Meta confirmed during its Q3 2025 investor conference that its automated advertising ecosystem Advantage+ had now reached a run-rate of $60 billionor approximately 51 billion euros. This rapid growth illustrates a structural shift with campaigns optimized end-to-end by artificial intelligence models.
The end of manual targeting
With Advantage+, the platform automates almost all of the levers formerly entrusted to media agencies: choice of audiences, budget allocation, placements, and optimization of results. The tool, integrated into the “Sales”, “Leads” and “Applications” modules, continuously learns performance to adjust parameters in real time. According to Kenneth Dorell, Director of Investor Relations at Meta, it is an “end-to-end” system capable of selecting the best combination of channels and formats to maximize ROI.
This automation is reinforced by GEM And AndromedaMeta’s new advertising recommendation models. GEM would now be four times more effective to generate advertising performance with equal resources, while Andromeda improved the quality of Facebook ads by 14% after its deployment on feeds and Reels. In 2026, these architectures will be generalized to Instagram and Messenger.
The agency facing new arbitration
Automation significantly transforms the role of media agencies and advertisers, so the value is no longer in the operational management of campaigns, but in the creative strategy, proprietary data and performance measurement. Suffice it to say that agencies that have not taken the turn early enough will be faced with a rapid erosion of their economic model. Teams historically focused on purchasing space and manually optimizing budgets are seeing their know-how replaced by algorithmic architectures that learn on their own, continuously. Margins on operations management are collapsing, while value is shifting towards design of data ecosystemsthere production of creative assets adapted to AI and the ability to interpret signals from platforms.
To survive, agencies will have to reposition themselves in technology integrators and performance strategistscapable of combining the brands’ proprietary tools with the closed optimization systems of Meta, Google or Amazon. Those that do not do so will become simple peripheral providers in a market now driven by algorithms.
Another point is that budgets, previously fragmented between networks, formats and audiences, are converging towards a logic of management by objectives (“sales”, “app installs”, “leads”). Meta thus controls the entire optimization loop, from delivery to result.
For agencies, differentiation will involve the ability to:
- design creative assets adapted to distribution AI,
- exploit the performance signals provided by the platform,
- articulate brand proprietary data with Meta’s advertising APIs.
The boundary between creation and media planning is becoming blurred, algorithms choose where to broadcast, creatives must now understand how they “read” content.
Towards an “AI-first” model of advertising performance
Meta also deploys multimodal models to better understand visual content. According to VP Finance Chad Heaton, these models “specify the keywords and themes of videos and posts, improving the correspondence between interest and distribution”. In other words, the platform becomes capable of automatically associating an advertisement with a relevant context without explicit targeting.
For brands, this changes the grammar of performance, behavioral signal replaces socio-demographic targeting, and content becomes the strategic visibility variable. The algorithm now determines the meeting between message and audience.
A concentration of advertising power
By internalizing data, models and optimization, Meta further reduces the mediation space left to agencies. The advertising ecosystem is concentrated around a few platforms with the volume of signals and AI infrastructure necessary for learning.
If advertisers gain immediate effectiveness, it is at the cost of structural dependence. Advantage+ campaigns perform well, but within an environment opaquewhere access to segments and control over arbitrage remain limited. This total integration, already visible at Google and Amazon Ads, devotes new hierarchy of the advertising market : agencies are sliding from the role of operators to that ofarchitects of data and customer identity ecosystemswhile brands find themselves managing their growth within a framework defined by the platforms themselves.
The question is no longer just that of performance, but of marketing sovereignty : will this model still allow brands to develop sustainably, or does it aim above all to make them captive of a system that they no longer controland where the customer is not king.
