The advertising industry has a predictable relationship with AI: first denial, then panic, then overcorrection toward the belief that everything has changed. The truth is more useful than either extreme. Some things have changed substantially. Some things have not changed at all. Understanding which is which determines where you spend your strategic attention.
What has already changed
The most significant change in paid social over the past three years is not in any single platform feature. It is in where the human decision-making actually matters. Manual bid management, granular audience segmentation, and campaign structure optimisation have been substantially absorbed by platform automation. Smart Bidding on Google, Advantage+ on Meta, automated budget allocation across ad sets: these are not experimental features. They are the default mode of operation for accounts at scale, and they consistently outperform manual management in controlled tests.
The creative layer has been affected differently. AI-generated image variations, automated copy testing, dynamic creative optimisation, and platform-native generative tools have expanded the volume of creative output available without proportionally expanding the strategic insight behind it. The gap between creative volume and creative quality has widened. Brands that mistake volume for strategy are producing more content that performs worse than before.
Targeting has been restructured by privacy changes and by the platforms’ increasing reliance on first-party signals. The iOS 14 privacy changes removed a significant portion of the behavioural data that made Meta’s audience targeting so precise between 2015 and 2021. The platform has partially compensated through modelled audiences and broader signals, but the precision of demographic and interest targeting has declined. The platforms’ response has been to lean harder into conversion optimisation: show the ad to whoever is most likely to take the action, rather than to whoever matches the audience specification.
What has not changed
The offer is still the thing. No algorithm can make an overpriced, undifferentiated, or poorly positioned product perform. The platforms are exceptionally good at finding people who might convert; they cannot manufacture desire or create relevance where none exists. Accounts that perform well in an AI-optimised environment have offers worth advertising, not just well-configured campaigns.
Creative insight has not been automated. The ability to identify what an audience believes, what they fear, what they want to be true, and how to speak to those things in a way that is both honest and compelling is not something that machine learning produces. AI tools can generate a hundred variations of a concept. They cannot generate the concept. The human insight that precedes the brief is still entirely irreplaceable.
Measurement has not become simpler. The proliferation of channels, the decline in cookie-based attribution, and the introduction of modelled conversions have made the measurement question more complicated, not less. Knowing what is actually working, as opposed to what the platform’s attribution model credits, still requires independent thinking and often independent tooling.
The privacy tension
The direction of travel on data privacy is clear. Regulatory pressure, browser-level changes, and platform-level responses to that pressure have reduced the signal fidelity that paid social advertising depended on for its precision. This trend is not reversing. First-party data, which your business owns and controls, is becoming proportionally more valuable as third-party signal diminishes.
Brands that have invested in building first-party data assets: email lists, CRM data, loyalty programme data, are in a structurally better position than those who relied entirely on platform-owned audience signals. The ability to upload customer lists, create lookalike audiences from owned data, and exclude existing customers from acquisition campaigns all depend on first-party data quality. This is an area where business investment in data infrastructure directly translates to advertising performance.
Where human strategy wins
The automation of bidding and targeting has not eliminated the need for human strategy in paid social. It has changed what that strategy needs to address. The questions that matter most now are not operational, they are architectural: which channels, in what combination, for which audience stages, measured against which outcomes, with what creative approach, and with what offer.
These are not questions the platform’s algorithms answer. They are not even questions the platforms ask. They are the strategic layer that sits above the machine, and they require a human who understands the business well enough to make good decisions about how to configure the system.
The advertisers who will perform best in an AI-optimised world are not the ones who have mastered the most platform features. They are the ones who can ask the right questions about what the business needs, build a measurement framework that answers those questions honestly, and give the platform’s automation the structural context it needs to work effectively. That combination has always been rare. It is now the primary source of competitive advantage in paid media.
0 Comments