos rather than in fixed placements, Banting says.
Also, according to a report from Sprout Social and Meltwater Social Listening (social listening software) covering 2023-2025, hate and misinformation content has grown rapidly across these platforms, with a 350% increase on TikTok and Instagram and a 275% increase on YouTube.
In terms of measurement, rather than only tracking reach or impressions, success was defined through “interceptions”: moments when people meaningfully engaged with the content or were prompted to stop and pay attention. The campaign recorded around 1.4 million interceptions, along with strong engagement results on YouTube informational placements, according to Banting.
Overall, the work reflects a shift in contextual marketing: brand safety tools are increasingly used to map and understand content environments, not just avoid them.
Contextual targeting is becoming more active, where adjacency is used to inform media decisions rather than be excluded.
Read the full article on the Media in Canada website here.
