For four years, Firely and HL7 International have asked the global FHIR community the same questions. Each year more countries answer, more put FHIR into production, and more back it with regulation and funding.

This year 101 experts across 63 countries told us what’s happening on the ground. That’s our biggest sample yet, and more than triple where we started in 2023.

That growth shapes how to read the rest of this report. Of the 11 countries new since last year, many are earlier in their FHIR journey, so several headline percentages look more conservative than a like-for-like comparison with 2025 would suggest.

It isn’t that established programs are slowing down. It’s that the group we’re measuring keeps getting broader. Where a trend is clear we’ve called it, and where a comparison isn’t clean we say so.

When you factor that in, the overall picture is an encouraging one. FHIR has moved from a standard countries were still evaluating to one they now run in production.

One of this year’s most notable findings is what the report calls the “AI paradox.” You might assume that as AI advances, structured data like FHIR matters less. The community is saying the opposite.

So we asked a deliberately pointed question. Given how far AI has come, is it still worth investing heavily in FHIR and structured data, when AI can read unstructured data directly?

The pushback was firm. 72% disagreed or strongly disagreed that AI makes the work less worthwhile, and only 10% agreed. On a separate question, about five times as many said AI is already helping their FHIR work as said it’s holding them back.

Most respondents say AI hasn’t changed much on the ground yet. Among those who do see an effect, the biggest is increasing interest in FHIR as a foundation for AI, named by 29%. Another 15% are using AI to speed up data mapping and transformation, historically one of the most expensive parts of moving to FHIR. A smaller group does think AI reduces the need to invest, so the picture isn’t unanimous.

There have been real questions about whether AI will let us skip the hard work of structuring health data. This year’s survey points the other way: the FHIR community is using AI to help build its foundations, not to avoid them.

Among countries that regulate health data exchange, 80% now mandate or advise FHIR specifically. That’s up from 74% last year, and it continues a steady four-year climb, even with new entrants who are earlier in their FHIR journey joining the sample.

Naming a standard isn’t the same as resourcing it, and the support behind the mandate is uneven. Among countries with regulation, compliance deadlines are becoming more common, reaching 49% this year.

Public funding tells the opposite story. The share of respondents reporting funds to stimulate adoption has edged down, from 50% in 2024 to 40% now, even as more countries take part. Either funding isn’t keeping pace with the spread of regulation, or people aren’t aware of what’s on offer.

Penalties lag too. 55% of countries with regulation attach no fines for non-compliance, and 30% don’t know whether fines exist. Rules are spreading faster than the means to enforce them.

R4 remains the most common main standard, but its share keeps falling: 36% this year, down from 40% in 2025 and 58% in 2024. R5 is moving the other way, named as the main standard by 8 respondents this year, up from 2.

Two things are happening at once. Some countries are moving up to newer versions, and many new entrants are starting on a mix of versions. Both pull R4’s overall share down, for different reasons, so the decline reads as the field broadening rather than mature programs stepping back.

Those eight R5 countries are a sign of progress, but version diversity cuts both ways. With R4 still dominant, R5 growing and a tail of older versions in use, national systems increasingly have to manage more than one version at once, a persistent challenge for interoperability.

As main standards, older versions are marginal. Only the Netherlands names STU3 and only Kenya names R4B, while no respondent names DSTU2. They still show up at lower levels of use, but the center of gravity has clearly moved to R4 and R5.

R6 is the one to watch. It’s still pre-release, so adoption is minimal, but there’s early movement up the scale. Where responses used to sit at the bottom, a few now report using it at some level and one names it as their main standard. Expect that to grow once the normative R6 is released, though a full transition is likely a few years off.

Starting with what successful use cases delivered. The clearest payoff, by a wide margin, is better access to information, cited by 63% of respondents, far ahead of any other benefit.

Improved care workflows (33%) and improved health outcomes (18%) follow, with lowered cost (8%) named less often. The first win is usually just getting the data to move.

Looking back at the year’s work, in 2025 the most common achievement was launching pilots. In 2026 developing new FHIR standards for specific use cases jumped to the top (51%), a sign of the work moving from piloting toward standardizing.

Looking ahead, the priorities are much the same as last year. Expanded adoption (61%) and new use-case standards (57%) lead the list, just as they did in 2025.

Two expectations stand out for rising as a share of respondents. Expected new regulation climbed from 33% to 45%, and a national standards organization from 15% to 19%. The momentum is in the policy and governance layer, where the next phase of adoption gets formalized.

The 2026 State of FHIR survey collected 101 responses from 63 countries between February and April 2026. It’s the fourth annual edition, run by HL7 International and Firely. Respondents are national experts and informed stakeholders, with the aim of at least one knowledgeable voice per country.

A few things to keep in mind. The sample has grown a lot, from 32 responses in 2023 to 101 this year, and new countries join each year. Many newcomers are earlier in their FHIR journey, so raw year-on-year percentage shifts can understate progress in established programs.

Some questions are new in 2026 or 2025 and have no prior-year equivalent. Multiple respondents from the same country sometimes disagree. Results are best read as qualitative rather than precise measures.