MedSense’s AI journey from one-size-fits-all to personalised medical journalism

Medical Dialogues, an Indian digital publication serving doctors, nurses, and healthcare professionals with medical news and research updates, is building an AI-powered system that personalises content based on users' specialties and reading behavior

By: Meghna Singhania, Editor-in-Chief, Medical Dialogues and Vinoth Poovalingam, CEO, Blinkcms AI

When healthcare professionals visit Medical Dialogues, they come looking for information that matters to their practice. A cardiologist may be interested in the latest developments in cardiac devices, guideline updates, or interventional cardiology. A dermatologist may be looking for updates in biologics, acne management, or aesthetic dermatology. Yet, like many digital publishers, we have traditionally delivered content in a largely uniform way. 

At the same time, we observed another challenge. Publishers and healthcare organisations often rely on generic surveys and manual research methods to understand the medical audience sentiment and behaviour. These approaches can be expensive, time-consuming, and often fail to reach the most relevant audience segments. 

We believed that these two problems were connected.

What if the same intelligence used to personalise content could also help us understand audience interests, opinions, and emerging trends?

This idea became the foundation of MedSense AI, our AI-powered content personalisation and sentiment analysis system. We set out to realise it through the JournalismAI Innovation Challenge, supported by the Google News Initiative.

The problem we set out to solve

The modern healthcare information ecosystem generates an overwhelming amount of content every day. Doctors have limited time and increasingly expect information to be relevant, timely, and aligned with their professional interests.

Traditional content management systems are designed to publish content, but they are not necessarily designed to understand readers at a granular level. As a result, a cardiologist, dermatologist, and orthopaedic surgeon may all receive a very similar content experience despite having very different information needs.

We wanted to explore whether AI could help bridge this gap by identifying patterns in reading behaviour, specialty interests, and engagement signals to improve content relevance.

At the same time, we wanted to investigate whether AI could support a more intelligent approach to audience feedback and sentiment collection, allowing surveys to reach the right audience at the right time.

The team behind Medical Dialogues

What we have done so far

The first phase of the project focused on understanding both the market and the technical landscape.

We conducted competitor analysis, reviewed existing healthcare engagement tools, and interviewed stakeholders to better understand audience research requirements and survey workflows. These conversations helped validate the problem and refine our MVP scope.

At the same time, our technical team evaluated Medical Dialogues' existing infrastructure, user data model, and integration requirements. This work formed the basis for a new architecture that combines behavioural data, content metadata, specialty-wise information, and survey responses within a unified environment.

We also completed the design phase for the survey experience, including survey creation workflows, user-facing interfaces, audience targeting rules, and recommendation pathways.

More recently, development efforts have focused on building the foundations of the platform, including survey APIs, audience segmentation logic, question management systems, and early AI experimentation around interest inference, recommendation models, and sentiment analysis.

One of the most valuable lessons from the programme so far has been the importance of narrowing scope. Our initial vision included a wide range of features, but through mentorship and iteration we learned to focus on the smallest set of capabilities needed to validate the core concept.

Alongside this, we have been prototyping lightweight AI approaches for topic clustering, content relevance scoring, and sentiment analysis.

Lessons learnt

One of the biggest lessons from this journey has been the importance of starting with the problem rather than the technology.

When discussing AI, it is easy to become focused on models, algorithms, and infrastructure. However, our most valuable conversations have often been with editors, users, advertisers and stakeholders who helped us better understand the underlying challenges we are trying to solve.

We also learned that personalisation in journalism is not simply a technical problem. It requires careful consideration of editorial integrity, transparency, user trust, and privacy. Relevance should enhance discovery, not create information silos.

Another important lesson has been the value of building iteratively. As the project evolved, we refined our scope to focus on validating core assumptions through a practical MVP rather than attempting to build every possible feature from the outset.

The road ahead

The next phase of the project is where many of the most exciting elements begin to come together.

Over the coming months, we will focus on integrating the frontend experience, deploying recommendation services, and refining AI models that infer audience interests from reading behaviour.

We will continue developing our personalisation engine, automated audience targeting capabilities, and sentiment analysis workflows. The goal is to test these systems with real users, measure their effectiveness, and understand how they can improve both audience engagement and audience intelligence.

As development progresses, personalised content recommendations will be introduced for pilot users, allowing us to evaluate whether AI-driven relevance improves engagement and content discovery.

The final stages of the programme will focus on pilot deployment, user testing, performance evaluation, and documentation. These activities will help us understand not only whether the technology works, but whether it delivers meaningful value to both readers and publishers.

While the current phase focuses on a limited MVP, we believe the long-term potential extends beyond Medical Dialogues. Many specialised publishers face similar challenges in understanding audience interests and delivering relevant experiences at scale.

The JournalismAI Innovation Challenge, has provided us with the support, mentorship, and framework needed to explore these possibilities responsibly. As we continue building MedSense AI, we look forward to sharing more learnings about how AI can strengthen audience understanding while preserving the core values of trusted journalism.

_____

This article is part of a series providing updates from the second cohort of the JournalismAI Innovation Challenge, supported by the Google News Initiative. To read articles from our other grantees, click here.