The most challenging part of building an AI chatbot often isn’t the technology itself—it’s the knowledge base. This foundation of specific, accurate information is what a chatbot relies on to answer visitor questions effectively. Most teams treat this as a brand-new project: identifying common questions, drafting approved answers, and then configuring the chatbot. However, for this healthcare provider, that groundwork was already laid. It wasn’t intentional at first; years of high-quality SEO content had simply done the heavy lifting before a chatbot was even on the horizon.

The Client

The client is a healthcare provider specializing in hyperbaric medicine, a form of pressurized oxygen therapy used in wound healing, decompression sickness, and a range of other clinical applications. Patients researching this specialty typically arrive with highly specific questions before making contact. Educational content is particularly valuable here, driving search traffic and helping prospective patients work through their questions before reaching out. The client has worked with Hypha on a long-term retainer, with content as the foundation of the engagement from the start.

The Challenge

The original scope was SEO-focused. Rank for hyperbaric-related searches, drive patient inquiries, build organic visibility for a specialty most people haven’t encountered until they need it. Hypha produced high-funnel educational blog content, posts explaining treatment conditions, clinical applications, and what patients should expect, and that content performed.

The chatbot question came separately, when HubSpot’s AI chatbot tools became viable enough to implement. The question wasn’t whether to add a chatbot but where the answers would come from. For a healthcare provider, chatbot responses can’t be pulled from a generic language model, but rather, they need to draw from accurate, specific information about the provider’s own services and specialties. Building that knowledge base from scratch, for a niche clinical specialty, was going to take significant time.

Layer 1: Content Built for Search

High-funnel content about hyperbaric medicine carries a specific requirement. Patients reading about clinical treatments aren’t casually browsing. They’re evaluating whether a therapy applies to their condition and whether the provider seems credible enough to contact. The blog posts Hypha produced were written to rank, but accuracy was the underlying standard. For a specialty most patients are encountering for the first time, content that doesn’t address their actual questions won’t hold attention regardless of how well it ranks.

Layer 2: Repurposing the Archive

When HubSpot’s AI chatbot capability became ready to implement, Hypha restructured the existing SEO content into HubSpot knowledge base articles. Content written and formatted for search was reorganized into a format the AI chatbot could reference when responding to visitor questions.

A chatbot drawing from years of specifically written content about hyperbaric oxygen therapy can answer clinical questions in a way a generic AI response can’t. The accuracy that made the content useful in search is the same quality that makes it useful in a live conversation. Once the knowledge base was in place, the content library was effectively serving two functions: supporting organic search visibility and powering AI-driven conversations with visitors.

Layer 3: Chatflow Meets HubSpot AI

The live system combines two components: HubSpot’s AI chatbot and traditional HubSpot chatflow routing. They serve different purposes and the implementation keeps them clearly separated.

The AI chatbot handles open-ended informational questions. A visitor asking what hyperbaric oxygen therapy treats, or whether it applies to a specific condition, gets a response drawn from the knowledge base content. Traditional chatflow handles structured requests: appointment inquiries, direct contact, handoffs to the care team. Those conversations follow rule-based paths that produce consistent, predictable outcomes.

The Hypha team member who led the project described the combination this way: “It kind of blends, like, the old school setup with the new school capabilities and lends into our content, which has always done pretty well for them.” That framing is accurate to how the system actually works. The AI handles the information layer. The chatflow handles the action layer. Neither component is doing the other’s job, and the system is better for that clarity. The chatflow went live in May 2026.

The Outcome

Visitors can now get substantive answers to clinical questions outside business hours, through content the provider has been producing and refining for years. When a visitor’s needs require a human, the chatflow routes them to the right team. The two systems work together without requiring a visitor to notice the transition.

Going forward, any new content produced for search also expands what the chatbot can answer. The content calendar and the knowledge base are no longer separate concerns.

What This Pattern Means Past This Implementation

The problem most organizations face when adding an AI chatbot isn’t the configuration. It’s the knowledge base. Building it from scratch is slow, the scope is hard to define, and the project often stalls waiting for a content initiative that could have been avoided.

This client didn’t start from scratch. The knowledge base was assembled from years of educational content already in place, already performing in search, and already accurate enough to train a chatbot on a clinical subject. For organizations with a substantial content history, auditing what exists is worth doing before treating the knowledge base as a net-new initiative.

Healthcare made this approach particularly well-suited. Content written to accurately address clinical questions serves the same function whether it’s indexed by Google or referenced by a chatbot. The delivery mechanism changed. The underlying requirement—accuracy—didn’t.

If you’re evaluating HubSpot’s AI chatbot tools, don’t assume you need to start with a blank knowledge base. The content you’ve already invested in may be closer to chatbot-ready than you think. We’re happy to help assess what’s already there. Talk to Hypha