> For the complete documentation index, see [llms.txt](https://docs.dstillery.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.dstillery.com/data-and-audience-products/data-and-audiences/how-our-multimodal-ai-works.md).

# How our multimodal AI works

See how Dstillery connects behavioral, language, and partner signals to understand real audience intent.

{% embed url="<https://www.youtube.com/watch?v=3NKepKcsQKw&t>" %}

{% hint style="success" %}
**Build once. Activate across supported channels.**
{% endhint %}

### See the complete customer journey, not isolated channels

Modern consumers move between channels. They search on mobile, read reviews on desktop, watch streaming TV, and buy in-store.

Traditional programmatic targeting treats these moments separately. Dstillery’s multimodal AI connects them into one understanding of audience intent.

### The power of a shared data language

Dstillery's Multimodal AI uses custom vector embeddings to map different data types into a shared intelligence layer. Keywords, web visits, CTV habits, and purchase data can then be evaluated together

This produces a richer picture of people who are in-market for your brand.

### Powered by four foundational signal pillars

{% stepper %}
{% step %}

#### Web and mobile visitation

Behavioral patterns across more than 400 million devices reveal intent at scale.
{% endstep %}

{% step %}

#### Opted-in search and intent panels

About 2 million privacy-safe users show the search and browsing sequences that lead to buying decisions.
{% endstep %}

{% step %}

#### LLM and language intelligence

Language models connect cross-channel relationships. They can link CTV viewing to web research or turn a written brief into targetable patterns.
{% endstep %}

{% step %}

#### Enriched partner data

Specialized signals add depth for CPG, healthcare, B2B, and retail use cases.
{% endstep %}
{% endstepper %}

Learn how these signals work together in [Our data & DS-1: how they fit together](/data-and-audience-products/data-and-audiences/dstillery-data-and-ds-1-how-they-fit-together.md).

### Start anywhere. Activate everywhere.

The shared intelligence layer gives you flexibility in how you begin and where you deploy.

1. **Flexible seed inputs:** Start with a keyword list, URL, product category, CRM list, CTV exposure, or plain-language brief.
2. **Always fresh:** Custom predictive models reach your DSP within 24 hours. They re-score daily to help reduce data decay.

### Key benefits

* **Omni-channel consistency:** Build your audience model once. Deploy it across CTV, audio, display, and web without rebuilding segments.
* **Zero data decay:** Models train on live behavior, reach your DSP within 24 hours, and re-score every 24 hours.