Reuters Events | Pharma

Thought leadership and innovation for the Pharmaceutical Industry

AI-powered chat that looks set to upend the patient search experience online is both threat and opportunity for pharma

By Andrew Stone on Jun 19, 2023

Just as pharma was beginning to understand and work with the way patients gather healthcare information online, the paradigm is set to change utterly. Thanks so very much ChatGPT.

Given the sudden mass global adoption of AI tools powered by large language models, rapid disruption looks certain. The tools, now increasingly on our smartphones, offer highly plausible and personable ways to find convenient answers to the billion health related questions humankind asks daily.

AI-generated health advice is out-ranking physicians on both quality and empathy.

The allure is clear. AI is proving more empathetic than GPs and it can provide some startlingly in-depth insights – biohackers are already using AI to give them context on their health data, such as bloodwork. No HCPs required.

Empathetic AI 
Stakeholders, from patients to HCPs, have always tended to start their search with a question, asked in their own natural language. AI-driven, chat-based search platforms transform this experience significantly from a messy and somewhat random tour of ranked websites into a concise, apparently authoritative answer presented in a conversational style that is highly engaging to the searcher.

The personalised and conversational nature of AI chat mimics a relationship that in turn engenders additional dimensions of trust.

These responses flow from question to question across a topic in a seamless digital information experience that is frequently infused with statements of empathy for patient concerns. 

It is entirely plausible that patients will be spending more time with ‘Doctor GPT’ _and ‘Doctor Bing’ _in the very near future than they will with ‘Dr Google’ or with their actual in-real-life doctor.

The urgent question before every pharma organization is how to respond as AI-generated ‘conversations’ replace traditional search to influence decision making across the healthcare ecosystem.  

The implications are profound. 

As it stands today, in most markets outside the USA, pharma has little to no voice online when patients are searching for answers in conventional search resulting in a significant digital disconnect. Chat-style AI that scrapes content from the visible web poses an even bigger challenge. “There is a genuine urgency to understand and influence the new digital information experience to ensure patients and healthcare professionals are receiving locally relevant, evidence-based answers when they engage chat tools for health advice,” says Julia Walsh, CEO of search listening agency Brand Medicine International who recently convened a cross-industry roundtable which has now reported its findings in a white paper Responding to AI Disruptions in Search.

This urgency is compounded by the decidedly uneven results chat-style AI offers, creating even graver risks of misinformation than the already broken search experience. 

For example, in a recent test of sample cancer research queries Brand Medicine’s Director of Analytics and Innovation, Dr Lillian Rigoli conducted, half of the citations from ChatGPT and all of those from Bard were invalid.

Part of the issue is that these tools are trained only on the ‘visible’ web, says Walsh. “This is the content with high current domain authority that sits in front of fire- and pay-walls.”

These partial selections include invalid sources and are further compromised by AI ‘hallucinations’, also known as ‘making stuff up’.

In many ways it’s an even murkier experience than existing search yet delivered with conviction and apparent empathy. “We are now navigating an entirely new digital landscape with diminished visibility of the digital information experience from the end-user point of view,” says Walsh. 

“There’s also little accountability over content given that some of these tools do not even offer the sources used for their answers, and for those that do, an alarming number of those references cannot be verified.”

From SEO to SCO
Current approaches to understanding and managing the patient (and HCP) search experience are will soon need to be superceded. Search Engine Optimisation (traditional SEO) is likely to diminish in relevance as Search Chat Optimisation is poised to become the new path to digital customer engagement.

This requires a rapid reorientation by every pharma leader working on engagement and a commitment to a long-term response as AI services evolve, advises Walsh. “We must accept it is happening. Search has changed forever and will not revert. We must tackle it as one would prepare for a marathon: we train, and then lean into the challenge to adapt to this new digital landscape.”

One long-term task facing the industry at large is persuading lawmakers to adapt the regulatory landscape so that pharma has a voice, she adds. “Legislation has failed to ensure the delivery of locally relevant, evidence-based health advice within traditional search in most markets, so regulatory reform needs to leap ahead to protect searchers from the influence of mis- or disinformation in AI-driven, chat-based search platforms.”

A focus for lobbying here is to argue for the need to permit sponsors of medications to openly publish customer-responsive content online (for both patient and HCP stakeholders) that accurately answers key queries in a locally relevant context. This will offer evidence-based content for AI chat tools to scrape from the internet to inform their answers.

Opportunity as well as threat
The threat that AI could make it even harder for pharma to make search results safe and accurate is clear, but there are also opportunities.

Developed intelligently, AI could serve as a powerful tool to support the delivery of patient-responsive content that accurately answers specific questions patients ask, providing a positive online experience that swiftly guides patients from initial symptoms to accurate diagnoses, and on to effective treatments.

There is already some evidence that it is possible to improve the health-related digital information experience. For example Med-PaLM 2, developed by Google, shows how AI could help provide genuinely authoritative insight by facilitating the understanding of complex medical texts. 

It has been recorded as accurately answering medical questions and achieving an ‘expert’ level performance on United States Medical Licensing Examination (USMLE) -style queries with 86.5% accuracy.

With such examples in mind, it is possible to imagine how pharma can harness AI to provide a far superior patient experience by putting patients at the heart of patient education and customising the chat experience for them.

“When we provide relevant content in an effective format that takes into account varying levels of health and digital literacy, we empower patients and build trust. We can design custom chatbots to manage the digital information experience by exclusively using approved, locally relevant, evidence-based content,” says Walsh. 

“Recognising that patients will use these tools regardless of warnings against doing so, it is imperative to intercept their digital information experience to offer precise responses. By providing a similar conversational experience, we can preserve the integrity of information delivery in relation to prescription medication.

“Currently there is little transparency on who is providing the information – _is it trustworthy? Is it applicable? …Up to date? We need to create a pathway to personalise it for our stakeholders and ensure the correct resources are used as a source of content for chat-based search responses.

Hybrid consultations
Another significant opportunity lies in leveraging patients’ existing search behaviour and augmenting it with expert advice. It’s possible to imagine solutions HCPs could use to conduct hybrid consultations using AI tools. 

By harnessing the vast knowledge accessible through these tools during a consultation, there is potential to expedite the patient journey by reinforcing the HCP’s perspective with the wealth of online knowledge available in the specific area. 

This approach enables HCPs to be present during the interaction, identifying quality information from misinformation in real time before the patient has the chance to be affected by any false information, or worse – to act upon it.

More broadly many small efforts within pharma have the potential to collectively drive a significant shift away from misinformation and reduce hallucinations associated with AI-chat, adds Walsh.

The way forward
A key finding from the workshop convened by Walsh with a panel of industry leaders is that Pharma’s response should be collaborative, should involve all key stakeholders and should urgently address how to harness the opportunities while minimizing the threats of AI.

“Given the scale of activity and the implications of a poor digital information experience on quality use of medicines, patient outcomes, asset (brand) reputation and commercial returns, putting our heads in the sand and staying out of the fray is not an option.”

The recommendations from the workshop include:

1.    Establishing a ‘Digital Health Literacy & Communications Advisory Board’ that includes pharmaceutical companies, patient advocacy groups, and other regulatory and government stakeholders can help maintain focus and momentum. This representative body, ideally also connecting directly with representatives from generative AI providers such as Microsoft and Google, could start by issuing a position statement on AI tools and continually review and respond to the evolving landscape.

2.    Retooling existing educational resources to be AI friendly. This offers a benefit to patients if industry can design their own resources to provide accurate information in a way mirrors the dynamics of chat that patients are embracing. By ensuring that rigorously reviewed patient (and HCP) education tools can compete with AI-chat on the open web in terms of user interface we can retain customer engagement on evidence-based educational platforms.

3.    Creating dedicated teams within pharma tasked with a timely response to AI. This includes consider splitting internal teams in toone ‘business-as-usual’ _team and another ‘AI-directed communications pilot’. This pilot, ‘skunk works’ -style approach, with a second team that comprises ‘Frustrated Enthusiasts’ _and ‘Digital Change-Makers’, would be empowered to trial new solutions to engage customers, and given permission to fail with an understanding that this is new territory that members of the team are unlikely to have experience in. This will require a culture shift from leadership down towards innovation and experimentation where there is permission to fail. 

Crucially, in the face of this profound disruption to the way questions are exchanged for health advice, all stakeholders must stay agile and responsive to change to maintain relevance in the rapidly evolving digital landscape.

Download the Responding to AI Disruptions in Search (Health) white paper here.

To register your interest to attend a webinar on this topic, submit your details through the above link.