The claim that word games improve vocabulary gets thrown around a lot, usually without much backing. It is worth looking at more carefully, because it turns out to be mostly true, with some important caveats about which games help and how.
What vocabulary research actually shows
Vocabulary is built through encounters with words in context. The more times you encounter a word used correctly and in varied contexts, the more firmly it is stored and the more readily you can retrieve it. Reading is the most efficient way to encounter a large number of words in context. Word games are a different kind of encounter, but still a useful one.
The key difference is that word games tend to involve active retrieval (you have to produce or recognise the word yourself) rather than passive exposure (you read it in a sentence). Active retrieval is generally better for memory consolidation. This is why testing yourself on vocabulary improves retention more than re-reading the definitions does.
Which games help most
Not all word games improve vocabulary equally. Here is what the research and practical experience suggests about specific types:
Word ladders are particularly good for cementing knowledge of common words because they force you to think about words at a letter level. You do not just need to know what BARE means. You need to know it is related to CARE, DARE, FARE, HARE, MARE, WARE. You build a network of words rather than a list.
Crosswords are good for recognising vocabulary, especially less common words, but rely heavily on knowledge of crossword conventions. The clue style is its own skill separate from general vocabulary. Still useful, but the learning transfers less directly.
Anagram games improve spelling and letter pattern recognition. Rearranging letters to find words requires you to hold a word's structure in mind, which reinforces the spelling. Research on this is limited but results suggest modest benefits for spelling accuracy.
Word search puzzles, despite being the most commonly used word game in educational settings, have the weakest evidence for vocabulary improvement. You are recognising pre-given words in a grid, not producing them or thinking about their meaning. They may help with spelling recognition but probably not much else.
The production effect
One reason word ladders and crosswords help more than word searches is something called the production effect. Words you say, write, or actively think about are better retained than words you simply see. In a word ladder, you are actively trying to produce new words from existing ones. Each step requires you to generate, not just recognise. That production makes the words stickier.
Consistent play matters more than session length
The vocabulary gains from word games accumulate over time rather than from single sessions. Someone who plays a daily word ladder for three months will notice a measurable improvement in word recall. Someone who plays twenty word ladders in one weekend will notice nothing lasting.
This is consistent with what we know about memory in general. Distributed practice, returning to material repeatedly across time, produces more durable retention than massed practice, doing a lot at once. The daily puzzle format happens to be well-designed for this even if that is not why puzzle sites chose it.
Spelling specifically
Spelling improvement from word games is more modest than vocabulary improvement. Most word games do not explicitly correct your spelling, which means a misspelling can go unnoticed. Word games where spelling errors prevent progress, like word ladders where you need to input the correct spelling to advance, are better for spelling than games where you select letters from a set.
If spelling is your specific goal, games where you type words from memory will help more than games where you rearrange given tiles.