AI Generated Advertising Content

Due to recent progress in the fields of Artificial Intelligence (AI) and Machine Learning, many of the creative tasks within advertising, such as writing ad copy or ad image selection, are increasingly being performed by machine rather than by humans. The rise of AI generated content stands to shake the advertising world, as some professional roles become obsolete. The ways that consumers and brands interact are also rapidly changing as a result. To understand this phenomenon, we must delve into the benefits and pitfalls of AI generated content.

Some advertisers dream of a time when they can enjoy a three-hour work week, utilizing a myriad of AI tools to streamline and automate their workflows to extreme lengths. While this particular scenario isn’t very likely to happen, it’s not hard to understand the desire: This would be quite the leap from the current day-to-day slog that many advertisers find themselves struggling through. In the digital era, marketing departments must churn out dizzying numbers of variations of digital ads for the various social media platforms currently popular, each with slightly different imagery and calls to action. Wouldn’t it be nice to automate this process, and let robots handle the boring bits? Well, that might seem like some manner of science-fiction futurism, but it is actually a possibility today.

AI can be used to completely generate both the advertising copy and the visual imagery for the ad, and when combined with customer profile data, AI can even customize the ad to be more persuasive to that particular viewer. These ads do not exist before the target consumer is ready to view the ad, then in an instant an ad is automatically generated just for that particular viewer. The AI takes into account the viewer’s interests, behaviors, and demographics. The result is a very tailored communication, which will likely be more effective than a traditional one-size-fits-all ad. It also has the benefit of saving the brand a fortune in advertising costs. All the man hours that would have traditionally been spent in crafting the ad, and then creating endless derivatives for every possible platform, was all accomplished without any man hours spent at all (well, besides the initial setup of the AI campaign that is).

Multiple AI technologies can be used in tandem for particularly creative results. Companies like DataGrid and Rosebud.ai have developed AI technology that allows advertisers to utilize completely artificial models that are almost indistinguishable from real, human models. These virtual models can be used as actors in commercials, or as fashion models for brands. You could even showcase fashion products on a generated model that looks identical to the viewer, letting the viewer know how those items would look on them specifically. The possibilities are almost endless. Albert.ai is another AI brand, one that autonomously plans and executes paid search and social media campaigns. Tech company OpenAI (co-founded by Elon Musk) launched the AI tool “GPT-3,” which can write copy so well that it’s hard to tell that the text wasn’t written by an actual human. Using tools like these, brands can save advertising costs, allowing smaller brands the ability to make advertisements that rival the quality of larger brands. They will also be able to experiment with more creative advertising, since the cost to experiment will be much lower than using traditional methods.

However, the technology isn’t all AI generated roses. For example, according to Google’s Search Advocate John Mueller, content automatically generated with AI writing tools is considered spam and against webmaster guidelines. It is possible platforms will begin banning AI generated content in the future, which would certainly dampen the technology’s potential. As of now though, no ban exists for this emerging technology, and platforms like Google are unable to detect if ads were AI generated or human created. Another concern is that AI generation tools might lead to a stall in the advertising job market, as many traditional roles are replaced with AI counterparts. However, it is also possible that freeing advertisers from the more tedious aspects of ad creation will have a positive effect, allowing them to spend more time and resources on more creative pursuits. This could lead to higher quality advertising for consumers, and more high-level positions for prospective advertisers. A more pressing fear is that AI generated content might lead to empty, uncreative, repetitive advertising. This could further crowd an already crowded market, and could erode brand trust with consumers, if the technology is used to generate low-quality content. 

Despite these potential pitfalls though, the future of AI generated content is going to be too lucrative to ignore. Advertisers that learn how to put these new tools to work for them will enjoy an advantage over brands who aren’t able to capitalize on the power of generated content. And as this technology matures, that will only increasingly be the case. Even as it stands now, if some parts of this paper were AI generated using the tools currently available, would you be able to tell?  

Machine Learning in Advertising

Innovations in Machine Learning are having a radical effect on how consumers and brands interact. Machine Learning also greatly expands the toolkit companies have, both to gather analytics, and to use that data to craft a deeper understanding of their consumers. Better customer profiles lead to better sales and happier customers, but there are pitfalls to be wary of too.

Machine Learning, a subfield of artificial intelligence, consists of algorithms that can iterate through large sets of data and then make smart conclusions and helpful correlations that would be difficult for a real person to detect. They can also learn over time to become better at their given task and improve their results with more exposure to data. These algorithms can digitally replicate the mind of a consumer, or the mind of an advertising researcher, or they can be utilized to communicate with consumers on behalf of brands. 

Gone are the days when armies of support staff wearing phone headsets are the only option companies have to communicate with inquiring customers. Now chatbots are the norm for many companies; Machine Learning allows chatbots to converse with human beings in a natural, human-like way that doesn’t frustrate users. The algorithms used by chatbots are able to pull bits and pieces from previous interactions and use them to infer answers to future questions, so they only get more effective at communicating over time. Chatbots are attractive to brands because they greatly reduce the overall costs they spend on customer service. Chatbots provide fast answers to the day-to-day queries of customers, resolving their queries immediately and simplifying the user decision process. Suppose a customer comes to your site and has trouble locating a certain product. In such a scenario, a chatbot on your site will solve the customer’s dilemma quickly. And unlike human agents, chatbots provide round the clock services.

Machine Learning can be used for a lot more than just chatbots though. Predictive targeting is a marketing technique that uses Machine Learning to predict customer decisions based on behavior patterns. Algorithms iterate through large data sets to predict the probability that a customer will take a certain action. They can help answer questions like: will a customer likely purchase this item or that? Will they be interested in engaging with a certain campaign, or would they react negatively to an advertisement? Machine Learning tools can help advertisers analyze the performance of ads, or help optimize marketing content, or they can analyze images posted on social media.

Social Media platforms can be an incredible source of data for advertisers. People tend to use Social Media to talk about their interests, comment on the places they have visited, and share their personal experiences with products and services. For example, a system that uses Machine Learning might conclude, after analyzing some social data, that young women who like a particular tv show and who are also interested in a certain celebrity are statistically more likely to purchase tickets for a certain vacation package. That could be very handy information for a brand. Suppose instead that you wanted to search through photos posted on Instagram to find images that contain your branded products: Machine Learning based tools could do this, and then analyze the content of those images to form a complex customer profile by observing characteristics about the people and environment in them. Using these customer profiles, a company could then launch new advertising campaigns that are more persuasive to this finely targeted audience.

There are however some potential downsides of using Machine Learning in advertising. For instance, Machine Learning only works with very large data sets, which will have to be harvested before any analyzing can take place. Machine Learning also struggles to reliably analyze sentiment because, despite advancements, algorithms simply aren’t capable of human intuition. Some tasks that are simple for humans are very difficult for a program to replicate. Sometimes AI is not a human enough replacement for an actual person. There are also companies who use machine learning in less-than-ethical ways. For instance, by sending out bots that can post deceptive content on social media platforms to falsely promote their brand, or falsely besmirch a competitor.

Despite these challenges though, there are many tasks that better suited for a Machine Learning based approach, and there is great potential in this emerging field. As the technology advances, there will be lots of new creative ways for advertisers to capitalize on the unique benefits of Machine Learning. Companies will need to adapt to these new methods if they are to stay competitive in the marketplace.  

Hyper-personalization in Advertising

Hyper-personalization utilizes artificial intelligence (AI) and data harvesting to deliver targeted advertising content that is closely tailored for each user. It can be a useful tool in an advertiser’s toolbox; however, it also raises several privacy concerns, and there is an inherent risk of alienating consumers.

Hyper-personalization is an evolution of personalized marketing, which consisted of such practices as adding a first name to an email opening line. However, personalized marketing never attempted to understand the individual preferences of each customer. Instead, it merely created the appearance that marketing messages were aimed at the individual. On the other hand, a hyper-personalized version of that same email might take a customer’s browsing behavior and location data into consideration, to offer products or services that would likely be of more interest to that particular customer. A hyper-personalized approach could also make use of real-time data, to make recommendations that were perhaps only relevant for a small window of time. For instance, a person who spent their morning googling car reviews, might be more receptive to an ad that promoted a local car dealership.  

Before venturing into hyper-personalization, relevant consumer data must first be gathered. This data can be harvested from a variety of sources, including social media, browsing tracking, purchase history, consumer trends, and data from IoT (internet of things) devices. Take Amazon for instance: their hyper-personalized recommendation engine is responsible for over 35% of their conversions. The engine’s algorithm is called ‘item-to-item collaborative filtering’ and it suggests products based on four data points:

  • Previous purchase history
  • Items in the shopping cart
  • Items rated and liked
  • Items liked and purchased by other similar customers

Using these data points, all harvested from their own site, Amazon can create a detailed user profile and then email this user ads that are specifically relevant to them. Thus, greatly increasing the likelihood of success.

Once consumer data has been collected and analyzed, companies can then segment their consumers into different subgroups. These subgroups can be based on a variety of factors, including demographics, location, brand interaction history, satisfaction, average amount spent, etc. Segmenting will allow for targeted communication, which will increase the potential for conversions. When customers are given offers that are unique to them, it’s likely that a solution will surface to their problem.

Returning to the idea of using hyper-personalization to promote local car dealerships, let’s take a look at an innovative way it could be used. Imagine a car dealership wanted to increase sales, they could target local consumers who had recently searched for car reviews, directions to a car dealership, or directions to a mechanic (if they are experiencing car problems, they may respond positively to messages to buy a new car). Once the car dealership had a group of local consumers that were likely interested in buying a car, they could then segment this group into subgroups based on their income levels, demographics, and browsing behavior. This would allow the dealership to send promotional content that included the specific vehicle (perhaps even in the specific color) that would mostly likely appeal to that consumer. 

There are challenges when it comes to hyper-personalization, however. Hyper-personalization requires a modern digital infrastructure. Most legacy systems fail to offer this capability, so many traditional mom-and-pop retailers (who may not have the resources to overhaul their existing platforms) have trouble adopting this new approach. There are also huge limitations that come from increasingly strict privacy regulations and consumer adoption of data protections. Consumer privacy laws, such as the California Consumer Privacy Act (CCPA) and the General Data Protection Regulation (GDPR), are putting cookie tracking at risk (the key component to tracking consumer behavior online). Free adblocking plugins and proxy servers can also be utilized by any consumer concerned about their data privacy, nullifying data collection and making hyper-personalization impossible. Users also often share devices, making it difficult to create a reliable user profile based on browsing behavior in those cases. Then there is the danger of eroding trust in brands if they appear to be too invasive. A poorly designed hyper-personalization campaign could push consumers away, if the communication makes it seem as though the brand has been spying on them. A recent survey showed that 86% of consumers are concerned about their data privacy, with 40% claiming that they don’t trust companies will use their data ethically.

Even with these challenges, hyper-personalization can be a profitable tool when utilized intelligently. Knowing the behavior and interests of your audience in real-time is very valuable, and new methods of capitalizing on this information are still being developed. Brands who are unwilling or unable to put hyper-personalization to use in their advertising strategies, will be at a disadvantage to brands who use it successfully to cleverly communicate with their audience.

RSS
Twitter
LinkedIn