Competitor Keyword Research for ASO: A Practical Method

Your boss drops a competitor's listing into the channel and asks why they win that term and you do not. You export their keyword set, add a few, wait two weeks, nothing moves.

I have run that loop often. The lesson from competitor keyword research: this is not about copying answers. A keyword carries no information by itself; what matters is why it sits there, what it earns them, and whether you could hold it.

Two questions follow. Which words do they win that you are missing, and which look big but are a grind for everyone?

The short version

Study what competitors win, not what they write.

Collect candidates from four public surfaces, not just a tool export. Filter each one: does it fit, how strong is the page holding it, how big is it. Fail one and it stays off. Rank by distance from the position, not by volume.

Tools hand you words. Judgment is yours.

Copying words is the wrong goal

A word on a competitor's page tells you they made a choice, not that it worked. They may be living off it, or they typed it once and forgot.

Two gaps are worth writing down.

"They have it, I do not." The word sits in their title and never in yours, for the same feature and the same user. Usually that is something you offer but are invisible for.

"They are grinding." The word looks large until you look closer: the position changes hands often, the page keeps getting rewritten, ratings drift. Crowded, unstable, and the same effort buys less. Mark it parked.

Where the evidence is hiding

Titles and subtitles. The most expensive space on the page, so anything there survived an argument. Pull the history: arrivals and removals are both signals.

The full description on Google Play. A real text surface. Phrases that repeat across a long description show the territory a team wants to own, especially what never reached the title.

Search suggestions. Type a root term into the store search box and read the dropdown. Those come from what people type, not from what a competitor wants you to believe.

Version history. Every release is a chance to change metadata. Log the changes and line them up against the ranking; you will see which word they bet on.

Three filters, in this order

Intent first, because it kills the most words.

Type the term and read the results. If the top is a different category entirely, the intent does not live with you; someone searching it wants another thing. A budgeting app chasing "budget" is the classic case: many searchers want corporate finance software. Right word, wrong intent.

The reverse holds too: an off-category term can return pages of apps like yours, which means users name the problem differently than you do.

Then the strength of whoever holds it. Is the word the main term in their title, or a passing mention in the description? A mention means it is not their front and there is room; at the front, it is a core asset, so expect a fight. Recent ratings and the first screenshots fill in the rest.

Last, rough size. Volume estimates vary wildly between tools, so read them as brackets. The results page is a fair substitute: small apps with modest rating counts suggest a soft term; household names suggest volume.

The counterintuitive part: borrowed authority

You see a competitor ranking well and conclude the term works. So you add it. Often nothing happens, because they were never ranking on that word. The page carries weight, and the word rode along.

Weight comes from how they perform elsewhere: installs, retention, ratings, brand search. None of it transfers when you add a word to a title.

You copy the outcome and skip the cause. A term lifted by a strong page can do nothing on a weak one.

How to tell: look for deliberate work. The word is in the title, the screenshots speak to it, the description hammers it. If it appears once in the opening line and nowhere else, it was carried.

When this method breaks

All of the above assumes the right reference. Pick wrong and it is noise.

The category is too far away. Different problem, overlapping features, different language. Their users do not search like yours.

The competitor runs on brand. Some apps look dominant until you separate brand searches from the rest. That traffic is people typing a name you cannot copy.

The app is one or two releases old. Early metadata is guesswork, and using it as a reference treats a hunch as evidence.

One more, about you: if your page does not convert, no research saves the campaign. Fix the page first, and check the current official guidance before you ship.

From notes to a ranked list

The output is not a keyword dump but a short table: the word, where I found it, fit (high, medium, low), who holds it, and one line on what it maps to.

Then rank by subtraction. Cut low fit. Cut the heavily defended, unless the term is existential. Start with held-but-undefended, the best return on effort. Check that survivors have an honest home in your metadata; if not, watch list rather than rewritten copy.

You will end with a first tier of a handful. That is correct. A list with dozens of top priorities has no priorities.

What to do next

If you want to start now, work in this order:

1. Pick your references. Two or three direct competitors, screened against the failure cases above. No brand-driven apps, no new listings. 2. Collect. Walk all four surfaces: title and subtitle history, the Play description, search suggestions, version changes. Record only, judge later. 3. Filter. Ask in order: does it fit, is it defended, is it big enough. Keep what survives all three and write the one-line mapping. 4. Rank and ship. Put held-but-undefended words first, take the top handful, place them where they honestly belong. Log a baseline, then read the trend.