Pets do not improve mental health on average

Most people believe a pet is good for mental health. The largest studies do not find that. They find something narrower and more useful.

On average, owning a pet makes no measurable difference to adult depression, anxiety or loneliness. Some specific things a pet brings do show up in the data. So do some specific costs.

What the largest studies find on average

A 2025 meta-analysis pooled 21 studies and 159,322 people. Pet owners were no more or less likely to be depressed than non-owners. The odds ratio was 1.03, with a 95% confidence interval of 0.995 to 1.07.

A second meta-analysis the same decade reached the same place from a different direction. Across 32 studies the effect of pet ownership on mental health was 0.07 of a standard deviation. The authors called it "of low magnitude" and concluded pets do not seem to influence mental health.

The two English cohort studies agree. One followed 7,617 adults over 50 for six years and found no change in depressive symptoms linked to pets.

Another followed 8,785 adults for two years. It found no clear association with depressive symptoms, memory or any marker of ageing.

One researcher summed up the field fifteen years ago. A general pet effect on health, he wrote, is "not a fact but an unsubstantiated hypothesis". The studies since have not moved that verdict.

Sources: Moshfeghinia et al. 2025, Martins et al. 2023, Sharpley et al. 2020, Batty et al. 2017, Herzog 2011

Why the belief persists

The people who get pets were already different before the pet arrived.

In the English cohort, each extra depressive symptom raised the odds of owning a dog by 7 per cent. People feeling low were more likely to get a dog. Over the following six years the dog made no measurable difference.

Dog ownership is also partly inherited. In 85,542 Swedish twins, around half the variation in who owned a dog was genetic. Comparing owners with non-owners compares two different kinds of people.

Pet owners also tend to be younger, partnered, employed and living in less deprived areas. Each of those predicts better mental health on its own. A study that does not adjust for them credits the pet with what the circumstances did.

One review looked only at people with a diagnosed mental health condition. The numbers were mixed, with "positive, negative and neutral impacts".

The benefit the authors reported came from interviews with people who chose to keep their pets. That is real experience. It is not an effect size.

Sources: Sharpley et al. 2020, Fall et al. 2019, Brooks et al. 2018

What pets measurably change

Three things survive scrutiny. None of them is the pet itself.

More walking

In a UK household survey, dog owners were four times as likely to reach 150 minutes of activity a week. The odds ratio was 4.10, with a confidence interval of 2.05 to 8.19. Other exercise did not fall, so the walking was added rather than swapped.

More social contact

Across four cities and 2,692 people, pet owners were more likely to get to know their neighbours. The odds ratio was 1.61. The study was partly funded by a pet food company, which the authors declare.

A softer landing after loss

A US panel followed adults over 50 through the loss of a partner. Those without a pet gained 2.58 depressive symptoms on average. Those with a pet gained 1.21.

This is the one result where a pet seems to help with mental health directly. It is self-reported and observational. It applies to a specific moment.

Sources: Westgarth et al. 2019, Wood et al. 2015, Carr et al. 2020

What a trained dog does

A psychiatric service dog is a different question from a pet. The evidence is stronger and also narrower.

In a 2024 trial, 156 veterans with post-traumatic stress disorder were compared over three months. Those partnered with a trained service dog scored 11.5 points lower on the standard symptom scale than those still waiting.

The confidence interval was 16.2 to 6.6 points. Assessors who did not know which group each veteran was in confirmed the direction.

Allocation was by waiting-list position rather than random. The follow-up was three months.

A meta-analysis of the wider field rated none of 41 studies as good quality. The service dog result stands out because its design is better than the rest, not because the rest agree.

Sources: Leighton et al. 2024, Hediger et al. 2021

What a pet costs

The costs are rarely measured, which is not the same as being small.

Grief

Among 177 people whose cat or dog had died, around 30 per cent experienced severe grief. Grief was worse after euthanasia. It was also worse when people felt others dismissed the loss.

That dismissal has a name in the research: disenfranchised grief. Grief that is real but goes unrecognised has fewer places to go.

Caregiver burden

One study compared owners of a chronically ill animal with owners of a healthy one, matched for age and sex. The owners of sick animals reported more stress, more symptoms of depression and anxiety, and poorer quality of life.

A pet that needs care is work, and the work has a measurable weight.

Sources: Adams et al. 2000, Spitznagel et al. 2017

The hormone story that did not hold

The most repeated mechanism is oxytocin. A dog gazes at its owner, oxytocin rises in both, and the bond deepens.

The paper behind that claim tested 28 dogs and 11 wolves. A reanalysis found the key comparison between dogs and wolves was not statistically supported.

A separate 2019 study tested dog and owner pairs across five kinds of interaction. Oxytocin did not change in either species.

The hormone may still matter. The evidence for it is not there yet. A plausible mechanism is not a measured effect.

Sources: Nagasawa et al. 2015, Marshall-Pescini et al. 2019

What that adds up to

A pet is not a treatment. The largest studies find no average effect on mental health. The belief that they do comes mostly from comparing people who were already different.

A pet is also not nothing. It reliably gets people walking, and it makes conversation with strangers more likely. It seems to take the edge off a bereavement.

Those are real. They are also the same things that help without a pet.

A pet can also hurt. Grief for an animal is common and often unacknowledged. Caring for a sick one carries a cost that shows up on the same scales used to measure depression.

No study can randomly assign a pet to one person and not another. Everything here is observational, which means everything here is about association. That is the honest limit of what is known.

This note describes evidence. It does not give advice about you or your animal.