How to pick keywords for your app

Every character of your title, subtitle and keyword field is a bet on a search phrase. Picking keywords is deciding which bets to place, and most apps lose not because they optimize badly, but because they bet on searches they were never going to win. This article is the decision framework: how to judge whether a keyword is worth your metadata.

Start from real searches, not a brainstormed list

The natural first move is to sit down and list phrases you think describe your app. Resist it, or at least treat that list as raw input only. Brainstormed keywords fail in a predictable way: most phrases that sound right to you get close to zero real searches, while the phrases people actually type are shorter, blunter and often not the words you would choose.

So start from evidence. Run the keywords through research and let the numbers filter the list. Look at what your competitors rank for, what autocomplete suggests, and what the popularity score says about each candidate. A keyword with a popularity of 5 is not a keyword, it is a hope.

Demand vs difficulty: the core tradeoff

Two numbers decide most keyword choices:

• Popularity tells you how many people search the phrase. No demand, no downloads, no matter how well you rank.
• Difficulty tells you how strong the apps already ranking are. High demand is worthless if position 1 through 10 belong to apps with a million ratings each.

The mistake beginners make is sorting by popularity and targeting the top. The keyword worth targeting is the one where both numbers work in your favor: a popularity 30 keyword you can realistically reach the top 5 for beats a popularity 70 keyword where you will sit at position 80 forever. Position 80 of a huge search is zero installs. Top 5 of a modest search is real, compounding traffic.

ASO Atlas folds this tradeoff into a plain-language verdict on every keyword: Strong opportunity (high demand, beatable competition), Worth targeting, Competitive, Hard to rank, Low demand, or Not enough data. The verdict is the tradeoff pre-computed, so you can scan a list of fifty candidates in seconds instead of weighing two numbers per row.

Relevance: the third axis

Demand and difficulty are measurable. The third test is one only you can apply: would someone typing this phrase actually want your app?

Ranking for an irrelevant term converts to nothing. Worse than nothing: the store notices that your impressions for a search do not turn into installs, and that hurts you. A meditation app that squeezes into results for "sleep games" wins a vanity rank and loses the algorithm's trust. Before any keyword goes into your metadata, ask what the searcher expects to find, and be honest about whether your screenshots will look like the answer.

The long tail is where new apps win

Head terms ("photo editor", "budget") have the demand and the entrenched competition to match. Multi-word phrases ("photo editor for dogs", "budget app for couples") have modest demand, far lower difficulty, and a bonus that is easy to miss: they convert better, because the search is specific and your app is specifically the answer.

A new app's realistic path is to win a set of long-tail phrases first, build download velocity and ratings from them, and use that strength to climb the broader terms later. Best opportunities surfaces exactly these keywords for your tracked apps: real demand, low difficulty, within reach of your current position.

Balance the portfolio

Treat your keyword set like a portfolio, not a wishlist:

• A few ambitious head terms. Put one or two in your title and subtitle where relevance weight is highest. You will not rank top 10 soon; you are planting a flag to grow into.
• A base of winnable long-tail keywords. These carry your actual downloads for the first year. Most of your metadata characters belong here.

All long-tail and you cap your ceiling. All head terms and you get nothing while you wait. The mix is the strategy.

When to give up on a keyword

A bet that is not paying should be replaced. Cut a keyword when:

• It has shown no rank (or a rank past 50) for two or three months despite being in your metadata.
• Its difficulty sits high and the ranking apps are established brands you cannot out-download.
• It ranks, but drives no installs, which usually means the relevance test failed.

Freed characters go to the next candidate from your research backlog. Keywords are not a marriage.

Iterate monthly

Demand shifts with seasons, competitors update their metadata, and your own strength grows with every rating. A keyword that was Hard to rank at launch can become Worth targeting six months in. Re-run the loop monthly: check what moved, prune the dead bets, promote the climbers, and pull fresh candidates from research. Picking keywords is not a launch task, it is the ongoing core of ASO.