How Outdated Science Stays Alive — And Why AI Is Making It Worse
A podiatry project, a Google AI overview, and what happens when the consensus moves on but the internet doesn't
Hi! I’m Rachel Pascal. I’ve spent more than a decade writing health communications, so I understand exactly how it gets shaped and packaged before it reaches you. Make Sense Health pulls back the curtain on that process, so you can read health information more clearly and make better decisions with it. Subscribe to get more like this straight to your inbox.
A few years ago I damaged my foot pretty badly — a Lisfranc injury and a cuneiform fracture.
The Lisfranc is a sprain or rupture of the ligament that holds the midfoot together, named after the French surgeon who first described it in Napoleonic soldiers falling from their horses. The cuneiform is one of the small bones of the midfoot, and a fracture there often accompanies the ligament damage. Both are rare (the doctor in A&E said he’d never seen this particular injury before) and slow to recover from.
In my case, the cause was slightly less dramatic than falling from a horse. I was doing a spinning kick in a martial arts class. Same injury, different century.
I wasn’t able to put weight through the foot for months, then wore a ‘Jurassic Park’ boot (as my colleagues at the time called it) for several more months (even at my own wedding). After that, othotics were part of my recovery for a long time.
If you’d asked me then how they worked, I would have said: they correct the alignment of the foot. They support the arch, guide the foot into a more natural position.
That’s what I assumed. It’s what most people might assume.
It is also, as I’ve discovered working on a project over the last few months, not quite right.
I’ve been working with a podiatry client for several months on a full content and strategy project — reviewing the clinical evidence base, rewriting his website, and now producing a series of SEO and AEO-optimised articles for prescribing clinicians. It’s one of those projects where you go deep into a subject and keep finding new layers.
This week, I’ve been writing a blog post that tackles a question that sounds simple: do orthotics correct alignment?
The short answer, according to the current evidence, is no — not reliably, and not through a mechanism the research supports.
But let’s reverse for a moment (podiatry joke there, for anyone who appreciates it)…
The framework that dominated podiatric education for many years proposed that orthotics work by returning the foot to a neutral position, correcting the mechanical deviations causing pathology. It gave the profession a shared language and a logical rationale for orthotic design.
But, when researchers started testing its central claims, it turned out to be wrong about the mechanism.
What the evidence consistently shows instead is that orthotics work by changing the forces acting on specific tissues rather than repositioning the foot. The clinical implications of that shift are significant: rather than seeking a neutral position to return the foot to, the question becomes which tissues are being loaded beyond their tolerance, and what modification would reduce that load enough to allow healing.
While I was writing, I noticed something important.
The Google AI overview for “do orthotics correct alignment” said this:
“Yes, orthotics correct foot and ankle alignment, which can help manage posture and relieve strain up your body.”
I’d spent the day working through the evidence that says exactly the opposite. So I decided to test it — to see how far I’d have to push AI before it recognised the problem with its own answer.
I copied the overview into Claude and asked it to analyse it against the current evidence.
It spotted the problem immediately. “Correcting alignment” and “guiding feet into a more natural position” is based on the older framework — one the research has been challenging since the mid-1990s.
One prompt. No pushing required.
The AI summary had synthesised the consensus — the muddled, outdated consensus that still exists across thousands of web pages, product descriptions, and patient information leaflets. It knew the right answer when asked directly. It just hadn’t applied that knowledge to its own output until I asked it to.
This is the AEO problem in miniature.
AI overviews summarise what’s already out there. If what’s out there is thirty years out of date, the overview will be too.
The only way to correct it is to put better content into the ecosystem. Well-structured, well-evidenced, clearly authored content that gives the AI something more accurate to draw on. That’s what we’re building for this client. One clinical article at a time.
As for my own orthotics, I now understand more precisely why they helped, and probably could still help me with the ongoing pain my foot gives me!
The mechanism was always kinetic, I just didn’t know what to call it.
And neither, it turns out, does most of the internet.
If you work in healthcare, run a clinical business, or produce health content of any kind — and you want someone to go to the evidence, build the argument, and produce content that holds up to scrutiny and performs in search — I’d love to hear from you. Drop me a DM or check out my professional website here
.
I must admit that, when I search for medical information, the Google Gemini results which appear first do not always use what I would consider the most reliable sources. I assume that this is because a lot of medical research papers and other medical evidence are not freely available to the public and require some sort of payment or subscription.