While most sectors are dipping cautiously into artificial intelligence (AI), ad tech entered with both feet two decades ago and has been circling ever since. This is a field built on AI. It's all about processing huge amounts of data at lightning speed and at scale. This combination is only possible with the help of highly sophisticated algorithms. Tracking their evolution provides a clear glimpse into the next big leap in automation.
2005: The battle for predictable clicks
In the mid-1990s, the Internet battle lines were centered around who had the best pay-per-click (PPC) business. Google, Yahoo!, and Ask were competing for search advertising supremacy, with success determined not only by the number of users but also by how accurately their PPC models could predict whether users would click on a link or, as the models evolved, take an action.
These early machine learning models digested a glut of data from search impressions and discovered subtle correlations between search terms and user behavior that could influence the probability of a click or conversion. For the first time, advertisers can buy impressions based on the likelihood of a desired outcome, and sellers can leverage this potential to increase the value of their inventory.
Such predictive models are also not isolated for search. Facebook's advertising arm has grown thanks to an even richer supply of behavioral data from its users, while companies like Criteo have launched similar models across display advertising. This meant building data collection pipes on domain-agnostic platforms rather than closed-loop search and social media. This was an incredible feat for developers working without Blueprints.
2015: Pushing privacy into probabilistic models
By the mid-teens, smartphones and apps had fragmented the online ecosystem, scattering audiences across a maze of platforms and devices. Identity Graph was born out of necessity, deploying machine learning models to stitch together user profiles across devices and channels. Without these, there is no way to know if an IP address, email login, and device ID all represent the same person.
But as the profiles assembled by these machines became more detailed, they attracted the attention of regulators, who intervened to ensure users consented to data collection. GDPR came into effect in 2018, eradicating many questionable data practices and putting consent front and center. Since we only analyze and use consented data, expanding our audience has become much more important. This process involves taking a small group of known, consented users and incorporating them into a probabilistic algorithm to extrapolate them into a larger pool of prospects. This means that advertisers and publishers can achieve viewership scale within all regulatory parameters thanks to machine learning.
2020: AI will start seeing and reading like us
In the second decade of this century, AI began adding qualitative capabilities to its quantitative repertoire. CAPTCHAs have been used to train computer vision models in conjunction with large-scale language models (LLMs), allowing machines to process, interpret, and classify media and text with near-human levels of understanding, but at superhuman speed and scale.
In digital advertising, this meant that rather than relying on manual keyword tagging, algorithms could enable contextual classification and targeting methods. This also meant that the model could predict and refine the performance of ad creatives by identifying the quality of its visuals and text and mapping them to results by tracking the performance of previous campaigns.
2022: Generative AI brings machines a mainstream moment
Previous advances in AI have focused on interpretation, but in 2022, the world will have AI that can create. Generative models use the same principles that power computer vision and LLMs, flipped to generate rather than analyze media. Midjourney made waves (and memes) with its machine-generated images, but it wasn't until the launch of ChatGPT in November 2022 that AI became mainstream.
This opens up some very exciting possibilities for agencies and brands working on digital advertising, allowing them to create large numbers of assets for multiple campaigns at once. Meanwhile, the ability to interact with digital advertising platforms through natural language prompts has made them more accessible and efficient to operate. Tools that were once considered technical are now instantly accessible, faster, and more collaborative.
2025 and beyond: Thinking machines challenge conventional thinking
AI has been driving the growth of ad tech since its inception. The question for many is: what happens next? Will the field continue to enjoy incremental improvements, or are we on the cusp of a new industrial revolution? Advocates of the latter are pinning their hopes on AI agents, autonomous models that can perform tasks on your behalf with little or no manual intervention.
At its core, the AI agent works by dividing a goal into a series of tasks, each performed by a specialized component designed to deliver a specific outcome. Agent AI is characterized by its ability to consider strategy rather than simply following a set workflow. The true realization of a thinking machine. A souped-up version of A/B testing, once the bread and butter of advertisers, allows you to consider and test multiple potential approaches and choose the best performing option.
However, these AI agents are currently isolated to specific platforms. Enabling them to reach their true potential will require cross-industry collaboration like we've never seen before. Ad tech vendors, agencies, brands, publishers, media owners, and platforms need AI-enabled APIs that allow agents to come and go without hitting a wall. Whether the next chapter in ad tech’s AI timeline is written will depend more on the industry’s willingness to collaborate on an unprecedented scale than on the technology.
