VeriSci: Defining What Health Content Creators Really Need From AI

Team

Project Manager
Product Strategy
UX Designers

My Role

UX Researcher

Duration

3 Months

BACKGROUND AND PROBLEM

‍What started as an implementation project (Phase 1) quickly revealed a more fundamental need for discovery and definition (Phase 0). Our research uncovered that we needed to step back and:

‍Define the Business Need: Understand exactly how health content creators struggle with research efficiency
‍Clarify Scope: Determine which parts of the content creation workflow needed the most support
‍Align Stakeholders: Bring together content creators, scientific advisors, and platform developers
‍Assess Feasibility: Evaluate which AI capabilities could realistically streamline research without sacrificing accuracy
‍Develop Preliminary Plans: Create a roadmap for building and testing prototypes
‍Allocate Resources: Determine the necessary investment in AI development and user research

Based on responses gathered through our survey, we discovered a key insight from the current live prototype: health content creators spend nearly half their workday on research alone. Specifically, 25% of respondents reported spending around 2 hours per task, while 15% said they spend over 3 hours just gathering credible scientific information—before they even begin creating content.

What’s surprising is that, despite this heavy lift, they still manage to publish around 12 evidence-based posts each month across platforms like Facebook, YouTube, and Instagram.

This kind of output isn’t easy—and it highlights the need for tools that streamline the workflow without sacrificing scientific accuracy or audience engagement. That’s exactly where VeriSci comes in.

What is Verisci?

VeriSci is an AI-powered tool developed by Science to People to help online creators craft clear, shareable, and scientifically accurate health messages with ease. Designed to streamline the content creation process without compromising on credibility, VeriSci supports creators in producing high-quality content more efficiently. While the current platform doesn't generate images, our Phase 0 team conducted extensive user testing and laid the groundwork for future capabilities based on real-world needs and feedback.

Generating accurate health information based on scientific sources

Creating images and visual content that represent health data

Fact-checking user-created content

Reviewing content for accuracy and engagement

Based on user testing, these features address key pain points in the health content creation process, helping creators save an average of 42 minutes per post with projected usage of 9.25 times per month, potentially impacting 6.73 posts monthly.

What do the users want?

Through multiple research methodologies—including surveys, usability testing, and concept testing—several key user needs emerged. Testing involved both a Figma prototype and the live tool, each serving a unique purpose. The live tool provided valuable insights that helped inform the workflow for future development, while the Figma prototype allowed for rapid iteration and user feedback. The prototype that was ultimately handed off to us served as the first validated version of the experience, setting the foundation for the next stage of product development.

Content Creation Goals:
‍
Content creators prioritize creating content that connects with their target audience by addressing relevant, timely issues, such as current events, trends, and topics that align with the community's needs. They maintain content calendars and gather feedback from their audiences to inform future content.

AI Integration Expectations

Users approach AI tools with specific expectations:
‍
• Content creators seek AI for time-saving purposes, to generate ideas, and to provide a starting point for their social media posts, ensuring the content is factual and relevant to their audience

• They typically look for novel information, recent publications, credible authors, and content that is easy to understand

Concerns and Priorities

When introducing AI into their content creation process, creators expressed concerns about AI's accuracy, algorithm, bias, and cultural sensitivity. They emphasize verifying content through reliable sources and use AI tools cautiously.

Creators emphasized that tone plays a significant role in how their content resonates with their audience, with some preferring humorous approaches and others more serious or informative tones depending on their audience and content purpose.

Expanded Use Cases

The research found users utilizing the platform beyond its originally anticipated use cases. According to Sprint 3 findings, users were creating print materials (pamphlets), developing educational content for university lectures, and using the platform for business development support, demonstrating the tool's versatility beyond social media content.

FOCUS AREA

User Interface Approaches

The team tested two distinct design approaches. This was significant because the user testers had backgrounds primarily in research, which likely skewed results toward a research-focused UI. For future development, this tool needs to have a balanced blend of content creation and research capabilities to serve its intended purpose effectively.

Design A: Separated UI for research and content creation

Design B: Integrated UI combining research and content creation

Through usability testing in Sprint 4, the analysis showed no clear winner between the designs, with a slight edge given to Design A (separate UI) for content creation. However, Design B (Integrated UI) had an advantage for research tasks, suggesting different strengths for each approach.

Workflow Integration

A critical focus area was the integration of research and content creation processes. Users strongly prefer integrated workflow, particularly appreciating side-by-side research and writing capabilities: "I think that would feel really sleek to have different options of what type of social media attribution or tweet or LinkedIn post that I wanted"

Content Customization

Creators emphasized that tone plays a significant role in how their content resonates with their audience, with different preferences for delivery style (humorous, energetic, serious, informative).

Design improvements needed to address:

Providing more options for creators to customize the tone and delivery style of their content, such as preset styles (energetic, serious, etc.) or adjustable sliders for tone intensity
Simplifying the process of adding and customizing captions
Offering more tools for visual engagement

Fact-checking and Source Transparency

Fact-checking emerged as a major priority for health content creators, with participants emphasizing the importance of verifying the reliability and accuracy of information sources, particularly when making claims in their content.

Users expressed a strong preference for well-established, reputable sources and prioritize cross-checking AI-generated content against original research or data. They also noted the significance of providing clear citations and links to sources, which helps to validate the information and build trust with their audience.

Complementary Features

Mobile Experience: Valuable, But Secondary

While our research did show that one user asked for mobile optimization, most users completed their content creation on desktop using tools like Photoshop, Adobe After Effects, and other similar software.

An important insight that emerged during our research was the information-gathering process users go through—such as interviewing parents and doctors. To help creators save even more time, we explored the idea of a mobile companion app. This would allow users to input notes, voice memos, and other artifacts on their phones, and then, when they switch to desktop, the content would be ready and formatted for them to use.

Taking a step back, we realized that with the limited time available during this phase, a more strategic MVP approach would be to produce a browser extension first. This would help us validate the need for a mobile app or gather insights to improve the mobile app concept before full development.

Grammerly and SciSpace extension shown

Templates and Content Planning

Based on competitive analysis, there are opportunities to develop:

A library of customizable templates for common health communications
A content calendar and planning system
Collaborative planning features for team-based content creation
Milestone tracking and campaign planning capabilities

Image is based on Arclet

Audience Targeting

The research suggested adding:

Tone analysis tools for different target audiences
Audience segmentation capabilities

Platform Integration

A key complementary feature would be integrations with social media platforms for direct content publishing, streamlining the workflow from creation to distribution.

Image is based on Arclet

Image Generation and Customization

Users expressed interest in the ability to generate graphics, with two important caveats: assurance that each graphic is unique to its specific query, and the ability to edit the graphics to suit their style and needs. This is similar to how Genspark allows users to select an area for editing (see example below).

Image is based on Genspark

what's NExt

ReTROSPECTIVE

KEY TAKEAWAYS

Looking back at the beautiful chaos of this project—the late-night breakthroughs, cross-functional partnerships, last-minute in-person user testing at colleges, and rapid problem-solving—several clear themes surfaced that continue to shape how we think about building the tool for health content creators.

Balancing Act
‍‍
Successful health content creation requires balancing scientific accuracy with engaging presentation—VeriSci aims to make this juggling act more manageable through AI assistance.

Time is Precious
‍‍
Content creators spend significant time researching and creating health content across multiple platforms. VeriSci demonstrated potential time savings of 42 minutes per post, a substantial efficiency gain.

Trust Through Transparency
‍‍
Source citations and transparent fact-checking emerged as critical features that built user confidence in the platform and would help build audience trust in their content.

Tone Tailoring
‍‍
Health content creators need to maintain their unique voice and adapt tone based on audience and platform. Future development should enhance customization capabilities.

Beyond Social
‍‍
The discovery of expanded use cases (educational materials, print content, business development) suggests VeriSci has potential beyond just social media content creation.

Interface Integration
‍‍
During testing, users had mixed feelings about using separate or combined interfaces. But one thing was clear—they wanted a more seamless experience, with fewer disruptions and an easier way to connect their research directly to the content they were creating.

Additional Focus Areas I'm Exploring

Building LLM structure. Below you'll see my thinking process based on what I have learned.

The “purple stickies” in the Steps section could represent smaller LLMs that serve as fact-checkers for the primary model.
Localized content stands out in this phase. I’ve learned that health content creators often use the Nextdoor app to identify health issues specific to their communities—which I found to be a really unique insight. For example, consider the recent measles outbreak in West Texas: creators can leverage this localized information to tailor their content more effectively.
The LLM will prompt users with content scraped from trusted sources. My assumption is that users will be more likely to trust the AI-generated content if they can see and select from the sources the LLM references.

Building a Brand Guide. Will have this up on my homepage in the near future.
Vibe Coding! Instead of building traditional prototypes in Figma, I thought it would be fun to explore vibe coding using a small LLM—just to play around and see what emerges. I will have this up on my homepage in the near future.