"What are the best keywords for my website?" has a boring true answer: the ones people actually type, that you can rank for, placed where they count. Here is the process, and what changes for an app listing.
Not what you call the product. Expand from a few seed phrases into every wording people use: suggestions, related queries, and the keywords your competitors rank for. Then check real clicks, not only volume — Google Ads groups similar wordings into one number.
| Type | Example | Where it goes |
|---|---|---|
| Exactly this kind of product | "youtube video translator", "cbt app" | Title, subtitle, first lines of the description, homepage H1 |
| The category | "video translator", "mental health app" | Description, site pages — only in the title if you can win it |
| Informational ("how to…") | "how to translate a youtube video" | An article on your site with an install/signup button — never the listing |
| Someone else's brand | "headspace", "google translate" | Nowhere in your title; a comparison page at most |
For each candidate look at who is in Google's top 10 and, for apps, how strong the apps ranking in the store are. Demand you cannot capture is not demand. Keep the keywords where the top is weak, mark the rest as "later".
Founders usually have a phrase they believe in. Check it literally before building pages around it: it may have real demand, or zero — and then the question is what people type instead.
The App Store indexes the title, subtitle and a hidden 100-character keyword field — not the description; Google Play indexes the description too. So the same keyword list is placed differently: single words without repeats in the keyword field, phrases in the title and subtitle, full sentences in the Play description. Localized listings use the words people type in each country, not a translation.
One cluster per page: a main keyword plus its wordings. Two unrelated keywords on one page compete with each other; the report groups wordings so you can see which belong together.
Pull the queries the site ranks for in Google (a keyword tool or the Riwords report does this) — that list, with search volume and real clicks, tells you what works for them. Reading their page source for meta keywords tells you nothing.
Pick the ones with a weak top 10: forums, small sites, no big brands. Head terms held by media and Wikipedia are a year away. The report marks difficulty for each keyword with the top-10 composition.
Single words that are not already in the title or subtitle, no plurals of included words, no competitor brands, comma-separated without spaces, up to 100 characters. Apple combines words across the three fields into phrases, so the field is built from words with demand, not phrases.
No — localize. People in Germany or Brazil search with their own wordings, and an English keyword is often searched there too. The report finds the local keywords with demand per country you choose.