Users claim Twitter is 'ghost-deleting' potentially offensive tweets

The site, however, says that the update was mistakenly marked as spam

The update in question was designed to 'make fun of journalists freaking out over Milo Yiannopoulos’ appearance on Bill Maher' (Getty)

Twitter has denied reports claiming it has started “ghost-deleting” tweets as part of its crackdown on offensive content.

The company recently announced a new set of safety features designed to protect users from online abuse, including something called ‘safe search’.

This hides tweets containing “potentially sensitive content” and tweets from blocked and muted accounts from users’ search results, though Twitter said that the content will continue to exist.

However, according to Twitter user @SuperNerdCow, the site went a step further, hiding one of his tweets from all users, even when they enter the update’s URL.

He claims that he can still access the tweet, but it appears as a broken link for others.

Twitter has, however, told The Independent that the tweet was not actually deleted but mistakenly marked as spam.

HeatStreet reports that the update in question was designed to “make fun of journalists freaking out over Milo Yiannopoulos’ appearance on Bill Maher”.

It was a captioned meme, reading, “This about sums up the reaction to Bill Maher talking with Milo Yiannopoulos. Discourse is #NotTheEnemy.”

The image itself portrayed doctored Mr Maher and Mr Yiannopoulos standing near a crouching cartoon figure wearing a rainbow-coloured shirt and holding a MacBook, with the label “*Autistic Screeching*” floating above their head.

The term is often used to ridicule so-called ‘social justice warriors’ online.

@SuperNerdCow said that Twitter didn’t delete the tweet when he posted it without the hashtag, leading him to believe that the site had been monitoring the #NotTheEnemy hashtag.

Twitter has also started placing temporary usage restrictions on accounts it detects “potentially abusive behaviour” from, with the site explaining that it will consider an account's behaviour, rather than the language used in tweets, to determine whether or not they warrant a ban.