What Does Social Media Say About Perinatal Mental Health?

An analysis of the public discourse of perinatal mental health on social media

Writer's note: this piece was co-written with Dr Juan Chart-Pascual

"But doctor, that's not what it says on the internet."

This is something a patient being treated for bipolar disorder told my colleague Juan during a routine consultation. She wasn't challenging his advice; she was just telling him what she had read on social media and asking how it fit with his recommendations.

That one sentence stayed with him since and sent us both down a rabbit hole. If patients are forming their ideas about pregnancy and mental health on social media - often long before they sit across from us in the clinic - shouldn't we, as clinicians, actually know what information they are finding there?

Image Source: Berke Citak on Unsplash

We are Nuria and Juan, two psychiatrists and researchers who couldn't shake that question. So, we did the only sensible thing two curious clinicians can do with an unanswered question and far too much enthusiasm: we conducted research to find an answer. Earlier this year, we presented our research at the UK and Ireland Marcé Society conference; now, through this piece, we want to share with you what we found, but first, let us tell you what we did.

What Did We Do?

Firstly, we used an AI model to sift through over 300,000 public posts on X (formerly Twitter), spanning eleven years, from 2014 to 2024. Then, we rolled up our sleeves to figure out what the internet really says about perinatal mental health, which refers to the mental health of parents during pregnancy and up to one year after childbirth, and how much of it we can trust.

In practice, we used one AI tool to automatically group posts into recurring topics, and another one to gauge each post’s tone, and determine whether it was positive, negative, or neutral. We then checked a sample of the results manually to make sure the AI had got both the topics and the tone right.

That process gave us 122 fine-grained topics, which we then grouped by hand into broader, more meaningful themes by reading the top terms and example posts from each one ourselves. We came up with five broader themes that accounted for 70% of all the posts that could be categorised, as summarised in the image below.

Image by the authors.

What Did We Find?

Substance Exposure & Neurodevelopmental Risks

The single biggest theme we found — accounting for more than a quarter of all the posts we could categorise — was about substance use in pregnancy and its effects on the unborn child's developing brain.

Most of it concerned alcohol consumption during pregnancy and foetal alcohol spectrum disorder: a well-established, genuinely preventable harm. Its prominence is honestly a good sign, showing that people are paying attention to some real concerns.

However, tucked inside that same theme was a smaller, noisier cluster: posts linking paracetamol use in pregnancy or childhood vaccines to autism and ADHD in children. The science behind these claims is extremely limited and contested, yet online, they are often presented with the same urgency, the same capital letters, the same "did you know?" energy as posts warning about genuine risks.

If you are scrolling at 3 am, mid-pregnancy and anxious, social media does not hand you a neat, fact-checked list. Instead, it presents you with real risks and rumours side by side, dressed in identical fonts. To us, that is the finding most worth sitting with, as it makes evidence-based guidelines and unsupported claims difficult to distinguish, fuelling misinformation and undermining informed decisions.

Lived Experience of Perinatal Depression & Birth Trauma

The second-biggest theme was also the one that scored most negatively in our sentiment analysis. It was about people describing their experiences of birth trauma, sleeplessness, or simply feeling "hit hard”, which likely explains the sentiment score of this theme.

These social media posts did not use detailed clinical vocabulary, but raw descriptions of lived experiences. Maybe that's exactly why they gain traction among readers: not because they teach them anything new about mental health, but because they evoke the sense of someone else going through the same experience.

Image from Pavel Bekker on Unsplash

While sharing difficult experiences with others who have been through similar situations can provide validation and reduce isolation, it can also spread misinformation and expose people to unsolicited and questionable advice.

For example, one pattern that stood out to us as psychiatrists, was people using "baby blues" and "postpartum depression" interchangeably. They are not. Baby blues are a normal, time-limited dip that most new parents experience, while postpartum depression is a serious, potentially life-threatening illness. Blurring the two can cause damage both ways — turning real depression into something to just push through, or an ordinary hard week into a diagnosis.

The Quieter Conversations

The other three themes were smaller, but each had its own flavour.

One focused on anxiety, panic and somatic symptoms, which made up nearly 13% of posts. Pregnancy-related insomnia was by far the biggest driver here. Yet, the theme also covered birth anxiety, panic attacks, and postpartum anxiety more broadly, alongside quieter concerns like anxious thoughts about motherhood itself.

Another theme, clinical postpartum depression (PPD): screening, treatment and research, reflected a more clinical corner of the internet where people discussed screening guidelines, ongoing research, and treatment recommendations for postpartum depression. Unsurprisingly, this was the only theme that stayed fairly emotionally neutral throughout, highlighting the contrast between discussing clinical guidance and sharing lived experiences. If anything, that's encouraging, as it shows a real appetite for accurate, evidence-based information online — a chance for clinicians to meet people where they already are.

Finally, there was a reproductive and physical health theme, spanning from premenstrual dysphoric disorder to contraception side effects to the kind of miscarriage grief nobody quite knows how to talk about. This theme offered a very powerful reminder that "perinatal" mental health doesn't politely begin at conception and end 12 months after childbirth.

A Decade of Content Getting Harder to Read

Here's the finding that unsettled us the most. Over half of all 305,619 posts — 54%, to be exact — carried a negative tone. What's shocking is that it got worse, year after year, for ten years straight.

This wasn't just a vibe. It was a clear pattern in the data, one that held up no matter how you looked at it.

So, What Do We Actually Do With This?

Honestly? Show up more. People are getting genuine comfort from strangers online describing the same spiral they are in, and that's not something to roll our eyes at; it's something worth protecting.

But here is the catch: people don’t simply find comfort online, they also find inaccurate information, unsolicited advice, and negative content.

Misinformation doesn't walk in wearing a sign. Paracetamol and vaccine claims sit right next to real, evidence-based warnings and waiting until a myth goes viral to correct it, is too late. Clinicians and scientific bodies need to be in the feed before that happens.

Image by National Cancer Institute on Unsplash.

Why This Matters Beyond the Like Button

It would be easy to dismiss this as noise, but at this scale — hundreds of thousands of people, over more than a decade — it starts to look less like noise and more like a barometer.

This data tells us where anxiety concentrates, where misinformation finds an opening, and where people feel so unheard that they would rather tell a stranger online than their GP. Listening to that properly might be one of the more useful things we can do with all this AI everyone keeps talking about. Because the goal isn't to pull people away from social media, it's to make sure what they find there is accurate, and that reaching out for support never feels wrong.