Who do students go to when they need support with their mental health?
Increasingly the question is what, not who.
The Cibyl Student Mental Health Study 2026 – now in its sixth year, sponsored again by Accenture with Student Minds in support – finds that a third of students and recent graduates (33 per cent) now use an AI tool to help manage their mental health, and almost one in five (18 per cent) name a chatbot as one of the places they turned to when they were struggling.
Over the same period, every human route Cibyl has tracked since 2021 has declined – friends, partners and colleagues down from 57 per cent to 47 per cent, family from 44 per cent to 40 per cent, school or university medical teams from 22 per cent to 19 per cent.
The one human route that has grown is the one where there is most controversy and little to no standards – the proportion who have spoken to academic or teaching staff about their mental health has risen from 10 per cent to 16 per cent.
It’s a picture of students assembling their own support from whatever is to hand – a chatbot at 2am, a feed full of wellbeing influencers, a lecturer who happens to be kind – while the official infrastructure sits in the background, often unseen.
Among those using AI for their mental health, 22 per cent use it to find clarity, search for solutions or work out a series of steps for a difficult situation, 19 per cent to vent their worries and frustrations, 11 per cent for coping strategies and help managing stress, anxiety or low mood, and 10 per cent simply as a responsive conversation partner. Smaller numbers use it to track mood and triggers, or as a kind of journal.
Cibyl’s coding of the free-text answers shows students turning to it when they find it hard to talk, or don’t want to burden anyone else – and the answers are full of apology, from “it sounds stupid but…” to “it’s embarrassing to admit…”.
Many say they use it alongside other support rather than instead of it, a few have tried its advice and rejected it, and some have ethical or environmental doubts about using it at all. But use is highest among those with low mental health, no friends or poor-quality friendships, and nine respondents described using AI to get through an acute crisis.
That fits what we found when we rounded up the research on students using AI for comfort for SUs last month – lonelier students lean on chatbots more, chatbots ease distress in the moment, and in the best trial to date texting a random fellow student reduced loneliness where texting a supportive chatbot didn’t.
Social media is playing a similar role. 30 per cent now get mental health information or advice through social media, not far behind the 37 per cent who get it through their university – a figure that has slipped from 41 per cent last year. A quarter (23 per cent) now spend five or more hours a day on social media, up from 15 per cent.
Yet when it comes to something that actually requires another person, the preference flips. Three in five prefer in-person counselling to online (60 per cent against 19 per cent), and the same goes for mentor and buddy schemes. Students aren’t choosing machines over people – they’re using machines when people aren’t there.
When University Mental Health Advisers Network members were surveyed earlier this year, the most common thing they reported was students using AI as a counsellor or therapist, and advisers weren’t sure how to respond. Most of the policy attention on AI has gone on assessment and misconduct, and almost none on what a chatbot is good for when you’re anxious, or when to stop typing and talk to a human.
The lecturer finding matters just as much. If more students confide in academic staff, then whether that member of staff knows where to send them – and whether the hand-off is a person or a link – becomes a a critical part of the system.
One respondent tried to book counselling twice, found no space, and was “only sent links for the uni’s MH/wellbeing page”. One in six (17 per cent) aren’t aware of any support provided by their university at all.
The headline Cibyl leads with is improvement. The WHO-5 wellbeing score is reported as rising from 50.9 to 56.5, low mental health falling from 23 per cent to 17 per cent, and recent difficulties from 75 per cent to 67 per cent. Maybe – but there are reasons to hold the champagne.
The sample has changed a lot. There were 4,637 respondents, down from 6,685 last year and over 12,000 in 2024. Graduates have jumped from 6 per cent to 20 per cent of the sample, full-time students have fallen from 86 per cent to 64 per cent, and the over-25s have gone from about a quarter of respondents to nearly half.
The obvious suspicion – that adding employed graduates flatters the averages – doesn’t hold up. Last year’s report found low mental health among 22 per cent of students, 23 per cent of employed graduates and 37 per cent of job-seekers, and applying those rates to this year’s mix gives around 23 per cent, not 17. But the report doesn’t break its wellbeing measures down by age or mode of study, so the other shifts can’t be tested.
Where there is a like-for-like measure, things look flat – around 49 per cent worried about their mental health daily or weekly last year, and 48 per cent do now.
And the numbers that matter most haven’t moved. 36 per cent of those who answered the section have had suicidal thoughts, and of those, 16 per cent had attempted suicide.
15 per cent used A&E for their mental health. Use of 24/7 helplines has risen from 12 per cent to 20 per cent – just as PAPYRUS, one of the few crisis lines aimed at young people, closed with immediate effect in September.
Whatever is happening to the averages, the gradients are as steep as ever. Respondents from low socio-economic backgrounds were more likely to find the transition to university hard (43 per cent against 31 per cent) and far more likely to say their mental health had declined since starting (44 per cent against 25 per cent). 47 per cent have experienced suicidal thoughts, against 30 per cent of those from high socio-economic backgrounds.
Low socio-economic students are less likely to get involved in extracurricular life (67 per cent against 83 per cent), some because of cost (16 per cent), more because of anxiety (39 per cent).
Once they’re in work, the money comes straight back – 28 per cent skip workplace socials because they can’t afford them, against none of their high socio-economic peers.
It’s at least plausible that what presents at university as social anxiety is partly years of learned financial caution – in which case treating it as a clinical problem misreads a material one.
The support preferences point the same way, with low socio-economic respondents more than twice as likely to want telephone counselling (30 per cent against 17 per cent) and mental health discussed in one-to-ones (33 per cent against 15 per cent). Private, embedded support matters most to those most exposed.
Disabled respondents are heavily concentrated outside the Russell Group – the report puts 75 per cent of them at non-Russell Group universities, against a third of non-disabled respondents – and those with a neurological impairment such as ADHD are almost twice as likely as average to have low mental health (32 per cent against 17 per cent).
Disrupted sleep, tiredness and poor concentration affect two in five respondents, and the proportion saying they are too physically unwell to use the coping strategies they would normally rely on has nearly doubled since 2024, from 9 per cent to 17 per cent – one in six who can’t reach for the music, exercise or socialising that wellbeing campaigns tell them to.
Only 20 per cent of those with a mental health condition consider it a disability, only 14 per cent have received Disabled Students’ Allowance and 27 per cent didn’t know it existed – which matters if, as the Fonagy review argues, support is supposed to become needs-led rather than diagnosis-dependent.
The brightest spot is work, albeit on a smaller base of around 750 employed graduates. 47 per cent say their mental health has improved since starting work, up from 31 per cent, and 77 per cent say their employer supports people with mental health challenges – a figure that hovered around half for years. Employers, it seems, have worked out how to make support visible and belonging early.
If students are building their own support from chatbots, feeds and friendly lecturers, the big question is whether anyone is helping them build it well.
That points to some unglamorous things – mental health literacy repeated across the student journey rather than delivered once in welcome week, academic staff who know what to say and where to send someone, honest guidance on emotional use of AI and where the data goes, and early, cheap, low-pressure ways for students to find the people they’ll turn to later.
None of that substitutes for counselling capacity or NHS services that are there when needed.
But in an era where the flashy wraparound investments of the past look hard to sustain, it’s the little things that are likely to matter most.