In which ChatGPT again attempts to write one of my blog entries...and gets nearly every data point dead wrong.

Two years ago, with all the controversy over the frighteningly quick expansion of machine learning technology (popularly known as Artificial Intelligence (AI), even though it's not really that) over the few years into every facet of our lives, I decided to run a quick experiment using ChatGPT. My request was pretty simple:

"Write a blog post in the style of Charles Gaba."

I didn't include anything about healthcare or the Affordable Care Act...just my name.

You can see what it came back with at the link. I concluded at the time:

Aside from the fact that I'm not a pirate and don't generally use "themes" for my blog posts anyway...the larger issue here is that the text generated by ChatGPT doesn't include any specific information.

The entire thing consists of awkward metaphors and lots of filler text. This is basically the equivalent of a book report on Of Mice and Men written by a middle schooler who only skimmed the Cliffs Notes and had to pad their assignment to reach the 500 word minimum requirement.

The problem (and threat) isn't so much what AI can do right this moment. The problem is what it will be capable of six months from now or six years from now.

Well, it's been two years so I decided to give it another shot, with the exact same request:

"Write a blog post in the style of Charles Gaba."

The "good" news is that this time around, ChatGPT included plenty of "specific information." In fact, nearly everything generated was chock full of "hard data."

Unfortunately...well...here's what it came up with this time, again, verbatim. I'll add commentary as I go:

Sure! Charles Gaba is best known for his data-driven analysis of U.S. healthcare policy, especially the Affordable Care Act (ACA), with a clear, analytical, and sometimes sarcastically blunt tone. He often uses charts, footnotes, and direct references to hard numbers, and he's not shy about calling out misinformation. Here's a blog post in that style:

ChatGPT now adds a preface briefly explaining who I am before getting to the actual response, and its description is fairly accurate, I'll admit.

ACA Enrollment Hits New Record in 2025… Again. Here's Why That Matters (and Why Some People Are Still Lying About It).
By Charles Gaba, ACASignups.net | May 6, 2025

I'm not thrilled about it including an actual byline, seeing how no, I didn't actually write this.

Well, here we are again. The final 2025 Open Enrollment Period (OEP) numbers are in, and guess what?

Yet another all-time record.
Seriously.

According to the Centers for Medicare & Medicaid Services (CMS), a total of 22.1 million Americans enrolled in ACA exchange plans for 2025—that’s up from 21.3 million last year and more than double what it was just five years ago.

Um...actually, the only part of this which is accurate is that the 2025 OEP numbers are an all-time record. Aside from that,:

• The final numbers are not in yet...as CMS still hasn't published them.
• The most recent hard total I have is from 1/17/25, which put the grand 2025 OEP total at 24.2 million QHP selections, not 22.1 million.
• That's actually up from 21.4 million last year, not 21.3 million (close but still off by ~100,000 people)
• It is, however, "more than double what it was just five years ago" (11.4 million people selected QHPs via the exchanges during the 2020 OEP).

Here’s the Breakdown:

• HealthCare.gov states (33 of them): 16.9 million
• State-Based Marketplaces (SBMs) (17 states + DC): 5.2 million
• New Enrollees: ~5.7 million
• Renewals: ~16.4 million

• Actually, only 31 states were hosted via HealthCare.Gov this OEP, not 33

• Those 31 states enrolled 17.1 million people, not 16.9 million

• The other 19 states +DC actually enrolled more than 7.0 million people, not 5.2 million.

• New enrollment was around 3.9 million, not 5.7 million.

• Renewals are around 20.2 million, not 16.4 million.

Oh, and guess what? That doesn’t even include the 13–15 million people enrolled in off-exchange ACA-compliant plans, nor the over 100 million people with Medicaid/CHIP, many of whom became eligible thanks to Medicaid expansion.

Good heavens, this is massively wrong.

• There's likely only around 1-2 million people enrolled in off-exchange ACA individual market plans, not 13-15 million!
  (even if you're extremely generous and include the small group market, the entire small group market is only around 10 million people or so including both on- and off-exchange)

• And no, there's not "over 100 million" on Medicaid/CHIP; the most recent data I have on that is either 78.8 million or 80.4 million nationally, depending on whether you're including the U.S. territories or not.

So, if you’re doing the math, that’s somewhere north of 35 million Americans getting their coverage directly from ACA-related provisions. And that’s before you even touch the under-26 crowd on their parents’ plans or the pre-existing condition protections that now apply to everyone.

• It's actually more like 47-49 million, not 35 million (~24.2M in exchange QHPs; ~1.8 million in BHPs; ~21M in Medicaid expansion; and perhaps 1-2 million in off-exchange ACA compliant individual market policies which you may or may not "count" as being "ACA-related").

Why the Surge?

Let’s be real: It’s not magic. It’s policy.

• Enhanced subsidies extended through 2026 under the Inflation Reduction Act 2.0.

• The enhanced subsidies only run through the end of 2025, not "through 2026."

• There's no "Inflation Reduction Act 2.0," just the original one.

• The family glitch fix went into effect last year—finally allowing millions of dependents with unaffordable employer coverage to qualify for subsidies.

• The family glitch was fixed starting with OEP 2023...2 years ago.

• Aggressive outreach and navigator funding restored and expanded by the Biden Administration.

True.

• Continuous Medicaid redetermination policies post-pandemic leading many to transition into marketplace plans rather than dropping off a coverage cliff.

True.

And yes, Premiums did go up an average of 5.8% this year… but for most enrollees, their net premiums actually dropped due to increased subsidies and benchmark shifting. (Math. It matters.)

• I actually estimated the weighted average unsubsidized ACA plan rate hikes at between 6.1 - 7.3%, although other sources may vary.

• I'm actually not sure whether or not "most" enrollees net premiums dropped or not for 2025. They may have but we don't know because, again, CMS hasn't published the official 2025 OEP Public Use Files yet.

Meanwhile, In Misinformation Land…

Right on cue, some bad-faith actors are out here yelling “Obamacare is failing!” or “People are being forced into coverage!”—as if being offered affordable healthcare is some form of punishment.

The irony of ChatGPT calling out "misinformation" in a blog post filled with misinformation is off the charts.

And while the claims of "Obamacare is failing/people being forced into coverage" were real attacks for a long time, I haven't really heard those tossed around by Republicans for years now.

Let me be perfectly clear:

• No, people aren’t being “forced” onto the exchanges. They’re enrolling because coverage is more affordable and accessible than ever.
• No, the ACA isn't a "government takeover of healthcare." It's a regulated private market with public subsidies. You know—capitalism with guardrails.
• And no, these aren't just “handouts.” The vast majority of enrollees are working-class Americans who pay a share of their premiums and out-of-pocket costs.

All of these are true, but again, these are countering attacks which haven't really been a thing for the past 5 years or so.

What’s Next?

Well, unless Congress acts, the enhanced subsidies will expire at the end of 2026, which could cause massive premium spikes and a sharp drop in enrollment. Again.

• THE ENHANCED SUBSIDIES EXPIRE AT THE END OF 2025, NOT 2026.

• The impact of that happening is accurate, however.

But hey, there's still time to extend them. And now there are millions more reasons to do so.

Stay tuned. Numbers don’t lie—but politicians often do.

Data Source: CMS 2025 OEP Final Report. Enrollment figures accurate as of 04/15/2025. For historical context and state-by-state data, check out the ACA Signups Spreadsheet.

• THERE IS NO CMS 2025 OEP FINAL REPORT TO REFER TO AS CMS HASN'T PUBLISHED IT YET.

To summarize, by my count ChatGPT got around 4 accurate statements out of 22, with another 3-4 claims which are either unknown or open to interpretation.

Put another way, 82% of ChatGPT's factual claims, most of which are supposedly "hard numbers" are flat out wrong, in many cases wildly so.

And this is what millions of college & high school students are now using as their primary source for term papers & the like.

I weep for the future.