They don't, but the spammers have become more sophisticated over time. They use Cyrillic letters that look like Latin letters, they hack old unused accounts (sometimes verified ones), they post the spam as a second-level answer, they use images instead of tweeting text, they have started adding noise and various transforms to the images to make them harder to automatically classify as spam, and they probably have many more tricks up their sleeves. Fighting against spam is hard.

That's a good question - there was a time, not too long ago where every single tweet of Musk was spammed with BTC giveaways.

Shocking that they don't take even basic precautions like image hashing to cut down on this.