The contextual intelligence engine behind every relevant impression.
Meet Mirrors. It reads what a video is actually about, classifies the context around it, and hands you the controls to decide where your brand belongs.
Built on content. Never on cookies or user-level identity.
Language coverage refers to model content detection and processing. Processing scale refers to volume processed, not classified or activated.
One engine, from the strategy to the read-out.
Mirrors is Silverpush’s proprietary AI-powered contextual intelligence engine. It lets brands plan, activate, automate, optimise and analyse contextual campaigns across YouTube, Social, the open web and programmatic, without cookies and without user-level identity.
It is not merely a targeting tool. Targeting is one moment inside a longer job that runs from building the strategy to explaining what the strategy did.
Strategize
Build a granular, precise targeting strategy for relevance and impact.
Activate
Automate ad execution against the media plan and targeting strategy, in brand-safe, relevant environments.
Optimize
AI-driven automation of pace, margin and KPI, alerting the relevant teams when performance deviates.
Analyze
Strategic, interactive insights that show which contexts carried the campaign.
Analyse. Classify. Control.
Three steps, applied to the content rather than to the person watching it.
Analyse
The engine reads video-level and channel-level content, enriched by a creator’s off-platform presence, to understand what a piece of content actually is.
Classify
Semantic understanding turns that reading into contextual signals: topic, category, suitability, quality and language, applied consistently across the surface.
Control
Inclusion through curated lists of videos, channels, creators and publishers. Exclusion through GARM-aligned filtering, custom exclusions and creator-level controls.
The signals it extracts
Every signal below describes the content. None of them describe the person watching.
Language
What language the content is in
Topics
What the video is actually about
Categorisation
Where the content sits taxonomically
Brand suitability
Whether the environment fits the brand
Keywords
Contextual cues drawn from the content
TV-like quality
Production standard of the channel
Shorts
Short-form format identification
Performance
How the content performs on the platform
We understand content, not people.
Most of adtech spent a decade getting better at identifying individuals. Mirrors was built the other way round, and that choice is why it still works as identity signals disappear.
Classification happens before the bid, from the content itself. Nothing in the engine needs to know who you are.
What Mirrors reads
- The topic and category of the video
- The quality and suitability of the channel
- The language and the contextual cues in the content
What it never uses
- Cookies or user-level identity
- Age, gender or device of the viewer
- Profiles of individual people
How a contextual strategy is built.
The same three-level structure holds on every surface. What changes is the signal underneath it.
Tactic
The strategic contextual approach aligned to the objective, from contextual alignment and conquesting to premium content alignment, brand safety and efficiency.
Sub-tactic
A refinement of that approach, narrowing the strategy to the part of the market that matters.
Signals
The precise contextual cues that identify the content: videos and channels on YouTube, hashtags and interests on TikTok, interests and behaviours on Meta, episodes on CTV.
What sets the engine apart.
Adaptive brand safety
Exclusions are built around the brand rather than applied from one generic list, and they keep adjusting as the content changes.
Contextual and interest targeting together
Contextual classification works alongside interest signals rather than competing with them, so the strategy is not forced into one or the other.
Automation for efficiency
Setup, pacing and optimisation are automated, which is what makes a granular contextual strategy practical to run at scale.
Insight you can act on
A dashboard that reports which contexts performed, so the next flight changes what actually needs changing.
What teams actually do with it.
Four applications that come up on almost every brief.
Reach a passion, not a demographic
Align a campaign to the content a audience genuinely cares about, rather than to a proxy profile of who that audience is assumed to be.
Protect the brand without shrinking the buy
Replace blunt category blocklists with suitability decided per campaign, so safety stops costing reach.
Run the same strategy on every screen
Carry one contextual strategy across YouTube, CTV, social and the open web instead of rebuilding it per platform.
Advertise without cookies
Keep targeting precision as identity signals disappear, because the decision never depended on identifying the person.
What the engine has delivered.
Sustainability and premium-apparel contexts on YouTube.
Competitor, keyword and review contexts on YouTube.
Alcohol-safe food, wine and bourbon contexts.
Each result is reported from that client’s campaign and belongs to its market and metric. Results vary by brand, market and objective, and are not a performance guarantee.
What connects to the engine.
Mirrors is the intelligence layer. These are the products that plan against it, scale it, run it and prove it.
Minted™
Curated channel segments for scalable reach. Minted is where to show up; Mirrors decides which contexts matter.
Campaign Manager
Automates campaign setup, monitoring, optimisation and reporting, in managed and self-serve models.
Measurement
Attention with Lumen, footfall with Adsquare, and platform or incremental lift studies.
See the engine on your own campaign.
Tell us what you are planning and a Silverpush specialist will show you the contexts worth buying, on the screens that matter to you.
Your data will be processed according to our Privacy Policy.
Mirrors, answered.
Mirrors is Silverpush’s proprietary AI-powered contextual intelligence engine. It lets brands plan, activate, automate, optimise and analyse contextual campaigns across YouTube, Social, the open web and programmatic, without cookies or user-level identity. It is not merely a targeting tool: it spans strategy through to analysis.
Content and context signals only. Video-level and channel-level analysis produces signals such as language, topic, categorisation, brand suitability, keywords, production quality and platform performance. Mirrors does not build or use profiles of individual viewers.
Audience targeting profiles the person. Mirrors classifies the content, so the decision is made from what is on screen rather than from assumptions about who is watching. It uses no cookies and no user-level targeting.
Verification tells you where not to appear. Mirrors also identifies where you should appear and why that environment is worth buying, with suitability built per campaign rather than applied from one standard blocklist.
YouTube, Social including TikTok and Meta, the open web, and programmatic. The engine is DSP-ready, with activation across platforms including DV360, The Trade Desk, TikTok, Meta and Google Ads. Availability varies by product and platform.
They are complementary but separate. Mirrors is the contextual intelligence that decides which environments and moments matter. Minted is a curated channel ecosystem for scalable reach, built independently of YouTube’s predefined lineups. Minted is not video-level contextual targeting.
The engine detects and processes content in 60+ languages. That refers to content detection and processing, not to the number of languages a campaign can be planned or activated in.