Behind Meta’s AI ad push lies a bigger shift in platform power

A quiet but consequential restructuring is happening inside the world's most powerful advertising machine. It does not show up in Meta's headline numbers, which continue to impress: global advertising revenue crossed $200 billion in FY2025, and average revenue per user reached an all-time high of $57.03. What it shows up in is the conversation among performance marketers, D2C founders, and app advertisers who have begun noticing something unsettling: the controls they once relied on are disappearing, results are growing harder to predict, and explanations are growing harder to find.

The advertising platform they thought they understood is being quietly, systematically rebuilt from the inside out and the blueprint belongs entirely to Meta. At the centre of this transformation is an accelerating shift toward what Meta calls AI-native advertising infrastructure. Advantage+, Meta's flagship automated campaign system, now handles targeting, bidding, placement, and creative decisions simultaneously, with minimal human override.

In January 2025, Meta removed detailed targeting exclusions entirely, citing a 22% improvement in campaign performance without the feature. In March 2026, it rolled out a major update to its AI delivery system, transitioning from auction-based placement to outcome-based optimisation that predicts downstream conversions rather than just clicks. And in late April 2026, Meta launched its official Model Context Protocol (MCP) server for ads. This enables AI assistants like ChatGPT and Claude to connect directly to advertiser accounts and manage campaigns through natural language, making Meta the first major platform to offer full read-and-write access from day one of such a rollout.

Each of these moves, taken individually, reads as a product improvement. Taken together, they represent something more structural: the steady migration of advertising decision-making from human operators to algorithmic systems, and the corresponding erosion of the transparency that once gave advertisers confidence in what they were buying.

Read On: Meta expands AI business assistant to global advertisers

When the Numbers Stopped Making Sense

The signals from the ground have been hard to ignore. Ovais Ahmad, a performance marketing consultant with visibility across dozens of ad accounts, flagged a pattern that caught attention across the industry. Around March 26 and again around April 9 this year, performance dropped significantly across multiple accounts at different spend levels and verticals. "Some ad accounts dropped in performance by up to 50%." The timing was striking: Meta officially released its MCP server for ads on April 29, and industry speculation quickly connected the dots between pre-launch infrastructure changes and the delivery disruption.

Pankaj Sharma, who works closely with agencies and performance teams managing Meta and Google spends, places the performance anxiety in a broader market context. The benchmarks that advertisers built their acquisition models on have shifted materially. Customer acquisition costs that once sat at Rs 400-500 are now benchmarked at Rs 100, and advertisers who were hitting Rs 100-150 are now seeing those numbers inch toward Rs 250-300. "There is saturation in audiences and high expectations from marketers," Sharma said. "Meta used to do wonders, but now it is not working — possibly people are not getting those leads, people are not filling in those leads."

Sharma also pointed to the cyclical reality of the Indian market. April to June is a low-spending season, the start of the financial year typically sees softer numbers, and geopolitical uncertainty suppresses consumer confidence. "Right now people are not buying as much as they used to. That also creates sentiment within the market and defines the performance of campaigns."

An industry source with broad visibility across the adtech ecosystem, who spoke to e4m on condition of anonymity, offered a more grounded read and shared a few possible reasons for the performance decline. Meta, they observed, might have been "cleaning up more irrelevant ways of running ads," possibly removing fake profiles and tightening delivery norms. The expert noted that after speaking with peers in his network, there were no official remarks on the matter. "Meta keeps changing its delivery and AI systems regularly, so some accounts are impacted more than others," the source said, noting that stabilisation typically follows—though the timeline varies. "Such announcements create high expectations, but it takes time to stabilise. A pre-testing phase could be a reason."

These are not mutually exclusive explanations. Market saturation, seasonal softness, audience migration from Facebook to Instagram, and active AI infrastructure changes can all operate simultaneously. The problem is that advertisers currently have no reliable way to isolate which factor is driving what.

The Visibility Trade-Off at the Heart of AI Advertising

This is, in many ways, the defining tension of Meta's current advertising architecture. The platform is asking advertisers to exchange granular control for aggregate efficiency, and the deal is harder to evaluate when the audit trail has been stripped out.

Gopa Menon, COO and Co-founder of Theblurr, puts it directly. "Meta has been step by step removing manual levers such as audience targeting, bidding controls, placement selection, and replacing them with automated systems. MCP is the latest step. You tell the machine your goal, hand over your account access, and have to trust it to deliver." The upside is scale and speed that human teams cannot replicate. The downside is structural. "What you lose is visibility into why decisions are being made, and the ability to course-correct with precision when things go wrong. It is a black box. The control you give up is real; the scale you get in return is also real. but it comes with Meta's algorithm sitting and guiding the entire process."

The architecture of Meta's MCP also carries implications beyond campaign management. Menon points to a dimension that most advertisers are not yet thinking about. "Every AI agent action on your campaigns flows through Meta's authenticated infrastructure. Meta becomes the connective layer between your business data, your AI tools, and your campaign execution. Over time, Meta accumulates behavioural intelligence from that position knowing what signals advertisers respond to, what data they prioritise, and how they make decisions. This compounds into something far more valuable than just ad inventory."

This compounds into a structural question about who actually owns the learning in an AI-mediated advertising relationship. When an advertiser's campaign data, audience signals, and creative performance history all flow through Meta's infrastructure, the institutional knowledge being built belongs, increasingly, to the platform.

Read On: WhatsApp, Meta AI app to get Incognito Chat feature in coming months

India: The Market That Cannot Afford to Wait and Watch

For India specifically, the stakes carry additional weight. Meta's India gross advertising revenue grew 29% year-on-year to Rs 29,392 crore in FY25, significantly outpacing Google India's 11% growth over the same period. An industry analyst tracking Meta's India monetisation notes that the company's revenue per user in India has surged 210% since 2019, against user base growth of just 70% over the same period. By 2025, indexed revenue per user is projected to cross 300 on comparable scale measurements, a number that underscores how India has evolved from a volume market into a monetisation engine.

The implication is direct: Indian advertisers are spending more, not less, on Meta's platforms. But the performance feedback they receive in return is increasingly opaque. If AI-driven delivery optimisation works better for high-ARPU markets where signal density is richer and Meta's training data is more developed, the question of whether Indian advertisers receive proportionate value from the automation becomes commercially significant.

The MCP Moment and What Comes After

Meta's MCP launch is not arriving in isolation. Google Analytics released its MCP integration in July 2025. Google Ads followed in October 2025, though with read-only access initially. Amazon Ads opened its beta in February 2026. The IAB Tech Lab Agent Registry listed 10 active MCP entries as of March 2026. What distinguishes Meta's version is full read-and-write access from day one, no developer app required, an OAuth flow that eliminates setup friction, and compatibility with tools like ChatGPT and Claude from launch.

For the martech and adtech ecosystem, this has immediate commercial consequences. Menon is direct about it. "A significant slice of the martech industry existed because Meta's API was technically hard to access. That friction created commercial space for bid managers, reporting tools, audience platforms, and campaign dashboards. MCP removes that friction entirely for free and for every advertiser globally. Tools that were charging to solve an access problem now have no access problem to solve." He adds that MCP's current limitations are that it can adjust campaign budgets but lacks the context to decide when or how much to scale, and cannot automate creative testing for fatigued ads. This means specialist tools that add genuine strategic intelligence remain relevant. "The ones that survive will be those adding genuine strategic or creative intelligence on top of the connection, not just the connection itself."

For advertisers themselves, the MCP era promises faster campaign launches, smarter audience testing, automated creative variations, and leaner teams with higher output. The real differentiator, as multiple practitioners have noted, will not be whether a brand uses AI - it will be how quickly brands integrate AI into acquisition systems before competitors do. Performance marketing is in active transition from manual execution to AI-assisted growth, and the window for early-mover advantage is compressing.

The deeper question, however, is not about tools. It is about what kind of relationship advertisers want to have with the platforms managing their growth. Meta's direction of travel is unambiguous, yet more automation, less manual control, faster iteration at scale, and an infrastructure that positions the platform as an indispensable layer between brands and their customers. The performance data, for many advertisers, continues to justify the investment. But the ability to explain, audit, and predict that performance is eroding in step with the very automation that is supposed to deliver it.

In the age of the algorithm, the most important question a performance marketer may face is no longer how to optimise a campaign. It is whether they still can.