work /

case study /

prevagen

Healthcare E-Commerce

From information site to AI-cited e-commerce platform.

Five years rebuilding Prevagen.com around the people who actually use it — and the machines that now answer questions about it.

Shopify

UX Design

WCAG Accessibility

E-Commerce

AI Visibility

Schema Markup

Front-End Dev

AI Chatbot

Client

Quincy Bioscience · Prevagen

Role

Senior Front-End Developer & Designer

Timeline

2021 – Present

Scope

Full platform rebuild + ongoing AI visibility

Platform

Shopify (converted from WordPress)

Environment

Regulated YMYL healthcare

01 · Background

A regulated brand with a trust problem.

Prevagen is a cognitive health supplement brand operating in one of the most scrutinized categories in consumer healthcare. The customer is a well-educated adult over 55 — typically a retired or still-working professional — who is skeptical by default, doesn't want to be sold to, and will leave a site immediately if anything feels unclear, cluttered, or evasive.

When I joined Quincy Bioscience in 2021, prevagen.com was a WordPress site with no clear commercial direction. It functioned as a product information page rather than a purchasing destination. Sales weren't being tracked. Accessibility hadn't been considered. The design reflected the assumptions of internal stakeholders, not the behavior of the people it was built for.

The work ahead wasn't a refresh. It was a rebuild — of the platform, the architecture, the customer journeys, and eventually the way the brand shows up in AI-generated answers.

02 · The Challenge

Three problems that were actually one.

On the surface there were three separate challenges: the site needed to become a functioning e-commerce platform, it needed to meet accessibility standards, and it needed to work for a demographic that most digital teams design past.

The Prevagen customer demands the same things WCAG demands: clear hierarchy, unambiguous navigation, sufficient contrast, no cognitive overload, one subject per screen. Designing for accessibility and designing for this demographic aren't separate disciplines. They're the same discipline applied to the same person.

The Lighthouse accessibility score when I started: 35. No ARIA implementation. Contrast failures throughout. No consideration of keyboard navigation, screen reader compatibility, or semantic structure. In a YMYL health category — where trust signals carry legal and commercial weight — this wasn't just a compliance problem. It was leaving money on the table.

03 · Discovery

What the research actually showed.

Before touching the platform I ran in-person research with real members of the target demographic. Not surveys — actual conversations with people from the audience, watching how they used the existing site and listening to what they said about sites like it.

Three findings directly contradicted what internal stakeholders believed.

Reviews don't build trust — they destroy it.

Stakeholders wanted customer reviews front and center on the homepage, drawing on Amazon behavior patterns and the assumption that social proof converts. The research showed the opposite. Several participants said directly that they don't trust reviews on company websites and don't believe the people in brand commercials are real users. Putting reviews in the main sections of the site didn't signal credibility — it signaled a brand trying too hard to sell them. We kept reviews off the primary page sections. It was a fight worth having.

The hero Buy button was causing people to leave.

The existing site had a purchase button in the hero image. Stakeholders assumed prominence meant conversion. Most visitors didn't see it at all — and among those who did, several reacted negatively. The phrase that came up repeatedly: feeling forced into a decision before they'd had a chance to learn anything. We removed the Buy button from the hero entirely and placed it in the navigation and as a dedicated CTA further down the page, after the educational journey had a chance to do its work.

The ingredient information was effectively hidden.

Visitors came to the site with a specific question: what is in this product and does the science support it? They couldn't find the answer quickly and several abandoned before locating it. For a health supplement brand in a YMYL environment, ingredient transparency isn't optional — it's a trust prerequisite. I gave ingredients a front-and-center presence: a dedicated panel in the product images and prominent placement in the FAQs. This decision served the human user and, as it turned out, the AI visibility work that came later.

Each of these decisions was driven by behavioral evidence, not aesthetic preference. The two journeys that became the site's organizing principle — education and purchase — emerged directly from watching what this demographic needed before they were willing to buy.

04 · The Decision

Winning the argument for the user.

The most significant decision in this project wasn't technical. It was organizational.

Internal stakeholders had strong opinions about the site's direction — opinions shaped by what they wanted to communicate, not by how the actual customer made decisions. The research made clear these two things weren't aligned. The customer wanted fewer choices, clearer paths, and less noise. Stakeholder preferences tended toward more information, more options, more visual complexity.

We can design for what we want to say, or we can design for what the customer needs to find. The first approach had produced a site that wasn't converting.

I brought the research to those conversations as evidence — behavioral data, direct quotes, observed navigation patterns from real users in the target demographic. Winning that argument gave me the latitude to make the structural decisions the site needed. The two-journey architecture that followed — education and purchase, nothing else — was the direct result of that conversation.

05 · The Work

Platform, product, accessibility — all of it from scratch.

The WordPress site was replaced with a Shopify build — designed and developed by me. Not a theme implementation. A ground-up build around the customer journey research and WCAG requirements, treated as a single brief.

Platform Migration

Full WordPress to Shopify conversion. Custom storefront built on the two-journey architecture — education and purchase as the only organizing logic.

Accessibility Remediation

Complete ARIA implementation across all components. Contrast system rebuilt to WCAG AA. Semantic HTML established from the root level. Lighthouse score: 35 → 100.

Subscribe & Save Program

Built on RECHARGE, customized beyond the platform defaults to match how the Prevagen customer shops. Result: 80% renewal rate. A large share of new visitors convert to subscription customers.

Rewards Program

Designed and implemented to extend customer lifetime value and give the subscription base a reason to stay engaged between purchase cycles.

05b · Conversational AI

Building a compliant AI chatbot in a YMYL environment.

Designing a consumer-facing AI chatbot for a cognitive health brand in a regulated environment is a different problem from building a standard e-commerce bot. The visible functionality — answering questions about products, pricing, ingredients, returns, rewards, subscriptions, and order support — is the easy part. The hard part is everything the bot must never do.

Prevagen operates in a YMYL category. A chatbot that insinuates a connection between the product and medical outcomes, or that answers a health question in a way that implies clinical authority, creates compliance exposure. The guardrail design wasn't a feature. It was the core of the brief.

200+ documented scenarios — sources & coverage

Product questions and pricing

Ingredients and formulation

Returns and order support

Rewards and subscription how-tos

Where to buy and availability

CSR escalation triggers

Compliance guardrails

The bot does not answer medical questions, does not connect Prevagen to health conditions or cognitive outcomes, and does not engage with questions outside the Prevagen scope. Business-opportunity and off-topic questions route to a live customer service representative. The escalation logic was a deliberate design decision — keeping the bot within the domain where it can be fully trusted rather than stretching it into territory where a wrong answer has consequences.

Scenario sources included customer service records, social media listening, and direct customer research. Every scenario was tested against two questions: does this response serve the customer, and does it stay within the regulatory boundary? When those two requirements were in tension, the compliance requirement won — and the response design found a way to be helpful within that constraint.

06 · AI Visibility

The brand now owns its narrative in AI.

For a health brand in a YMYL category, what AI systems say about you is no longer a secondary concern. Third-party review sites, misinformation, and low-quality content had been filling the space that prevagen.com should have owned. Schema work, content restructuring, and a deliberate two-site entity strategy changed that.

The governance structure. Quincy Bioscience's site was repositioned as a technical authority layer — a more clinical, reference-grade property that functions as a structural support beam for prevagen.com. AI systems reading both properties see a coherent entity: a parent organization providing technical credibility and a consumer property providing product and purchase information. That relationship tells AI systems there's organizational depth behind the brand in a way a standalone consumer site cannot.

Schema and structural implementation. Product schema, ingredient schema, FAQ schema, and BreadcrumbList markup were implemented across key pages. The ingredient pages were a specific focus. Apoaequorin — Prevagen's key ingredient — had been a gap in the brand's AI presence. Restructuring the ingredient content and implementing appropriate schema meant that when users ask AI systems about the ingredient, prevagen.com now appears as the cited source rather than third-party commentary.

Content restructuring for entity clarity. The FAQ restructuring that served the human user — putting ingredient information front and center — created the kind of structured, attributable content that AI systems extract and cite. The accessibility work and the AI visibility work were drawing from the same structural substrate.

ChatGPT

Cited as the place to buy online. Product and purchase option links surface in conversation when visitors ask about Prevagen.

Google AI Overviews

Shows as a cited answer. Products, site, and promotions appear in Google Merchant results for cognitive health queries.

Secondary Search

Owns the secondary terms. 'Prevagen reviews,' 'prevagen ingredients,' 'is prevagen safe' — prevagen.com is now the authoritative source.

The work is ongoing. Entity representation in AI knowledge graphs isn't a one-time implementation — it's a maintenance discipline. The structural foundation is in place and the early citation results are measurable, but the broader goal — establishing prevagen.com as the default authoritative source across AI systems for questions about the brand, its ingredients, and its category — is still being built. That's the current work.

07 · Results

What the numbers say.

The sales increase built over two consecutive years as the structural decisions compounded. A site that had never tracked sales as a primary metric became a functioning e-commerce platform with measurable conversion, retention, and recovery rates.

35→100

Accessibility score · Lighthouse

30%

Sales increase · two consecutive years

80%

Subscribe & Save · renewal rate

45%

Cart recovery · rate

3%+

Conversion rate · site-wide

Engagement improvement · time on page + click-throughs

The accessibility score moving from 35 to 100 is the number that matters most to me. In a regulated health category, with an audience that makes trust decisions in seconds, that score represents structural rigor — not checkbox compliance.

Why this matters for your team

What this engagement demonstrates is a decade-long ability to make structural decisions under real constraints — regulatory, organizational, and technical — and trace those decisions to business outcomes. The research-to-decision chain, the compliance work, the AI visibility architecture, the chatbot guardrails: these aren't separate skill sets. They're the same systems-level thinking applied to different surfaces of the same problem. If you're building in a regulated environment, taking AI visibility seriously, or need someone who has actually shipped a compliant AI product — this is the kind of work I bring to a team.

08 · Reflection

One thing I'd do differently.

I would have started the AI visibility work earlier. The structural decisions made for accessibility — semantic HTML, clear entity signals, well-formed schema — turned out to be exactly the right foundation for AI readability. The connection between those two disciplines became clear over time, but I was making the accessibility decisions for compliance and UX reasons before I fully understood their implications for AI interpretation.

If I were starting this project today, the accessibility audit and the AI visibility audit would be the same audit. They're measuring different surfaces of the same structural substrate. That's the lesson this project taught me — and it's now the lens I bring to every new engagement.