Buzzoola DSP (Demand Side Platform)

Buzzoola DSP is a tool for setting up and posting advertising campaigns.

2026: Implementing an AI model to increase advertiser reach in programmatic advertising

Buzzoola has implemented an AI model for semantic content analysis on internet pages. The developer announced this on February 17, 2026. The model analyzes the text of the pages of sites connected to SSP Buzzoola, determines their topics and assesses the risks by brand safety, after which it automatically decides where it is permissible to show ads. This purchase model allows you to increase the coverage of advertisers in DSP Buzzoola by an average of 25-30% without reducing brand safety requirements.

Buzzoola technology is aimed at solving one of the key problems of the programmatic advertising market: the limited available inventory when working with brands sensitive to the context of advertising. Previously, content filtering was based on static restrictions and manual moderation (white-lists), which narrowed the reach of advertisers. The updated AI model works with the text content of the pages, performs semantic classification and integrates directly into the decision layer DSP for automatic selection of sites in real time.

For the correct operation of the model, Buzzoola has developed KeyPages, an intelligent content categorizer that automatically determines the subject of web pages for context targeting and Brand Safety. How the KeyPages content categorizer works:

HTML pages are loaded and cleared of unnecessary elements (meta data, tags), after which keywords are extracted from the text using TF-IDF and graph algorithms.

Key phrases are checked by the model for relevance to the page title. Based on the title and keywords, prompts are formed for the LLM of the Qwen family, which conducts the primary categorization of content.

The results are lemmatized and mapped to the internal dictionary of DSP categories, after which a decision is made about the possibility of displaying ads.

On the basis of LLM Qwen, hundreds of thousands of pages have already been classified into 80 + categories: life style, business, politics, education, science, hobbies and interests. According to internal estimates of the company, the accuracy of determining the context of pages reaches over 85% and higher, depending on the categories of content.

The integration of the AI model into the DSP increases the available inventory, including site pages previously excluded from media plans. Advertisers expand coverage without losing brand security and increase campaign efficiency through relevant placement context and procurement optimization across key metrics. Publishers, in turn, are able to monetize more traffic on various resources, for example, on news sites or in online stores.

The rejection of rigid binding to white sheets is a logical stage in the development of online advertising. The market is moving from formal site-level constraints to dealing with the meaning and real context of content consumption. Analysis of each individual page allows you to scale advertising campaigns without compromising on brand safety and placement quality. We believe that in the coming years this approach will become a new standard for the industry, "said Vitaly Zhelyapov, CEO of Buzzoola.