Runway Connects AI Ad Creation to Publishing on Meta, Google and TikTok

An AI-generated ad no longer has to be the end of Runway’s marketing workflow. On September 30, 2026, the company introduced Runway Ads, which creates images and videos, publishes approved versions to Meta, Google and TikTok, and uses platform results to shape the next batch. Human approval is enabled by default; teams can choose automated publishing.

Runway is piloting the product with selected enterprise partners, including Lovable, before a wider rollout planned for the coming weeks. The system brings creative production, publishing and measurement together rather than stopping at a finished asset. Runway says it built the product on infrastructure used for its own paid marketing, particularly on Meta and TikTok.

Campaign results become the next creative brief

The starting inputs are an advertising account and a brand kit. Runway Ads uses brand guidelines, previous ads and product imagery to generate video and image variants. An automated brand check screens each version before it enters the approval queue. Approved assets can then move directly to the connected advertising platforms.

The return path is the product’s central mechanism. Runway Ads reads performance data from Meta, Google and TikTok, then generates another round around the creative that earned spend. Runway describes this as replacing a workflow spread across three or four tools, with people manually connecting the steps. The feedback is meant to guide production, not simply appear in a campaign report.

That is a concrete step beyond the workflow Runway described for Agent 2.0 in June. That earlier product could analyze ad metrics supplied by users from Meta, YouTube, TikTok or Google and create the next set of ads to test. Runway Ads builds on the company’s Agent work, adding direct publishing and a return path for platform performance.

Automation has campaign-level boundaries

Runway calls the product an autonomous engine, but its publishing permissions are adjustable. Teams can enable automated publishing by campaign or by variant type as they gain confidence. The default approval queue and automated brand check are separate steps: screening a variant against brand rules does not, by itself, remove the human approval requirement.

Other controls Runway describes

• Localization follows team-defined rules, including brand voice and terms that should not be translated. It covers on-screen text and product screenshots, not just voiceovers.
• Automatic resizing and reformatting adapt variants to different aspect ratios and channel specifications.
• Budget controls include limits on daily changes, a minimum share of new audiences and caps on retargeting.

The first performance case is Runway’s own

Runway’s business case rests partly on its internal marketing results. Since July, the company says weekly ad volume grew from 77 to roughly 900, while return on ad spend doubled. It also reports a roughly 34% increase in conversion, steady click-through rates and a 41% decline in cost per subscriber, despite higher spending.

Those are company-reported figures from Runway’s own program, rather than a published performance comparison across its enterprise customers. Co-CEO Anastasis Germanidis said combining generation, publishing and measurement gave the company fuller visibility into its paid advertising and helped it move faster. The numbers describe that internal experience; they do not establish the same gains for another advertiser.

For the pilot, Runway says Lovable is using the product to scale its creative output. Its proposed uses reach beyond that partner: ecommerce advertisers can generate ads from product feeds, while mobile-app marketers can localize a successful ad and read performance separately in each market. These are the workflows Runway is offering to automate as access broadens, not customer results demonstrating their effectiveness.