The Samaritans Radar web app runs on Twitter’s API. When enabled by a user, it scans friends’ tweets for specific keywords and phrases.
Anything deemed indicative of a suicide risk is automatically flagged up to the user via email, with a link to the tweet and resources on how to help.
The notifications are completely private, and never seen by a user’s Twitter followers. A user will only see alerts relating to friends they follow, and not those who follow them. The app will not be able to scan direct messages either.
Language that might trigger a red flag includes "tired of being alone", "hate myself", "depressed" and "need someone to talk to".
However, common phrases such as "help me" are also included, and the app may flag up sarcastic or humorous tweets. Users will be able to give feedback on whether an alert was correct.
In an FAQ on its website, Samaritans says the app can recognise specific phrases but "won’t get it right every time". Organisations that might tweet similar language – such as other suicide prevention charities – can also be whitelisted.
Samaritans Radar has been produced by digital agency Jam as part of Twitter’s ‘Ads for Good’ programme and marks a wider collaboration with the charity.
The app is targeted at 18- to 35-year-olds on the basis that they tend to share more social media updates indicative of their emotional well-being.
Samaritans has also worked with Facebook, launching the "Help a friend in need" campaign last month.
Joe Ferns, executive director of policy, research and development at Samaritans, said: "We know that people struggling to cope often go online looking for support.
"By not addressing this issue we run the risk of shutting these discussions down and driving them underground."