The Use of AI in Advertising and Marketing

AI and Advertising Marketing

Before the emergence of the internet, advertising and marketing were carried out through various traditional channels, such as television, radio, and printed media. In the age of the internet, the rise of digital advertising has caused advertising and marketing channels to become far more diversified, with both online and offline modes. Through the utilisation of both online and offline media, advertisers can better interact with potential customers. Additionally, the increasing prominence of Artificial Intelligence (AI) technology has greatly improved the efficiency of advertising marketing.

Google’s ‘Responsive Search Ads’ allows marketers to enter up to 15 headlines and 4 lines of description. Over time, Google Ads will test different combinations and learn which combinations work best. Additionally, this tool also works seamlessly with Google Cloud, which relies on AI and machine learning to deliver even more in-depth audience insights to marketers [1]. The International Business Machines Corporation (IBM) enables companies to use artificial intelligence to fine-tune marketing strategies. Their AI assistant IBM Watson locates potential audiences, selecting relevant creative content and engaging target audiences in one-on-one conversation. Due to this, companies can allocate marketing budgets more effectively, while also developing advertising campaigns around the interests of audiences [1]. Another example of AI being utilized in advertising is Coca-Cola, who uses AI algorithms to analyze when, where and how consumers talk about the brand on social media. The company studied 120,000 pieces of social media content to understand the demographics and behavioral characteristics of customers and users discussing its products [2]. Advertising and marketing incorporate multiple AI technologies, such as Multi-Touch Attribution (MTA), Recommendation System (RS), Reinforcement Learning (RL), and Large Language Models (LLMs).

The Beginning: MTA

The earliest application of AI technology in the fields of advertising and marketing is MTA. MTA is a method of marketing measurement that considers all touchpoints on a customer’s journey, assessing each the relative contribution of each touchpoint in the user conversion process. The results of the attribution will directly affect the channel budget allocation. This attribution-oriented budget allocation can help in reducing the Cost Per Action (CPA) and increasing Conversion Rate (CVR). MTA is widely used in advertising and marketing. This can be seen in the earliest related projects, Logistic Regression (LR) [3], which adopts the simple and effective algorithm of logistic regression. Additional Multi-Touch Attribution (AMTA) furthered this by borrowing techniques from survival analysis and using the hazard rate to measure the influence of an ad exposure [4], using these to improve the accuracy of conversion rate predictions. The Casual Attention Model for Multi-Touch Attribution (CAMTA) model further improves the accuracy by minimizing the selection bias in channel assignment across time-steps and touchpoints [5]. The figure below shows the path of the user’s touchpoint through different channels (Display advertising, Social Media, Paid Search) until a conversion is made. Common examples of a conversion may include an order, membership registration, a click, etc., depending on the goals of the advertiser.

Development: Reinforcement Learning and Recommender System

In the advertising and marketing landscape, the application of AI is not just limited to MTA, as RL and RS also shine. The recommendation system based on reinforcement learning abstracts the scene of the interaction between the advertiser and the user into the Markov Decision Process (MDP) model, which contains the elements: state, action, and reward, then designs advertising and marketing strategies according to this model. To summarise, the recommendation system will formulate a corresponding ‘action’ according to the ‘state’ of the user, (for example, recommend sports shoes advertisements to users who have searched for or watched sports and fitness information and videos), to get ‘rewards’ (for example, a click, or the placement of an order).

The figure above shows the process of promoting a digital payment application in an online marketing campaign using reinforcement learning and recommender system AI technology to control the distribution of cash coupons. AI technology will use previously learned strategies on large data sets to decide whether and how many cash coupons to issue for each user state, thereby maximizing revenue (the degree of promotion of digital payment applications and subsequent retention) under limited budget.

The popular short video advertisement in the picture above also applies AI technology. This is widely employed by various social media platforms such as TikTok, Kuaishou, and many more. For example, the Kuaishou team utilised a reinforcement learning-based recommender system to improve the average viewing time of videos while also meeting certain requirements (achieving a certain amount of ‘follow’, ‘like’ and ‘comment’) [6].

The Future: Large Models

The recently popularised AI chatbot ChatGPT has also shown its proficiency with regard advertising and marketing purposes. ChatGPT can be widely used in tasks such as personalising customer experience, writing product descriptions, serving as a customer service chatbot, and creating customer service feedback surveys.

Like the development of ChatGPT, the advertising and marketing model also tends to be a large model, and is not limited to the language model. This mode initially uses a huge user offline data set to train a huge model, and then puts the model in an online environment for the model to update and optimize the model through real-time interaction with users. This process involves another AI technique called fine-tuning. Of course, the ensuing challenges include the commonly seen black box problem that large models face, the memory and time consumption caused by huge computational complexity, and so on. If these problems can be solved in the future, it will bring huge benefits to advertisers and enterprises, which include increasing the conversion rate, automatically generating the optimal advertising strategy (reducing labour costs), and giving users a better experience.

The Key to Choosing AI Technology: Data

How we choose AI technology to aid us is also an important topic in the field of advertising and marketing at this stage. A key to AI models is the input, which is related to our data type. If the data is of a numerical type, use general models such as Multi-Layer Perceptrons (MLPs). If the data is text data, language models are recommended, such as LLMs. Image data is best suited for Computer Vision (CV) models. Large data sets require large models, such as GPT. Small data sets, on the other hand, call for general models, such as MTA. In general, AI technology meets our needs by learning through potential features in the data, so we must not become separated from our data during the process of using AI. Models based on user data are capable of better ensuring a more personalised experience for users, thereby helping businesses achieve expected profits.

The future of AI-based Advertising and Marketing

In the future, advertising and marketing will make use of increasingly abundant AI technology, with more user characteristics and larger model to meet our needs. Federated learning is an example of this, as it can use larger data sets to update models without violating user privacy. Transfer learning is another example, as it can undertake faster and better model learning by migrating a model trained in the source domain to the target domain.  Using various AI technologies and larger model architecture to improve model accuracy is part of the trend that AI based advertising and marketing is undergoing. As such, learning how to integrate various AI technologies and update large models, each with their own individual strengths and benefits, will become the most significant aspects of future advertising and marketing.


[1] Alyssa Schroer. AI in Marketing and Advertising: 19 Examples to Know.

[2] Erica Santiago. AI Advertising: Pros, Cons, Tips & Examples.

[3] Shao, Xuhui, and Lexin Li. “Data-driven multi-touch attribution models.” Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining. 2011. 

[4] Ji, Wendi, and Xiaoling Wang. “Additional multi-touch attribution for online advertising.” Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 31. No. 1. 2017. 

[5] Kumar, Sachin, et al. “Camta: Causal attention model for multi-touch attribution.” 2020 International Conference on Data Mining Workshops (ICDMW). IEEE, 2020. 

[6] Cai, Qingpeng, et al. “Two-Stage Constrained Actor-Critic for Short Video Recommendation.” Proceedings of the ACM Web Conference 2023. 2023.

The work described in this article was supported by InnoHK initiative, The Government of the HKSAR, and Laboratory for AI-Powered Financial Technologies.
(AIFT strives but cannot guarantee the accuracy and reliability of the content, and will not be responsible for any loss or damage caused by any inaccuracy or omission.)

Share this content


Units 1101-1102 & 1121-1123,
Building 19W Science Park West Avenue,
Hong Kong Science Park,
Shatin, Hong Kong

Products & Solutions


About Us


Copyright © 2024 Laboratory for AI-Powered Financial Technologies Ltd. All Rights Reserved.