When Large Language Models Meet Advertising: A New Revolution or a Bubble?

Artificial intelligence (AI) has perhaps never received such high attention. On 30 November, 2022, OpenAI officially released ChatGPT, an AI chatbot trained on the GPT3.5 architecture 1. ChatGPT is like an omniscient human, capable of reasonably answering questions about culture, history, art, technology and business. Therefore, compared to traditional search engines, ChatGPT is more user-friendly, which directly impacts search engine advertising. Additionally, ChatGPT is accurate, efficient, fluent and fast. More importantly, it can provide personalized responses based on users’ individual needs and preferences. This gives it the ability to assist businesses and individuals in designing and optimizing advertising content, further reducing the costs. With the release of GPT-4 2, this ability has been further enhanced.


1 The GPT3.5 architecture is an improved version of the GPT-3 model. GPT-3 is a natural language processing model developed by OpenAI, which has extremely high language generation and understanding capabilities. Based on GPT-3, the GPT3.5 architecture’s scale is larger, includes more parameters, uses more and richer training data, and has some improvements in the model structure.

2 GPT-4 is a large multimodal model (which can accept image and text input and generate text output). Although it is not as capable as humans in many real-world scenarios, it has shown performance comparable to human levels in various professional and academic benchmark tests.

Preliminary Applications of ChatGPT in the Advertising Industry

The advertising industry, especially digital advertising companies, has shown keen interest in ChatGPT. At Google’s I/O conference in 2023, Google officially released its new generation large language model, PaLM 2, and showcased its new search engine powered by PaLM 2, Search Generative Experience (SGE), which will revolutionize the future of online search experience. Tik Tok is also exploring the possibility of integrating large language models with its e-commerce business. Microsoft demonstrated a novel example at the New Bing release conference, asking New Bing on stage, “Can an IKEA Klippan sofa fit into my 2019 Honda Odyssey?” This indirectly solves customers’ questions and promotes sales, whether they want to buy a sofa or a car.

Traditional advertisers also explore new alternatives. McDonald’s asked ChatGPT “What is the most iconic burger in the world?” The response, ‘Big Mac,’ inspired McDonald’s to turn this conversation into a poster, which Burger King quickly emulated. These advertisements have achieved excellent results.

Challenges Faced by ChatGPT

The overwhelming media reports, the influx of funds regardless of cost, and companies scrambling to express their opinions, a revolutionary change seems to be in sight. However, as time goes on, many issues have emerged, posing real challenges to tools like ChatGPT.

1. Content Quality

After trying ChatGPT, Tom Roach, Vice President of Brand Planning at Jellyfish Global, wrote on his personal Twitter, “I recently used ChatGPT to create a list of possible ideas for a brand positioning. It created a bland and expected set of answers.” This is the biggest pain point of ChatGPT at present. It cannot generate eye-catching, novel and amazing ideas. ChatGPT may be able to help small shops design their own billboards, but can it produce slogans for large companies like “Just Do It” for Nike, or “I’m Lovin’ It” for McDonald’s? At least for now, we don’t see a great possibility.

2. Social Ethics and Laws

ChatGPT is responsible for providing the best answers, but the reality is that the best is not necessarily the most correct. ChatGPT cannot fully understand the various controversies, and even forbidden topics, of different countries, ethnic groups and societies when generating answers. If there is no subsequent human review, companies may inadvertently involve controversial content and triggering crisis, resulting in tangible losses.

There has already been a case in Germany. At the end of 2022, KFC’s app in Germany sent out a notification to customers, inviting them to celebrate the anniversary of the Night of Broken Glass 3 by ordering fried chicken and cheese. This quickly provoked public outrage and condemnation. KFC subsequently withdrew the related notification and apologized to the public, stating that the app had pushed AI-generated, unreviewed promotional content.

3. Uncanny Valley 4

Compared to real people, ChatGPT generates content at a rapid pace, and the sentences are produced in succession. This can make it evident to users that they are interacting with AI in the conversation. Similar to the uncanny valley effect, this feeling can be amplified over a long conversation, potentially leading to distrust, fear, or rejection of AI. Consequently, users may lose trust in any content generated by AI, and even develop a negative emotion to recommended products, which would be counterproductive.


3 The Night of Broken Glass refers to the pogrom against Jews carried out by Sturmabteilung paramilitary forces, Nazi Party members and anti-Semitic German civilians throughout Germany on 9–10 November 1938. The German authorities observed the event without intervening. This event is widely regarded as the beginning of massacre of Jews.

4 Uncanny Valley is a concept proposed by Japanese robotics scientist Mori Masahiro in 1970. It refers to the phenomenon where human emotional responses change as the appearance of AI and robots becomes increasingly lifelike.

Potential Applications of ChatGPT in Advertising

We must understand that, just like the connotation of “Copilot” in the name Microsoft 365 Copilot 5, a large language model is more like a co-pilot. Utilizing it to assist in our professional work is already a significant advancement.

In advertising, we need to understand that the strongest ability of ChatGPT is its rapid understanding and generation of natural language. This includes syntax, semantics, and most importantly, the understanding of context. According to the input text and its logic, ChatGPT can output content that matches grammar and semantics, and is close to user expectations. Specifically, we have the following potential applications:

  • Text Content Generation: The creation of advertising content is the most important and core requirement for clients. ChatGPT can be utilized for generating initial advertising ideas and assisting advertisers in revising existing storyboards to make them more attractive to target customers. Additionally, ChatGPT can automatically generate a set of ad slogans, detailed copy, and even storyboards for video ads that fit the brand image, based on specific requirements provided by advertisers.

    We can see that the storyboard below covers various scenarios, each generated to highlight the desired functionality. Each scenario is accompanied by appropriate narration. In contrast, many existing laptop ads solely rely on close-up shots with subtitles describing different functions, which is obviously more monotonous.
An example of using ChatGPT to generate an ad storyboard
  • Advertising Data Analysis: Another field where ChatGPT excels is data analysis, extracting key information that users need. After an ad launch, advertisers need to constantly track the performance and ROI of the ad, and adjust the ad platform, frequency etc. Now we can use ChatGPT to complete this demanding and repetitive task. Here is an example using ChatGPT to identify the three products with the highest click-through rates. The process is intuitive and fast, and the results are highly accurate.
An example of using ChatGPT for data analysis

  • Intelligent Customer Service: Prior to the introduction of ChatGPT, many different chatbots had been applied in intelligent customer service, but none could provide a user experience like a real human conversation like ChatGPT. ChatGPT can help users solve problems and provide support just like real customer service personnel. In addition to answering user questions, it can also collect specific customer needs and provide personalized services to individual customers, recommending comprehensive solutions provided by the manufacturer.

To conclude, compared to human labor, the biggest advantages of ChatGPT are fast response, low cost, and high degree of automation. In current economic situation, cost reduction and efficiency improvement are the focuses of many companies, and ChatGPT aligns perfectly with these objectives. It can be predicted that an increasing number of companies will allocate resources to adopt the technology.


5 Microsoft 365 Copilot is a data and AI assistant tool based on Large Language Models (LLM) and Microsoft Graph. It can transform user’s natural language input into efficient productivity tools.

The Future of ChatGPT in Advertising

The reshaping of traditional businesses by large language models is inevitable. Looking back at the history of human society, using machines to replace human labor for repetitive and low-tech jobs has always been the direction of social progress. ChatGPT’s problem is that it is just like a “new kid on the block”, and it still has many areas that need improvement. Therefore, it is essential for businesses not to blindly rely on it, but rather develop a comprehensive understanding of its capabilities and limitations in order to utilize it appropriately.

Large companies like Coca-Cola chose to directly cooperate with Bain & Company and the developer of ChatGPT, OpenAI. Coca-Cola stated that they will improve their business level through cutting-edge AI technology. Bain & Company also pointed out in their article that they will combine their rich digital implementation skills and strategic experience with OpenAI’s AI tools and platforms to help global customers discover and apply the value of AI.

It can be predicted that as large language models like ChatGPT gradually improve, certain positions may be replaced. However, possessing the knowledge of how to utilize these tools correctly and efficiently will stimulate the emergence of a new industry, ultimately creating a large number of new job opportunities. Before that, it is important for us to remain rational, stay informed about industry changes, be fully prepared, and embrace the arrival of a new era.

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

Address

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

Products & Solutions

People

About Us

Address

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