WebMD is no longer with the practice
- Elizabeth Restat
- Nov 18
- 4 min read
Updated: Nov 19
How AI is changing what patients gain - and lose- from online health information.
I had the opportunity last month through AI & Product Colorado to learn from Alex Liss, SVP of Data Science and AI at Huge, about optimizing web sites and brand positioning for AI.
AI is, to quote Alex, eating the web:
• There has been a 25% drop in Web Traffic, per Gartner
• 75% of searches are now being answered by Generative AI, per Brookings
• 60% of zero click searches are from Generative AI, per Bain
• There has been a 70% drop in traffic to so-called “niche” websites, per Pilot Digital
Interesting enough if you own a retail brand. But what if you own disease? I ended up wondering what it might mean for that most exalted informed-patient standard, the WebMD Phd.
Web-informed patients are, naturally, a spectrum. They might be asking Jeeves for the dosing instructions for oil of oregano or horse dewormer. They might find something well sourced but re-written for purpose, like WebMD. Best case, for a chronic condition they’ve visited the American Diabetes Association or National Kidney Foundation. An adventurous few - or those patients who are themselves scientifically trained, might have attempted to read a source study.
The grandaddy, WebMD, is deeply impacted by the AI traffic trend. It’s traffic is down between 15-50% per month, since it’s 2023 pre-ai search peak*. According yesterday’s New York Times, “A survey last year found that about one in six adults — and a quarter of adults under 30 — regularly consult an A.I. bot like ChatGPT for medical information." The ‘deliberately’ in that sentence is invisible and silent. There are quite a few more googling symptoms and getting answers from their phone’s on- board AI of choice. Google, for example, was single-handedly responsible for cutting WebMD’s peak numbers by 15% in 2024 after the introductions of Gemini.
Generative AI, LLMs, ChatBots - whatever you call them, they offer both risk and opportunity in a changed behavior.
Pro: Someone’s weapon of choice chatbot instance knows them pretty well. It knows their reading preferences and writing style. It’s going to know how to display information at the right level and how to talk to a human so they listen. According to the NY Times article I mentioned above, “so many users told us that a big part of the appeal was that the A.I. offered a kinder version of health care.”
It tells them they’re right. It’s nice to them. Who could resist something like that in their own home?
Con: It knows them well within a specific sphere. The same chatbot that I’m asking to fix my fantasy football team might have a really specific but lopsided sense of my strengths and weaknesses. That knowledge isn’t always benevolent enough to want to keep me alive.
“Should I drink this mystery fluid I found?” is enough to give it an impression of my intelligence level; depending on its ownership and motivation, it might think it’s worth it to tell me to jump.
Pro: Someone’s AI is more likely to give them a range of sources; if they trust the AI but have a bias against certain information it could serve to wedge open a closed mind.
Con: Anyone with a well-optimized and authoritative site can be indexed and creep in to an AI response. It’s easy to avoid looking at the source of the answer to your vital health question came from- in some cases it’s much harder to actual find attributions. Perplexity is kind enough to tell me with a footnote where it found what it tells me. Others less so. We’re more dependent than ever on the integrity of source material, but less likely than ever to know where it came from when our search engine is a commercial AI product.
AI might be more persuasive than the “raw” web and be pulling from even more bananas sources. Larger AI vendors have stopped including disclaimers or otherwise footnoting what might be erroneous health advice. This is before anyone starts trying to sell anything. Say what you will about Gemini, at least we can pretend we understand their business case and motivations. Quite a few others are still up for grabs.
We need to use the demise of traditional search as an opportunity to push higher quality information back down the bell curve of patient information.
What can be done? Optimistically speaking, with an open-minded moment, comes a brief but crowded hour for actual facts.
• Anyone running “niche” health sites - be they for a specific chronic condition or for an overall health network - optimize your site so it’s easily recognized as an authority during AI trawling. This effort does have a specific set of best practices, and for that I refer you back to Alex and his team at Huge for recommendations. At a high level, though, think structured information intended more to be “Read” than actually “Seen” by more and more of its consumers.
• Individual health educators - If you’ve been looking for yet another reason to get AI savvy, here it is. If you haven’t, start playing with Open AI/ Perplexity. Keep it available, preferably on a format that has attributions or one that’s provided with safe sourcing. You never know, the thing might actually have a good idea.
The gradual demise of WebMD and the rise of the AI informed patient requires vigilance—embracing AI’s charismatic bedside manner while remaining critical of sources and insistent on traceable, quality attribution. For those out there who are the sources, the challenge is to make yourselves as consumable as possible while still maintaining approachability for the patient.
*
Month | 2025 Estimate (visits) | 2024 Estimate (visits) | 2023 Estimate (visits) | Value Drop | Percentage Drop |
January | 52,000,000 | 65,000,000 | 70,000,000 | 18,000,000 | 25.7% |
February | 52,000,000 | 64,000,000 | 70,000,000 | 18,000,000 | 25.7% |
March | 51,000,000 | 63,000,000 | 69,000,000 | 18,000,000 | 26.1% |
April | 50,000,000 | 61,000,000 | 68,000,000 | 18,000,000 | 26.5% |
May | 32,000,000 | 47,000,000 | 67,000,000 | 35,000,000 | 52.2% |
June | 33,000,000 | 45,000,000 | 66,000,000 | 33,000,000 | 50.0% |
July | 42,000,000 | 50,000,000 | 65,000,000 | 23,000,000 | 35.4% |
August | 41,000,000 | 50,000,000 | 65,000,000 | 24,000,000 | 36.9% |
September | 52,000,000 | 51,000,000 | 64,000,000 | 12,000,000 | 18.8% |
October | 53,000,000 | 50,000,000 | 63,000,000 | 10,000,000 | 15.9% |
November | [Awaiting data] | 49,000,000 | 63,000,000 | ||
December | [Awaiting data] | ~48M | ~62M |
Sources: Public traffic estimate platforms (SimilarWeb, SEMRush, Ahrefs, Sensortower), LinkedIn, and news analytics.semrush+6
Comments