Schibsted Adds AI Optimization to Direct-Sold Ads to Compete With Meta

Schibsted Media Group is adding publisher-specific ad optimization to its direct sales pitch, seeking to win or defend advertiser budgets that might otherwise go to Meta. The Norwegian media company is working with London-based Nodals AI to build campaign models using its own audience data, and is packaging the technology into direct-sold native advertising products priced on a cost-per-click basis.

The effort reflects a broader challenge for publishers: reach and local market knowledge can attract advertisers, but brands increasingly expect the performance tools offered by large platforms. Schibsted’s director of ad data capabilities, Christer Ljones, said the company sees custom algorithms as a way to demonstrate campaign effectiveness alongside its audience scale.

Competing beyond audience reach

Schibsted owns VG, Norway’s largest digital news publisher, and Sweden’s Aftonbladet, among other media properties. It also acquired Swedish broadcaster TV4 and Finnish broadcaster MTV Oy from Telia last year. Together with its local newspaper network, the properties give Schibsted roughly 50% audience reach in the markets it serves, according to Ljones.

That scale has long supported the company’s direct advertising business. Direct sales account for about two-thirds of Schibsted’s ad revenue, with programmatic making up the rest. Online display advertising and branded content contribute most of the ad revenue, in roughly equal proportions. The balance differs across the portfolio: VG and Aftonbladet depend more on advertising, while local publications rely more heavily on subscriptions.

But audience size alone is not enough for advertisers focused on measurable outcomes. Meta’s optimization tools, including Advantage+, have helped establish expectations that platforms should help improve campaign results, Ljones said. Nodals founder and CEO Aly Nurmohamed said advertisers increasingly compare publisher proposals with platform performance, rather than evaluating media chiefly on a cost-per-thousand-impressions basis.

Schibsted has regarded Meta as a competitor, Ljones said. The company is also among publishers challenging Meta’s planned terms-of-service changes, which would have required publishers active on its platforms to share data for training Meta’s AI models. Schibsted and 180 other publishers joined the Norwegian Media Businesses’ Association in a lawsuit filed at the end of October. The Oslo District Court then temporarily froze the change in Norway while its legality is examined. Meta later paused the update across regions, according to the account provided by Schibsted.

Using publisher data to build campaign models

Nodals’ approach is to create optimization models for publishers using information they already collect about their audiences. Nurmohamed described inputs such as logged-in user data, reading and scrolling behavior, and interactions with content and advertisements. The model uses those signals to predict which users may be more likely to click or purchase in response to a particular advertiser’s campaign.

For Schibsted, the proposition rests partly on its direct relationship with readers. Ljones said search accounts for only a single-digit percentage of traffic on the company’s largest sites, leaving it less dependent on search referrals than publishers that rely heavily on that source. The company argues that its owned properties and first-party audience knowledge can support targeting and optimization without relying on third-party data vendors or identity graphs.

Nurmohamed said that because publishers own the models and the data used to train them, the approach should be less exposed to Apple’s efforts to restrict third-party audience data vendors and identity graphs. That is Nodals’ characterization of the model’s potential; the source material does not provide independent performance comparisons or quantify the effect of Apple’s restrictions.

Schibsted began testing Nodals’ technology last summer and has since incorporated it into direct-sales activity, Ljones said. The company is developing a line of native ad formats with AI optimization built in. The products are sold on a cost-per-click basis, shifting the pricing conversation from impressions toward clicks and campaign outcomes.

Testing incremental advertiser spend

Some traditional advertisers, including grocery retailers serving Norway and Sweden, have responded to the offer, according to Ljones. Those advertisers are seeking a combination of performance and brand awareness. Schibsted does not expect its system to outperform Meta on average, he said; its commercial aim is to deliver results at meaningful scale for particular advertisers and campaigns.

The publisher is pitching the products in part to clients that have increased spending on Meta but are seeing returns or conversions flatten or decline. Ljones described an initial proposal of shifting 10% of a client’s Meta spend to Schibsted, with the goal of maintaining much of the conversion volume while generating additional results through the publisher.

Schibsted says early tests are showing encouraging signs, but it has not disclosed campaign-level figures or a broader measurement of results. Ljones said the company is focused on bringing in new spending from current clients, protecting existing budgets and reconnecting with advertisers it had lost. The strategy is not to replicate Meta’s scale across the board, but to use local reach, first-party data and optimization to make a case for incremental direct investment.