Pew Research surveyed 5,111 Americans last October to find out how they find health information. The first source, their doctor, wasn't a surprise but the second most popular was "patients like me" at 66%.

When asking only the people that get information from these sources, the researchers can get a view into how well they work. While AI chatbots are at the bottom of the list for usage at 22% for sometimes+ they are understandable and convenient.

We need to make an observation about "people like me." While social media is at near the bottom for usage much of the information from other patients will come via forums or groups such as those on Facebook or Reddit. This highlights that for social trust depends more on who is speaking than the channel itself.

When asked if they see or hear conflicting information 76% say sometimes or more often. This number changes very little by age group.

When confronted by these conflicting pieces of health information, 54% of Americans have a hard time assessing which ones to trust. This also doesn't vary much by age, but we do see the hubris of youth showing itself very subtly.

The full report is 79 pages in PDF form and Pew Research's impeccable quality means that this gets the "often imitated, never duplicated" Klick Wire Must Download award (I've never seen anyone bother to imitate it -Ed.).

Qventus surveyed more than 60 provider CIOs, Chief AI Officers, CMIOs, and other senior IT leadership about their AI ROI. Important to remember is that Qventus is an AI purveyor for this industry.

When asked where they felt AI was needed, revenue came first followed by admin tasks. 

Epic has been making strides to integrate AI with its core EHR systems but not all vendors are at the same level. This is shown with the EHR vendors showing as the top obstacle to AI execution.

Healthcare is traditionally a very conservative industry when talking about new technologies. The researchers asked about agentic AI and 40% they are already investing with 57% staying "healthy skepticism."

Klavio surveyed 8,000 global consumers to determine their opinions about AI in the shopping journey. This isn't the same as the patient journey but is still interesting insight nonetheless. 

First, more people (49% vs. 21%) trust AI responses, likely interpreted as chatbot answers by survey respondents.

Next, consumers tend negative when they see AI being used in brand advertisements. The question here matters though, it was:

The word "clearly" is doing a lot of heavy lifting here. It implies that the materials in question are AI slop and this will necessarily skew the results negative.

What would have been the result if hey had asked "What is your reaction to brands that use tasteful AI-generated content in their marketing?" (no hypotheticals -Ed.).

The 2027 FDA budget requisition highlights many of the projects the agency has ongoing as well as the financial requirements to keep it operating. 

The FDA request has numbers from 2025 to the request for 2027. However, that's not a lot of data so the Wire authors dug up some historical data and added the latest numbers to it (the dashed line from 2024 to 2025 was a missing data point).

You won't find this chart anywhere else. Need a high-res version for a presentation? Just reply to this email.

The batch of NOV letters from September last year weren't a one-off occurrence, the agency is asking for legislative support for its actions. A section in the budget request is titled "Holding Firms Accountable Using Misleading Advertising to Drive Profits at the Expense of Consumers."

That pretty much sums up the FDA's approach. The requests for new legislation in the document aren't technically new powers, but they enshrine in law what is currently considered interpretation and guidance. 

The Klick analysis of the September 2025 letters also shows what the agency is looking for:

ZeroBounce surveyed 1,157 professionals in the United States and Europe in February to find out how they handle work email. The results show an always-on world, even for email.

The most surprising finding here isn't that only 23% say they check email during vacation (hold this thought for later) but that 18% don't check work email at work! Perhaps these are professionals such as physicians who are busy with patient consults rather than reviewing emails (that's speculation -Ed.).

Above, only 23% said they check email while on vacation but here when asked how they handle email when on vacation only 29% say they fully disconnect... so 71% check email?

Here is the third way the researchers asked about email on vacation. Here, they asked why individuals check email and only 19% said they never do.

So this study is a perfect example of "it matters how you ask the question" (not an actual quote -Ed.). We got three different numbers from three somewhat different questions:

Three ways of asking the question, three different answers of 23%, 71%, and 81% saying they check emails on vacation. Keep this in mind when interpreting all survey data.

When the researchers asked where else professionals check their email, the motorcyclist in me was disappointed but not surprised.

If you need a proof point for email in your next deck, you have a variety of answers here to choose from.

There are two stories about Medvi circulating:

It's neither. To explain, take our model of luck. To have "luck" you have to have three components:

Matthew Gallagher had all three components. He didn't have a magical "in" or an unrealistic connection that no one else has. He just saw what was possible. From the New York Times article:

The "issues" you'll read about Medvi are pretty standard mid-sized company problems. 

So, can someone duplicate Medvi and succeed? Likely not because they can probably pivot faster than the newcomer. However, duplicating the good idea isn't where the Luck is. Look for your opportunities and make sure you have the "ability to act" and don't listen to the green-eyed monster when others try and tear Medvi down.

OpenAI is alerting political leaders that AI will change the very structure of current capitalistic societies ... which, don't kid yourself, is essentially all of them (a bit political but I'll leave it -Ed.).

Looking at the roll-out of compute power and the advances being made in frontier models the above is not an overstatement. The paper compares the changes coming to "the Industrial Age, the Progressive Era and the New Deal" (I find the lack of an Oxford comma disturbing -Ed.).

OpenAI seems to think that human-centered work can absorb workers displaced by AI. Maybe.

The company also acquired Technology Business Programming Network (TBPN) as it expands its PR resources.

Digital health funding deals were 33% up in dollars and 10.5% down in volume from Q1 2025.

In their analysis of the first quarter, Rock Health included a chart of who shares data with whom. Perhaps surprisingly, Perplexity has the most integrations.

Reporting on a study in JAMA this month, CNN highlights what types of advertisers end up on medical misinformation websites. They charted the amounts from different health-related advertisers.

The study spanned four years and the amounts are totals across the entire span.

Individual entity spending was much lower than the categories, of course.

As John Wannamaker famously said: "Half the money I spend on advertising is wasted; the trouble is I don't know which half." 

These amounts are certainly very small percentages of the totals for these groups. While any waste isn't good, these need to be combatted because:

In a set of 10 questions about AI where three were negative and seven were positive, the negatives were at the top of the results for Americans. The one outlier was being more efficient at work.

Emarketers shows us that MMM is the top measure for incrementality of initiatives by advertisers.

I'm not embarrassed to admit that I had to look up "ghost bids." It's an interesting measurement tactic.

Gallup's latest global employment report shows that employees and managers have lower engagement in 2026 but other metrics are up. In particular this report shows how "best performers" can be far from the mean when it comes to engagement.

Meanwhile there is a correlation of employees who use AI and who agree their company is supporting AI. This is not necessarily causation, but it's still interesting.

If you're looking for engagement numbers for your specific country, this 251-page report likely has them.