Open any marketing trade publication right now and you will find breathless coverage of chatbots, predictive analytics, and algorithmic ad buying. But for a working marketer, it can be hard to separate hype from substance. What does artificial intelligence actually do for a marketing team once the buzzwords are stripped away, and where is it heading next?
A review published in the International Journal of Intelligent Networks attempts to answer that question by combing through the academic and trade literature on AI in marketing. The authors pull together research on how marketers are using AI today, which segments of the field it is changing most, and what obstacles still slow its adoption.
Who did the work, and what they were after
The review was led by Abid Haleem of Jamia Millia Islamia in New Delhi, along with colleagues Mohd Javaid (Jamia Millia Islamia), Mohd Asim Qadri (Galgotias College of Engineering and Technology), Ravi Pratap Singh (Dr. B.R. Ambedkar National Institute of Technology, Jalandhar), and Rajiv Suman (G.B. Pant University of Agriculture and Technology). The team, drawn from mechanical and industrial engineering departments, set out to summarize how AI is being woven into marketing practice rather than to test a single hypothesis.
The authors laid out four goals: to describe what AI is and why marketers are reaching for it, to catalog the places it shows up across different marketing segments, to map the changes it is bringing to the field, and to pull together a list of concrete applications.
For readers outside the field, a quick definition helps. Artificial intelligence is a broad label for software that performs tasks usually requiring human judgment, such as recognizing images, understanding language, or predicting what a customer will do next. Machine learning, a subfield, refers to systems that improve their performance as they process more data, without a programmer writing explicit rules for every situation.
How the review was assembled
The researchers describe their work as a literature-based study. They gathered articles, book chapters, and industry writing on AI in marketing from Scopus, Google Scholar, ResearchGate, and other sources. In total, they examined roughly 217 publications, reading through them and organizing the material into conceptual categories aligned with their four research questions.
This is not a meta-analysis that pools numerical results. It is a narrative synthesis, meaning the authors are describing and grouping what others have reported rather than running fresh statistical tests. The output is a map of the field rather than a measurement of effect sizes.
Where AI shows up in marketing work
The review organizes AI’s marketing footprint around familiar segments: pricing, product, promotion, place, strategy, and planning. Within each, the authors describe specific jobs AI is being asked to do.
On the customer-facing side, machine learning models sift through data from social media, websites, menus, and online reviews to help brands decide what content to serve and when. Programmatic advertising platforms use these models to bid on ad space in real time, matching messages to audiences algorithmically rather than through manual media buys. Facial recognition tools, the authors note, are being tested to track in-store visits and link shoppers to their social profiles, triggering personalized notifications.
On the back end, AI is being used for customer segmentation, churn prediction, dynamic pricing, and lead scoring. The review describes how predictive models can estimate the likelihood that a specific prospect will convert, or flag customers who appear to be disengaging before they leave. Chatbots handle routine service inquiries and, as a byproduct, gather data that feeds further personalization.
The authors also point to content-related uses. Natural language generation tools can draft email subject lines and short-form copy. Recommendation engines, like those powering Amazon or Netflix, surface products or media based on past behavior. AI-driven analytics identify which pieces of content perform best with which audience segments, and when to distribute them.
What is changing about the practice itself
Beyond specific tools, the review argues that AI is shifting the shape of marketing work. The authors describe a move from campaigns planned in advance and measured after the fact to a more continuous process in which models adjust messages, bids, and audiences in real time based on incoming data. Marketers, in this account, spend less time on manual segmentation and more time setting strategy and interpreting what the systems surface.
The authors list a set of applications where they see AI having the strongest effect. These include digital advertising optimization, reduction of routine errors, connection of end-to-end business processes, faster analysis of large data sets, automation of marketing workflows, intelligent customer support, predictive lead scoring, dynamic pricing, inventory control during demand spikes, and personalized email at scale. Each application is paired in their summary table with citations to the underlying studies.
One recurring theme: the review frames AI less as a replacement for marketers and more as a way to expand what a single team can handle. Tasks that once required a roomful of analysts, such as building customer personas from billions of interaction records, become feasible for smaller organizations once the tools mature.
The limits and the friction points
The review does not paint an unqualified picture. The authors devote substantial space to the obstacles marketers run into when trying to put AI to work.
Data quality is one. Machine learning systems are only as useful as the information fed to them, and the authors note that if records are inconsistent or riddled with errors, the resulting recommendations can misfire. They argue that marketing teams need to coordinate with data management functions to clean and maintain data before AI initiatives will produce reliable results.
Skills are another. The review points out that many marketing departments lack staff with data science backgrounds, which makes it hard to evaluate vendors, interpret model outputs, or troubleshoot when systems behave unexpectedly. The authors suggest that partnerships with outside firms are often necessary to get programs off the ground.
There are also concerns about transparency. Some AI approaches, particularly those using deep learning, produce decisions that are difficult to explain. The authors raise this as a practical issue for marketing teams that need to justify spending decisions or defend targeting choices to regulators and customers. They also flag privacy and data-use questions, noting that AI systems can push beyond what is legally permissible unless they are built with specific rules in mind.
Trust rounds out the list. The authors observe that AI is still a relatively new presence in many organizations, and that people are reluctant to rely on systems whose reasoning they do not understand. Poorly trained models can produce recommendations that are out of step with customer preferences, which erodes confidence further.
Practical takeaways for marketing teams
For marketers considering where to start, the review points toward a few patterns that recur across the literature. Applications with clear data inputs and measurable outcomes, such as email send-time optimization, lead scoring, and programmatic ad bidding, appear repeatedly as entry points. Personalization of on-site experiences and product recommendations is another area where the cited studies report gains.
The authors also suggest that investment in data infrastructure often has to come before investment in AI itself. Without clean, well-organized records, the downstream models struggle. And they note that transparency requirements should factor into tool selection: a system that produces an auditable report on why it made a given decision may be preferable to one that delivers slightly better performance but cannot explain itself.
As a caveat, readers should keep in mind that this is a descriptive review rather than an experimental test. The authors are synthesizing claims made across a large body of work, much of it industry-oriented, and the evidence base for any single application varies in rigor. The picture they assemble is a guide to where the field is pointing, not a verdict on which tools deliver the best return.
