How screen readers actually announce your images

Most marketers writing alt text have never actually heard one read aloud. That gap matters, because the moment you hear it, the reason alt text quality matters stops being abstract. Here's what a screen reader says in the three situations you'll actually run into.

When there's no alt text at all

Screen readers don't stay silent on a missing description — they announce that something is there and give up. NVDA typically says "graphic" with nothing else. VoiceOver often says "image," or worse, reads the filename character by character: "I-M-G underscore four seven three two dot J-P-G." Either way, the listener knows a picture exists and gets zero information about what it shows.

When alt text is a keyword dump

Stuffing alt text with search terms doesn't get skipped or filtered — it gets read exactly as written, word for word, comma for comma. A screen reader has no way to tell "marketing copy trying to rank" from "an actual description," so it reads the whole thing as one flat, run-on sentence.

Read both of those out loud yourself. The first is noise. The second takes about two seconds and actually tells you something.

When alt text is done well

Good alt text gets announced, then the reader moves on — no repetition, no filler words like "image of" (most screen readers already announce that an image is present before reading the description, so saying it again is redundant), no stumbling over hashtags or emoji. It sounds like a short, clear sentence because that's exactly what it is.

The takeaway isn't complicated: whatever you type into that field gets read exactly as written, to someone who can't cross-check it against the actual photo. That's a higher bar than most other copy on your page has to clear — and a good reason to actually listen to one of your own posts with a screen reader at least once.