And how AI is quietly changing the rules of performance
For many SMEs, Meta (Facebook and Instagram) has historically been a powerful growth channel. It offered strong reach, highly targeted audiences and the ability to scale results with relative predictability.
Over the past 12 to 18 months, however, something has changed. We are seeing more businesses investing consistently but struggling to scale results, experiencing volatile performance across campaigns and losing confidence in targeting and optimisation. There is a growing sense that the platform is increasingly determining outcomes, rather than the advertiser guiding them.
If that feels familiar, it is not just your campaigns. It is the platform itself.
The shift: from targeting to AI-led delivery
Meta is progressively transitioning toward a more AI-driven advertising model.
While advertisers currently still have the ability to define audiences, control placements and structure campaigns manually, the platform is increasingly encouraging a different approach – one where the advertiser provides the objective, creative inputs and budget, and the system determines how delivery is optimised.
Campaign types such as Advantage+ are leading this shift. They are designed to simplify execution, reduce manual inputs and leverage data at scale, with targeting, placement and optimisation decisions increasingly handled by AI.
While manual control remains available, it is becoming less central to how the platform is designed to perform. In many accounts, automated formats are now being prioritised, both in recommendations and in performance benchmarks.
Businesses are not losing control overnight, but they are being guided toward a model where control is exercised less through settings, and more through the quality of the inputs that shape the algorithm.
The challenge: reduced visibility and “black box” behaviour
One of the most significant implications of this shift is a reduction in transparency.
Advertisers are increasingly operating within what is often described as a “black box” environment. Decisions around targeting, placement and optimisation are made dynamically by the platform, with limited visibility into how those decisions are reached.
This is not just a technical limitation. It changes how performance can be understood and managed.
Traditional levers such as detailed audience segmentation, placement control and isolated testing are being replaced by broader, AI-led delivery. Reporting focuses more on outcomes than on the drivers behind them.
For many SMEs, this creates a sense of dependency. Performance is visible but the reasoning behind it is less so.
Why results are becoming less predictable
AI-driven campaigns optimise based on available data and signals. When those signals are strong and aligned to the business objective, performance can improve significantly.
When they are not, the algorithm will still optimise – but not necessarily in the way the business intends. This is particularly evident in campaigns with specific or niche objectives.
Scenario: Franchise recruitment for a national network
In this example, the objective is highly defined. The business is seeking potential franchisees with specific experience, financial capacity, mindset and values, within clearly defined geographic regions.
The campaign is structured accordingly, with messaging focused on business ownership, a dedicated landing page and audience signals aligned to commercial intent.
However, in practice, Meta’s AI begins directing spend toward broader audiences: consumers of the service, general interest groups and users outside the intended profile.
From the platform’s perspective, these users are easier to identify, engage more readily and generate lower-cost signals. The algorithm is optimising for efficiency. The business, however, is optimising for suitability.
The result is that lead volume may increase, but lead quality declines. Sales teams spend more time filtering, cost per qualified lead rises and confidence in the channel begins to erode.
This is a clear example of where the platform is functioning as designed, but not aligned to the commercial objective.
Why this happens
Meta’s AI relies heavily on historical performance data, engagement signals and conversion events. In niche campaigns, conversion volumes are lower and signals are less frequent. As a result, the algorithm broadens its search.
“Similar” users are identified based on behaviour, not necessarily commercial intent.
This creates a widening gap between who the platform can reach efficiently and who the business actually wants to attract.
Can this be corrected?
One of the most common approaches to improving performance is the use of first-party data. Businesses can upload customer lists, qualified leads or CRM segments, which Meta then matches to user profiles to create custom audiences and lookalike audiences.
In principle, this helps guide the algorithm toward more relevant users. However, while data can improve performance, but it is not a complete solution.
Meta can work with relatively small datasets, sometimes from as few as 100 users. However, larger datasets tend to produce more stable and reliable signals for optimisation.
Match quality however is a separate challenge. It depends less on volume and more on how well the data aligns to Meta profiles. In many SME environments, databases are built around work email addresses which are less likely to match personal social accounts, reducing the effectiveness of the audience. The result is that a portion of the data cannot be effectively matched, reducing the strength of the signal being fed into the algorithm.
Even with strong datasets, AI will continue to expand beyond initial audiences in pursuit of scale which means that data helps guide performance, but it does not fully control it.
What actually improves performance in this environment
As control shifts away from targeting, performance is increasingly driven by the quality of strategic inputs.
– Creative becomes the primary filter
With broader targeting, creative plays a more critical role in defining who engages. It must attract the right audience while discouraging those who are not aligned.
– Stronger signals improve optimisation
AI responds to what it can measure. Structuring enquiries, qualifying leads and aligning landing pages to real expectations helps shift optimisation from volume to value.
– The wider ecosystem shapes results
Campaign performance is no longer driven by ads alone. Brand clarity, website experience, messaging consistency and conversion pathways all influence how effectively the platform can optimise.
The bigger shift: Meta Ads is no longer a standalone channel
Meta can no longer be treated as an isolated advertising channel. Its performance is shaped by a broader system in which multiple elements interact.
Brand defines the audience relevant to the business. Creative signals determine who should engage. Data informs how the platform learns. The user experience determines whether interest converts into action.
When these elements are aligned, AI can amplify performance. When they are not, automation increases inefficiency.
Meta’s move toward AI-driven advertising is not temporary. Automation will continue to increase while manual control will continue to reduce. As a result, the gap between average and high-performing campaigns will widen because, while AI can optimise execution, it cannot define positioning, audience, value or commercial intent.
Those inputs remain the responsibility of the business and they matter more now than ever before.
Milestone-Belanova
At Milestone-Belanova, we work with growth-stage and mid-market organisations to navigate these shifts, aligning brand, creative and digital strategy so that AI-driven platforms such as Meta deliver meaningful commercial outcomes, not just activity.
Through our Relentless Performance™ pillar, we focus on strengthening the inputs that now matter most, including positioning, creative, data and conversion pathways, ensuring AI-driven systems are guided by clarity rather than left to optimise in isolation.
If your Meta campaigns are active but not delivering the quality or consistency you expect, we can help you reframe how performance is structured, moving beyond platform settings to strengthen the strategic inputs that drive results.
We welcome a conversation about how to bring greater clarity, alignment and control to your campaigns, so that automation works with your commercial objectives, not against them.
