The claim: A new Nature Medicine paper reports an AI system outperforming both frontier models and human clinicians on judged CBT competencies.

On paper, that’s a big result.

The details: Researchers tested a large language model wrapped in a “cognitive layer,” which is essentially a reasoning scaffold designed to guide therapeutic responses.

Findings included:

Important context: The lead affiliations are with the company that built the system (Limbic).

Industry research isn’t inherently invalid, but it does raise the bar for independent replication.

In mental health, “promising” and “ready” are not synonyms.

What the paper shows well:

AI can perform strongly in structured, text-based CBT interactions.

What it does not establish:

That gap matters. Psychotherapy is widely considered a high-stakes domain requiring cautious, phased evaluation and strong risk monitoring.

And the broader evidence base is still thin: a recent systematic review found LLM mental-health chatbot research methodologically fragmented, with ongoing concerns around safety, governance, and bias.

The deeper tension:
The mental health system desperately needs scalable support. That urgency can make polished evidence feel more settled than it really is.

The real takeaway:
This paper isn’t a coronation. It’s a signal.

The interesting result isn’t that a chatbot beat therapists. It’s that engineering scaffolds can make general AI models look clinically coherent in CBT-style conversations.

Useful? Potentially.
Settled science? Not even slightly.

The clinically sane response:
No panic. No applause.
Just evidence standards that remain inconveniently high.

Sources: Nature Medicine | PubMed | npj Mental Health Research | JMIR AI | American Academy of Arts and Sciences

Where to start. War rages on, and Nigel Farage’s touring Reform rally rolled into my hometown on Thursday, all pyrotechnics, strategic lighting and careful camera angles. (For those that don’t know, Nigel Farage is akin to the Donald Trump of the UK). The room, notably, was seated, and it was not full. Empty seats everywhere, which is perhaps why the shots were so artfully framed. Their Chair warming up the room, an ex-doctor no less, told the crowd he’d seen someone shooting heroin in a doorway on the way to the venue. At which point my husband had to verbally de-escalate me/threaten to cut my live feed. Farage himself attended, naturally. The whole thing was part spectacle, part grievance convention, part expensive attempt to make decline look charismatic.

The demographic was the one you’d sketch if someone asked you to draw “certain kind of man with Opinions.” The fake news was galling, but so is what happens afterwards, when the same mood music spills onto my councillor social media page. The abuse is vile. Misogyny in all its colours. Some of it stupid, some of it creepy, some of it so tedious you can almost hear the damp in the room it was typed from. My skin is getting thicker, which is useful, because it does push all the old buttons about appearance and being looked at before you’re listened to. There is always a price women are expected to pay for taking up public space, and the bill keeps arriving.

The young women I work with deserve so much better than this stew of swagger, disinformation and resentment. They deserve a world in which power is not performed through intimidation, where being loud is not mistaken for being correct, where their bodies are not public property and their fear is not political collateral.

I watched Louis Theroux’s documentary Inside the Manosphere, and the title feels too soft for what it is. This is not some fringe internet oddity bubbling away in a basement. It’s not “emerging.” It has arrived, unpacked, and started writing policy. That, for me, is the point. We are long past the stage of treating misogyny as a niche subculture for lonely boys with podcasts. It is shaping language, confidence, dating, politics, safety, the lot. Pitiful, yes. But also dangerous, organised and profitable. We do ourselves no favours by calling it pathetic and leaving it there.

So that’s me this week: watching a travelling theatre of masculine grievance roll through town, deleting abuse from my socials, sitting with young women who will inherit the fallout, and finally the pièce de résistance: tonight (Monday, when writing this) I am presenting a motion on potholes (yes, love, potholes) at full council. In a bid to make myself look less ‘dug up’ I hopped on Uber Eats to buy a face mask, only to be presented with a Butt Mask instead. Don’t get me wrong, butt acne is real- but this, with a bloody Corgi cartoon on the top- takes the cake. Megababe Le Tush cream anyone? FML

If you’re a therapist who likes to zoom out from the therapy room and think about systems, relationships, and culture, this conference tends to draw some of the most thoughtful conversations in the field.

The basics:

What to expect:

A chance to hear how therapists are thinking about technology, culture, and social systems shaping mental health

In a moment when therapy conversations are increasingly dominated by individual-level solutions (apps, AI, productivity hacks), systemic perspectives offer a useful reminder: people exist inside relationships, families, and cultures.

👉 Learn more: https://www.aftconference.co.uk/

Why it matters:
For many people, looking at you, residents of the UK and the northeastern U.S., this has been a long winter. Less daylight, lots of snow and rain, disrupted routines, and months of gray weather all contribute to Seasonal Affective Disorder (SAD).

Good moment to assess:
As days get longer, some clients feel relief. Others are still carrying winter-related depressive symptoms.

Questions you might ask:

And for therapists:
Seasonal mood shifts affect clinicians, too. If this winter felt heavier than usual, it may be worth your own seasonal check-in, more light, more movement, and more connecting with your psychiatrist if needed.

The bias came from the humans, not the machine.

A new study in Communications Medicine examined how bias showed up in clinical notes written by healthcare providers and later used to train pediatric mental health AI models. Researchers analyzed ~20,000 pediatric anxiety cases and matched controls (ages 5–15) from Cincinnati Children's Hospital Medical Center.

What they found:

What helped:
Researchers tested a data-centric de-biasing approach that included filtering out biased terms and normalizing language patterns. That reduced diagnostic bias by up to 27%.

Source: Communications Medicine, March 5, 2026

This week I read the latest Jenny Colgan novel, and of course, I loved it. If you’ve never encountered them, imagine a Hallmark movie set in Scotland, with an overworked, intelligent young woman from the big city who is forced to move to a small coastal town, where she slowly falls in love with the landscape, the community, the local food… and usually a strapping local man.

They are a light mental lift. Cozy and predictable in the best way, and full of bakeries, bookshops, seaside winds, and small-town gossip.

Yes, it’s the whole film version. No, I haven’t finished it yet. But please, just watch it!!!

Stories from the community

Response rate: our lowest ever. Embarrassingly low. 🙈

Correct answers: 1 (congrats to the reader who said “One Battle After Another.")

By focusing on process and not content, we got the community’s message loud and clear: more therapy content, fewer Hollywood predictions.

Note taken.