The three solve the same problem in different ways. Choosing well is, above all, knowing which of the three paths is yours.
Polylang: duplicate content
Polylang does not translate: organizes. Each entry has its twin in another language and you link them. It's solid, its base version is free, and for those who write each language separately —because the message changes according to the market— it's exactly what they need.
The limit: someone has to write that content. If you have three hundred articles and expect them to appear in English, Polylang alone won't do it; you need third-party add-ons for automatic translation.
WPML: the tailor-made suit
WPML is the veteran and covers almost any scenario: multisite, translation commissioned to professional agencies, approval workflows, compatibility with practically the entire ecosystem.
The price of that it's a dense initial setup and a real learning curve. It's the right choice for large projects with someone dedicated to maintaining them, and probably too much tool for a twenty-page website.
WP-Translate: translate and review
The approach is the opposite of WPML's: that the site is translated today and polished later. You launch automatic translation of everything, the plugin marks what's translated as pending review, and you correct on the website itself what is seen and what sells.
The limit: if you need each language to have distinct content —not translated, but written separately—, Polylang's approach fits better.
Quick decision table
- Each language with its own content, handwritten → Polylang.
- Large project, dedicated team, external professional translation → WPML.
- Same content in several languages, as soon as possible and without a learning curve → WP-Translate.
About the cost
Polylang has a free version; its Pro version and WPML are annual subscriptions. WP-Translate is a one-time payment with a lifetime license. If the project is going to last for years, that difference accumulates; if it's a one-off website, it weighs less than the convenience of each tool.
None of the three is a bad choice. What does turn out expensive is choosing by price and discovering two months later that the workflow wasn't yours.