Definitions

The interest in providing definitions that allow delimiting the field of study is typical of an emerging field. The definitions provided, even before the object itself (advertising transformed by the impact of AI), concern the concept of artificial intelligence. Rodgers (2021), for example, compiles in a systematic table up to 27 definitions of the term from disciplines as diverse as computing, psychology, mathematics, engineering, advertising, philosophy, medicine, health, commerce, law, education, and engineering. All of them share the features of automatism, effective problem solving, rational guidance through the application, or development of algorithms with the ability to adapt to the evolution of their task performance, and the intensive use of data (Qin and Jiang, 2019; Kumar et al. 2019; Li, 2019; Albinali and Hamdan, 2020).

In the analyzed context, AI implies, therefore, algorithm systems (or applied software) that allow the development of specific tasks and that have a certain learning capacity from training with data in the practice of tasks, so that they can correct and adapt their performance and improve their results (Mittelstadt et al., 2016). Learning can be supervised or autonomous and is developed through training with data sets (machine learning) or through recognition capacity and decisional autonomy (neural networks and deep learning) (Watts and Adriano, 2021). Traditionally in AI, three types of algorithms are considered according to their functionality (Hill, 2016): Predictive analysis (those that allow obtaining information about what will happen in a process and/or why it will happen that way), prescriptive analysis (those that allow obtaining information about what to do and why in a field of causally related possibilities); and automated decision or action (those that perform autonomous actions that require decision-making without human intervention).

On the other hand, the transformative impact of AI as a vector of innovation is the subject of intense theoretical and analytical production (Allan, 2016; Lee et al., 2019; Paschen et al.), also in the field of communication and the media (Song, 2019). This nature as a factor of innovation is associated, according to Paschen et al. 2020, with the ability to alter the ways and times of production processes and with the ability to generate value from the application of analytical or decision-making techniques and tools.

Hancock et al. (2020, p.1) identify communication mediated by artificial intelligence as that form of communication in which “an intelligent agent operates on behalf of a communicator by modifying, improving, or generating messages to meet certain communication objectives”. In this way, advertising mediated by artificial intelligence is seen as one in which technology intervenes to improve the achievement of communication objectives.

In this sense, three complementary concepts are observed in the analyzed literature regarding the impact of artificial intelligence on the advertising process: synthetic advertising, computational advertising, and smart advertising.

Synthetic advertising assumes an approach more focused on the product of advertising (the advertising message), placing the defining emphasis on “ads created or modified by the alteration with artificial intelligence” (Campbell et al., 2021, p. 1). It encompasses, therefore, the use ofAI toimproveimageediting, creation, and post-production processes (as in the case of so-called deepfakes) as well as the incorporation of technologically advanced image formats, such as virtual reality (VR) or augmented reality (AR).

Computational advertising, on the other hand, delimits an approach focused on the medium or tool. The concept designates a broad spectrum of phenomena around the confluence of systems, technologies, and computational methods in advertising:

Refers to an advertising ecosystem with fruitful computing capabilities that employ mathematical, physical, IT, and economic approaches to expressively represent a rich set of advertising objects and environments, model and analyze complex stakeholder behaviors, facilitate efficient, profitable delivery of advertising information to potential consumers across various media vehicles in different contexts of customer/product/brand/firm interactions and touchpoints, and provide efficient, optimal solutions for advertising decision making and market design. (Yang et al. 2017, p. 3).

The concept of computational advertising connects the three basic dimensions of advertising innovation: formats, data, and algorithms, and encompasses all the actors and interactions of its defining processes (Duan and Yang, 2018; Helberger et al., 2020).

The third approach takes a similar line, although more focused on technology and its ability to be customized. Qin and Jiang (2019) identify smart advertising with the application of AI to the design, production, management, and analysis of advertising content. Synthetically, Li (2019:334) defines intelligent advertising as a brand or product/service communication focused on the consumer, based on data exploitation, and mediated by intelligent algorithms. For some researchers, intelligent advertising constitutes a true subdiscipline of advertising that is located “at the intersection between cognitive sciences, computer sciences, and advertising” (Rodgers, 2021: 5). In any case, intelligent advertising constitutes a new phase of digital advertising, after the phase of interactive advertising and that of programmatic advertising (understood as automated buying and selling), marked respectively by the innovation of the Internet and Big Data (Li, 2019). Those texts in the sample that consider the impact of AI in the production and uses of the advertising image with immersive formats such as VR or AR maintain, however, the notion of smart advertising linked to the ability to personalize (Campbell et al., 2021).

It is worth underlining the close connection between data-driven advertising and smart advertising (Yang et al., 2017). The relevance of data processing in nuclear phases of the advertising process, such as the identification of insights, segmentation or research, and control of results, supports a natural connection between the contribution of Big Data (Skiera, 2016; Malthouse and Li, 2017) and the subsequent demand for increasing computing and automation capabilities that pave the way for AI (Chen et al. 2019). But this does not mean that other more qualitative processes of advertising activity, such as creativity, writing, or interaction with the user, are exempt from the disruptive influence of AI.

Innovation vectors

The synthesis or classification of innovation vectors derived from AI in advertising is another point of shared interest in the analyzed literature that explores the limits and scope of new advertising (Helberger et al., 2020). In this sense, innovation vectors are understood as those instrumental features of AI that have a transformative effect on advertising processes (Paschen et al., 2020) and would come to answer the question: what features or capabilities of AI produce innovation in the advertising ecosystem?

Precisely, the efficiency in the processing of large data sets and the identification of relevant patterns constitute one of the instrumental aspects of the contribution of AI that characterize its transformative condition (Kumar and Shaphali, 2016). The analytical power of artificial intelligence allows Big Data to be incorporated into prospective advertising processes, which require identification of objects (insights, profiles...) and decision-making (production, media purchase), particularly regarding unstructured data, which -unlike the structured data with which conventional marketing and advertising work- constitute the true source of value in the digital context (Kietzmann et al., 2018).

The impact on decision-making also includes the areas of creativity or strategic planning, transforming traditionally hierarchical processes, such as the decision on actors or prescription figures (celebrities), production logistics (outdoors), etc. (Lee and Cho, 2019). Both predictive and prescriptive analysis support a radical change in recommendation and segmentation capabilities (Enache, 2020). A change that contributes to transforming the processes of advertising in the interactions between key actors (users, agencies, and the media) but that also facilitates the emergence of new actors in the advertising ecosystem (consultancies, technological intermediaries) (Lee and Cho, 2019).

The ability to automate tasks, and optimize times and processes, is clearly identified in the literature as the second disruptive contribution of AI to advertising (Li, 2019). This contribution is particularly visible in the media planning and sales processes, which, with the development of programmatic purchases, allow managing the coupling of large inventories of spaces and networks of advertisers in real-time, making it possible to adapt the message to the user profile and situation (Busch, 2019). AI- induced acceleration sometimes makes it possible to merge processes that were previously consecutive and belonged to different phases (for example, between the creative and media buying phases, or between the printing and impact evaluation phases). The consequence of the acceleration induced by the automation of tasks is, therefore, that of a reorganization of the processes based on certain simultaneity loops that allow the feedback of decisions to be incorporated in real-time (Qin and Jiang, 2019; Enache, 2020; Ahn, 2020; Rodgers, 2021).

The analyzed texts also point to two areas of development characteristic of AI as sources of growing influence in the field of brand and product communication: complex object recognition capabilities, especially in the case of Natural Language Processing (NLP) and Image Recognition (IR) systems.

Intelligent systems for natural language processing (NLP) are essential for the development of automation of writing tasks (Qin and Jiang, 2019). They also constitute the basis for the development of language recognition apps that make it possible, for example, to effectively explore conversations on social networks to identify insights (Kietzmann et al., 2018) or transform interaction processes with users through conversational algorithms and chatbots, of particular interest in brand communication actions (Li, 2019). Natural language processing capabilities have also proven to be a useful resource in the automated management of the semantic interrelationship between ads and their context, an issue that has traditionally concerned advertising due to the consequences that an inappropriate context could have on the interpretive effectiveness of ads (Watts and Adriano, 2020).

Image recognition (IR), for its part, currently has relevant implications in the creative tasks of advertising: ontheonehandintheoptimizationofprocesses(imageinventorymanagement,editingandpost-production tools, etc.) and on the other, on the optimization of the effectiveness of creativity, making possible forms of automation and personalization of creative processes that allow adapting the image of the ad to the strategy of the advertiser/brand or to the user’s profile (from complementary aspects, such as color, to substantial issues such as the brand or the product itself) (Lee and Cho, 2019; Vrakatsas and Wang, 2020).

The transformation of the advertising process

The third focus of attention of the analyzed articles emphasizes how AI transforms the processes and phases of advertising and would include those contributions of the analyzed sample that answer the question: what aspects of advertising does the impact of AI transform and how does it transform them?

Qin and Jiang (2019) state that smart advertising is instrumental in restructuring advertising processes, including advertising research, strategic planning, creativity, copywriting, media buying, and control. Similarly, Li (2019) identifies insight discovery, ad design and creation, media buying, and campaign effectiveness evaluation as areas of advertising substantially transformed by artificial intelligence.

AI allows advertising research (discovery of insights and consumer trends) and market analysis to be merged in a single step (Qin and Jiang, 2019). The use of language recognition tools to explore activity on social networks in real-time has considerably improved the handling of structured data characteristic of conventional marketing and has led to an unprecedented qualitative leap in the treatment of unstructured data (Skiera, 2016).

Speech and Image Recognition Systems have made it possible to integrate the flow of unstructured data that circulates in the photos, videos, and audio that we share on social networks to the wealth of structured data built from the textual exchanges in which users participate. And, at the same time, the application of AI to the processing of this data makes it possible to identify relevant patterns and generate coherent profiles, which constitute the raw material for the use of predictive analytics on consumer behavior (Shumanov, Cooper, and Ewin, 2021). Kietzman et al. (2018) illustrate, for example, how the digital marketing company Astro uses Microsoft’s Azure system to process billions of data points in a profiling system that allows it to dynamically adapt its clients’ websites in real-time.

Although it is an environment little associated with the impact of AI, its innovative influence is also becoming more noticeable in advertising production and creativity. Several authors (Qin and Jiang, 2019; Vakratsas and Wang, 2020) point to smart copywriting systems as a specific case, which allows the writing phases to be automated and generate an adaptable inventory of slogans, claims, or significant elements of the advertising message.

Through the use of prescriptive algorithms, AI proposes a horizon of optimization of the adaptation of formats and advertising channels to the product, brand, or strategy (Lee and Cho, 2019). This makes it possible to improve decision-making on what type of advertising (outdoor digital, conventional display, desktop, smartTV, mobile, advergame, social networks, voice assistants...) is more appropriate for a certain brand, a specific product or service, or a specific campaign.

AI also makes it possible to enhance the adaptation of the advertising message to the recipient’s profile in real-time, not only in terms of message/user matching (what programmatic buying deals with) but also in terms of dynamic adaptation of the type and content of the message through Dynamic Creative Optimization (DCO) systems (Koren et al. 2020).

DCO systems establish a series of basic combinations between digital advertising message options (colors, aspect, image, text) structured in templates and profiles available through Data Management Platforms (DMP), which take into account basic variables such as time, weather, location, device, ad placement, demographics, and retargeting data.

In this way, a certain flexibility is achieved in adapting the message to the user and its context. This makes it possible, for example, to optimize product recommendation and inventory management systems by adapting the image of each product in the catalog to the user’s profile, as is the case with the covers of the Netflix series and movie catalog, which change their appearance depending on the user’s previous consumption (Song, 2019).

Perhaps the best-known aspect of the impact of AI in advertising is that of programmatic buying, which is leading to a profound transformation of the traditional media buying and planning system (Busch, 2019).

Programmatic buying designates the automation of the process of allocating advertising space in the media as a result of the activity of Ad Exchange networks in which two intelligent systems are involved: Data Management Platforms (DMP) and Demand Side Platforms (DSP). The agencies use the gigantic data collection of the DMPs to evaluate the strategic and opportunity potential of each advertising impression and bid for them through the intelligent algorithms that make up the DSPs (Carrillo-Durán and Rodríguez-Silgado, 2018).

The dominant form of programmatic buying is Real-Time Bidding (RTB), in which advertising networks bid in open auctions for advertising space in real-time, in such a way that the resulting impression in the media is adapted to the specific user profile (Martinez et al., 2017).

Programmatic Direct Buying (PDB) is the other automation format used in AI-powered ad exchange networks. In this case, the purchase is about a select inventory of impressions and media in Private Marketplaces (PMP) that imply a higher quality of data and impressions (Qin and Jiang, 2019). Figure 1 illustrates the ad path in a mobile environment in the context of a real-time auction (RTB).

Source: Own elaboration from Busch, 2019.

At this point it is worth highlighting the role of AI in the development of programmatic buying on television and video, where the complexity of handling unstructured data derived from the use of images and audio and its reconciliation with the logic of real-time processes poses some technical challenges of enormous scope (Malthouse et al., 2018).

Based on the potential for automation and exploitation of data repositories that AI offers, Chen et al. (2019) identify an extension of the programmatic buying model to production and creativity processes that extends the capabilities of Dynamic Creative Optimization (DCO) systems. The so-called programmatic creativity, thus, comes to form a new field of optimization coordinated with Content Management Platforms (CMP) that integrates the DCOs with Programmatic Advertisement Creation (PCA) systems. The result is a new programmatic advertising ecosystem that unifies the automation of creativity and media buying and selling (Chen et al. (2019) according to the scheme in Figure 2.

The optimization of processes derived from AI maximizes, in this case, scalability and adaptation in real-time: The ability to manage large repositories of combinable elements for the creation of advertising content allows managing large advertising inventories, as occurs in the case of the operator Verizon, which uses DCO and content management platforms to optimize its native advertising service (Koren et al., 2020). At the same time, the ability to adapt in real-time the appearance and content of the advertising message to the profile and context of the audience allows for improving response rates and the effectiveness of advertising communication (Vrasatkas and Wang, 2020), interconnecting in real-time creativity with impact analysis in a loop that is one of the touchstones of advertising effectiveness (Qin and Jiang, 2019).

Source: Chen et al., 2019.

In the sphere of the creation of complex images, the fusion between creation, impression and impact evaluation also has implications in the field of product placement. First, the increase in real-time image creation and video editing capabilities using AI expands the horizon of adaptation possibilities. The Microsoft-linked firm BEN (Branded Entertainment Network) uses AI to identify the best scenes and contexts of an audiovisual piece for product placements based on the brand or campaign. And the technology company Mirriad has developed in-content advertising platforms capable of identifying surfaces of objects in the image that may contain advertising or brands (Bradley, 2021). The conjunction between image processing technology, on the one hand, and the identification of profiles in real-time, on the other, allows, for example, to artificially “insert” an ad in a street marquee that a character from a movie crosses or change the beer that the protagonist of a series drinks in an episode to place the product or message that best suits the viewer’s profile (Campbell et al., 2021).

AI also makes it possible to apply facial image reconstruction technology (deepfake) to the creation of ads, resurrecting deceased actors or celebrities to turn them into current advertising prescribers, as happened in 2020 in Spain with the Cruzcampo campaign that used a deepfake of the singer Lola Flores (Palomo-Domínguez, 2021). But the key to its potential, beyond the spectacular nature or the effectiveness of the images created with AI, is that these systems can learn to improve their performance by integrating the analysis of their activity and the feedback derived from the user’s response. (Li, 2019).

The campaign evaluation process in terms of impact and effectiveness is also affected by the disruptive impact of AI. The automation of the processes and their execution in real-time allows you to obtain feedback on the campaigns at the very moment of their implementation, offering the possibility of making strategic decisions and optimizing the campaign on the go. Intelligent performance data processing and analysis systems “reduce the time between impact assessment and campaign optimization to a matter of seconds” (Qin and Jiang, 2019, p. 6). This is, perhaps, the most relevant disruptive element that AI brings to the advertising process: reinserting or merging the evaluation of effectiveness or impact as one more element of each phase. For the sample addressed, the fusion between execution and measurement constitutes the true revolution of AI in its contribution to the organizational dynamics of advertising (Lee et al., 2019). In this sense, the internal perception of the actors coincides with the processes that the research works reflect, pointing in both cases to AI and Big Data as the two engines that determine the recent evolution of the sector (Martínez and Aguado, 2019).

Beyond the traditional advertising process, the consulted literature also includes a potential niche for the application of AI tools and processes in the field of brand communication: the use of chatbots or conversational algorithms in social networks opens up the possibility, according to Thomas and Fowler (2020), for the development of artificial prescribers or AI influencers. Artificial conversations are already an important part of social networks: In a study on the presence of bots on Twitter, Liu (2019) has identified that at least 30% of the content on the social network is produced by algorithms. At the same time, companies are beginning to experiment with virtual influencers who embody brand values associated with the identity of their followers. According to Rodrigo-Martín et al. (2021), virtual influencers are “avatars created using artificial intelligence and augmented reality techniques that allow the communication and advertising possibilities of the traditional influencer to be combined with absolute control of their activity and “life” on social networks”. Rodrigo-Martin et al. (2021) point to the case of Lil Miquela, a 19-year-old fashion influencer, musician, and model who defines herself as queer and a Black Lives Matter activist, who has more than three million followers on Instagram, and who, in reality, is a virtual avatar created and managed by Brud, a Los Angeles technology company. The potential of AI in this field, as Thomas and Fowler (2020) point out, goes beyond the creation and characterization of the avatar, to include the adaptation of the behavior and the message of the algorithmic prescribers to the profiles and personality of their followers, thus maximizing the effectiveness of the prescription.