Contextual AI
Contextual AI is an AI-based, custom targeting solution that performs deep content analysis to find the best contextual placements for your ads.
Contextual AI uses Natural Language Processing (NLP) to digest the content of articles across the web and understand the text the way a human would. By processing entire websites of content, the AI overcomes the ambiguities in language that over-simplified tools such as keywords can overlook, accurately assigning meaning to the words in context. (For example, does the word ‘Apple’ in this article refer to a company or a fruit?). A neural network produces an embedding vector of the content, allowing full webpages to be indexed for contextual relevance:

Leveraging this rich contextual analysis, Advertisers can build custom lookalike segments from a few sample articles or select from our list of curated contextual segments for over 300 IAB content categories. These segments have an additional cost of a $0.25 CPM that will be added to the all-in cost of media.
To leverage a Custom Contextual segment, follow these steps:
- Define your target content:
- URL Input: Find 4-5 sample articles that represent your brand’s target content. The following are our recommendations for selecting your sample articles:
- Aim for a minimum of 3-5 example pages per contextual segment. Closer to 10 may provide good results depending on the subject but won’t be necessary in all cases.
- Choose the number of examples based on how broad or narrow you want to match. If you want to go for something really broad (for example, “current events”, an IAB category), you’ll want a number of examples covering many aspects of such a broad subject (in December of 2022 I might include examples covering inflation, the Ukraine war, midterm elections, twitter, gpt-3, Donald Trump, Joe Biden and the World Cup for example). On the other hand, if you want to get very precise, for example “1960s muscle cars” you can get away with fewer examples.
- Look for pages that are text-heavy with at least 5 sentences. Low quality pages (e.g. infinite scroll, interactive, or image heavy pages) will have lower performance in the matching process. News articles, blog posts, how-to articles, etc will provide more accurate matches.
- Text Input: Enter a descriptive paragraph of text about the topic of interest to build a list of contextually relevant URLs for targeting. Text should have a minimum of 50 words and maximum of 500.
- URL Input: Find 4-5 sample articles that represent your brand’s target content. The following are our recommendations for selecting your sample articles:
- Create the Audience: In the ‘Audiences’ tab, select ‘New Contextual Audience’ and enter your sample articles. Using NLP, the sample articles are analyzed to achieve a deep understanding of content and meaning. The algorithm compares your sample articles to millions of URLs across the web to find the closest contextual matches.
- Once your custom segment has generated, you can review a sample of the matched articles by clicking on the name of the Audience in the ‘Audiences’ tab. Under the ‘Site List‘ section, you will see the URLs for your sample articles. These can be edited at any time if the ‘Allow Updates’ box is checked. Below this section you will see ‘Example Audience Data‘, this is a sample of the matched articles that your ads will target when you use this Audience. Note, this is only a sample and not a complete list of matched URLs. The total number of matched URLs will be displayed under ‘Valid Records’.
- Target the Audience on a Line: Matched articles are compiled in a custom segment that can be added to your Line Item under ‘Audiences’ using the ‘Associate Segments’ button. These audience segments contain Full URLs for precise contextual targeting. It is not recommended to layer this segment with any additional inventory or audience targeting as this may restrict campaign scale.
IAB Curated Contextual Segments
In the ‘Audiences’ tab you will find over 300 Contextual AI segments curated by the Pontiac Team for each of the IAB categories. If you click on the name of the segment, you can see the input and the sample output for this segment. The ‘Site List’ section displays the articles that were used to ‘define’ the desired content for the category. The ‘Example Audience Data’ below shows a sample of the matched articles where the ads would serve if this Audience is targeted. This is a sample and not the complete list of URLs contained in the Audience.
Note, these IAB Category segments cannot be edited and they are already available for use across all of your Advertisers at a $0.25 CPM. These can be added in the ‘Associate Segments’ section of the Line Item, under ‘Contextual’.