Book Now

AI in Pelvic Health: How We Use It at Pelvix, and How to Use It Safely Yourself

Book an appointment
Person lying in bed at night looking at a glowing smartphone screen in a dark room.

Written by Megan Jackson, Specialist Pelvic Health Physiotherapist and Clinical Director at Pelvix.

Artificial intelligence is no longer a future technology in healthcare. It is in your consultation room, on your phone, and in the questions patients ask themselves long before they ever book an appointment. According to a recent UK survey, one in four UK patients now use AI for health information, and around one in seven have used it instead of contacting a doctor at all.

That deserves a considered response from those of us providing the care. Here is how we use AI at Pelvix, where it helps, and where it can be misleading.

How we use AI in clinic: meet Heidi

We’ve now introduced Heidi, an AI-powered medical scribe, during consultations. With your consent, Heidi listens to the conversation and produces a structured clinical note while we focus on you. The note is reviewed, corrected and signed off by your clinician before it goes anywhere near your record.

In practical terms, this means three things matter most:

  • Better attention. When I’m not typing, I’m watching you and I can properly listen instead of multitasking.
  • More accurate notes. AI scribes capture detail that humans, even good ones, lose during a busy clinic day. The notes are then quality-checked, so you get the benefit of the detail.
  • More time to think. Less time spent writing means more time spent reasoning, explaining and answering your questions properly.

On data privacy: Heidi is GDPR-compliant, data is held on UK servers, and audio recordings are deleted once the note is verified. Heidi publishes its full compliance and data privacy documentation openly, which I’d encourage anyone curious to read. We’ll always tell you when AI is being used in your consultation, and you can decline at any time without affecting your care.

AI as a research assistant for clinicians

Keeping up with the literature used to be laborious. Pelvic health is a small specialty within a much larger field, and significant papers can sit in obscure journals for years before they reach mainstream physiotherapy practice. AI tools have changed that. With careful prompting, they can summarise the current evidence on a specific question in minutes, flag systematic reviews, and locate research I wouldn’t have stumbled across on my own.

That said, this only works because clinicians know enough to spot when the AI is wrong. Which, as we’ll see, is more often than people realise.

Where AI gets it wrong: the fabrication problem

Large language models are exceptionally good at sounding authoritative. Being correct is a separate matter.

A 2024 experimental study found that GPT-4o, when asked to generate literature reviews, produced citations that were entirely fabricated in roughly 20% of cases, with a further 45% containing bibliographic errors. A broader benchmark across multiple AI tools and research domains found citation hallucination rates ranging from around 14% to over 90%, depending on how niche the topic was. The more specialist the subject, the more the AI tends to invent.

This matters in two ways. First, anything an AI tells you about a study should be verified before you act on it. The “study” may not exist. Second, the AI’s confidence is not a signal of accuracy. It will quote a fictional 2019 paper in the British Journal of Urology with the same authority it would use to quote a real one.

The pelvic health blind spot

Here is something patients should know. AI systems learn from the data they are given. Medical research has historically under-represented women, and pelvic and intimate health conditions have been particularly neglected. When an AI is trained on a body of literature that contains less data on, say, vulvodynia than on hamstring tears, it is less reliable on vulvodynia.

This isn’t a hypothetical concern. Recent analyses suggest that AI tools can downplay or misinterpret women’s symptoms in ways that mirror documented patterns of medical gaslighting. For pelvic health specifically, a field already characterised by long diagnostic delays, the risk is that AI gives confident, plausible-sounding advice that perpetuates the same gaps.

This is one reason your clinician’s expertise still matters. We know where the literature is thin, where conditions are commonly missed, and where AI is most likely to be confidently wrong.

AI for patients: useful, with caveats

I’m not trying to talk anyone out of using AI for health information. It can be useful for explaining unfamiliar terminology, helping you formulate questions for an appointment, summarising a long patient leaflet, or helping you understand what to expect from a procedure. Plenty of people now find us with the help of AI, which can only be a good thing.

Where it goes wrong is when it becomes the only voice you listen to, and especially when it’s asked leading questions. Ask AI “could my back pain be cancer?” and you’ll get an answer that lists cancer as a possibility. Ask it “what are the most common causes of back pain in a woman in her forties with no red flag symptoms?” and you’ll get something far more useful. The framing matters.

Some prompts that tend to produce better answers:

  • “What is the current evidence base for [treatment]? Include the strongest peer-reviewed sources and tell me where the evidence is weak or conflicting.”
  • “Steelman the case for and against [intervention] using only peer-reviewed sources. Tell me when the studies were published, and flag any that may now be outdated.”
  • “What questions should I ask my pelvic health physiotherapist about [symptom] to get the most out of my appointment?”
  • “Please give me the most likely benign explanations for [symptom] first, before listing rare or serious ones, and tell me which red flags would warrant urgent review.”

And one rule worth holding to: ask the AI for sources, then check whether they actually exist. If a reference can’t be verified, treat the surrounding claim with appropriate scepticism.

Red flags: when to put the phone down and book in

AI is a useful second brain, but it isn’t a clinician. There are symptoms that warrant a real assessment, not a chatbot:

  • New or unexplained bleeding
  • A sudden change in bladder or bowel control, especially with back pain or numbness in the saddle area
  • Pain that is worsening rather than fluctuating
  • A new lump, bulge or vaginal pressure that wasn’t there before
  • Pain that wakes you from sleep
  • Persistent symptoms that haven’t responded to anything you have tried

If you’ve spent more than an hour disappearing down a forum or chatbot rabbit hole at 11pm, that’s also a sign you’d benefit from speaking to a real person. Anxiety is a powerful symptom amplifier, and AI can throw fuel on that fire.

A balanced view

AI used well makes me a more present, more accurate, better-read clinician. Used carelessly, it can mislead with real confidence, and pelvic health is one of the fields where it does that most readily, because the underlying evidence base is patchier than the AI’s tone suggests.

Our position at Pelvix is straightforward. We use AI where it helps, and we treat its output the way we’d treat any other source: carefully, and with the assumption that it could be wrong. We’d suggest you do the same.

If you’ve been using AI to make sense of pelvic symptoms and want a real assessment rather than another opinion from the internet, you can book an appointment online or contact the clinic.

References and further reading

This article is for information only and does not replace individual medical advice. If you are unsure about a symptom, please contact your GP or a clinician with relevant expertise.