For years, the standard advice for spotting a scam text was simple: look for the typos. In 2026 that advice is obsolete. Generative AI writes most scam texts now, and it writes them fluently, warmly, and in whatever tone is most likely to get an answer from someone like you.

What AI changed

Three things. Volume: one operator can now run conversations with thousands of targets at once, each reply drafted by a model in seconds. Personalization: leaked marketing data feeds the model your name, age bracket, and shopping habits, so the message references a store you actually visit. Polish: grammar, punctuation, and tone are indistinguishable from a real person or a real company.

What AI scam texts look like

A pharmacy refill reminder for a pharmacy you use. A neighborly wrong-number text that grows into a friendship over weeks. A bank fraud alert that mirrors your bank's exact wording. The tell is no longer in the writing — it is in what the message asks you to do.

The tells that survive

Every scam, however fluent, eventually needs one of three things: a tap on a link, a code or password, or a payment. Treat any unexpected message that asks for one of these as hostile, no matter how human it sounds. Verify through the company's own app or a phone number you already have — never through the message itself.

Filtering is the counterweight

You cannot out-read a machine that A/B-tests its lies, but you can refuse it the audience. Blockr's keyword packs are updated against the AI-generated campaigns in circulation, and content-based rules catch the patterns that repeat beneath the varied wording. Filtered texts land in Junk instead of in front of you.

Calm beats clever

AI made scam texts sound human. It did not change what they are after. Slow down, verify independently, filter aggressively — the machine loses interest in numbers that never answer.