App Store Keyword Research: A Playbook for Indie Developers (2026)

Keyword research is where indie ASO either becomes a system or dissolves into vibes. The vibes version picks terms that “feel right,” ships them, and re-guesses when nothing happens. The system version treats the App Store itself as the dataset — autocomplete for demand, competitor listings for supply, Apple’s own popularity scores for volume — and produces three defensible keyword bets in an afternoon, for free.

This is the system version. It pairs with the keywords field deep-dive (where the chosen terms go and the rules they obey) and the ASO playbook (the strategy these terms serve).

What you are actually researching

A finished research pass produces a three-tier output, not a list:

• Primary terms (3): the bets. They get the title, the subtitle, and your measurement attention.
• Secondary terms (8–12): the supporting cast that fills the 100-character keywords field with non-duplicating, autocomplete-verified words.
• Combinations (free): multi-word queries Apple assembles by mixing words across your title, subtitle, and keywords field. You never write these — you design the parts so the right wholes exist.

And know which fields feed the index you’re optimizing: on Apple, title + subtitle + keywords field + IAP names — not the description. On Google Play, the title, short description, and the entire long description. The same research feeds both stores; the placement differs, per the cross-store comparison.

The 5-step indie keyword research workflow

Step 1 — define the app in one plain sentence

“A budget tracker for couples who split expenses.” “An interval timer you can read mid-workout.” No brand, no adjectives. If the sentence won’t come, the positioning isn’t ready and research would just encode the confusion.

Step 2 — brainstorm 20 candidates from it

Category words, audience words, problem words, verb words. For the couples’ budget tracker: budget, expense, split, shared, couple, money, bills, settle, track, spending, household, allowance, savings, partner, finance… No filtering yet — filtering with no data is just taste.

Step 3 — verify demand in autocomplete

Type each candidate into App Store search, letter by letter, and record what Apple suggests. Suggestions are ranked by real query volume; a candidate that never appears at any prefix is a word nobody types, however right it feels. This step usually kills a third of the brainstorm — that’s it working.

Step 4 — map supply with a competitor frequency pass

Open the top 30 apps in your category and tally which surviving candidates appear in their titles and subtitles. A term in five-plus competitors’ premium fields is owned; a term in zero to three is contestable. While you’re there, eyeball difficulty: review counts and update recency of whoever ranks top-ten for each term — six-figure review walls mean move on.

Step 5 — commit to three

From the survivors, pick the three with the best intent-to-difficulty ratio that your screenshots can honestly sell. Slot one into the title, one into the subtitle, spread the rest of the survivors through the keywords field (the character counter keeps the budget honest, the subtitle helper shapes the subtitle), and write the other seventeen down for next quarter.

The three-keyword rule

Three terms, ninety days, top-ten or re-research. The rule exists because of two indie constraints. Attention: you can genuinely iterate on three terms inside monthly release cycles, and you cannot on thirty. Falsifiability: three focused bets produce a readable verdict — if none of the three cracks the top ten in a quarter, the research was wrong, and you want to know that cleanly rather than have it smeared across thirty excuses.

Each of the three must pass:

1. Intent: someone typing this is looking for what your app does — not browsing a category.
2. Winnability: the current top ten for the term contains apps you can realistically displace.
3. Provability:your first screenshot and subtitle can back the term’s promise without stretching — both for conversion and for guideline 2.3.7’s sake (the rejection index shows what stretching costs).

The workflow, worked end to end

The couples’ budget tracker, all five steps with real outputs:

• Brainstorm (20): budget, expense, split, shared, couple, bills, settle, spending, household, partner, money, allowance, savings, iou, roommate, fair, tally, divide, joint, balance.
• Autocomplete pass: kills tally, fair, divide, allowance (no suggestions at any prefix); promotes the modifier finds split bills, expense splitter, and shared expenses from the space trick.
• Competitor map: budget and expense tracker appear in 8+ of the top 30 — owned. Expense splitter appears in two, shared expenses in one, roommate in zero despite live autocomplete demand.
• The three bets: expense splitter (title), shared expenses (subtitle), and roommate (keywords field, first slot) — each autocomplete-verified, contestable, and honestly provable by a first screenshot of the split flow.
• The shipped field (89 characters, no duplicates with title or subtitle): roommate,split,bill,settle,iou,household,partner,joint,couple,money,trip,group

Total elapsed time: one focused afternoon. The same pass without the verification steps produces a field full of budget-family terms — plausible, owned, and inert.

Free research sources that actually work

1. App Store autocomplete. The demand signal itself, ranked by real volume. The next section is the full technique.
2. Category Top Charts. The supply map — what the winning listings target, free to read, current by definition.
3. Apple Ads keyword suggestions and popularity. Inside an Apple Ads Advanced campaign draft you can see suggested keywords and the 5–100 popularity score — Apple’s only first-party volume number — without funding or launching anything. (Apple Ads Basic, the automated tier, won’t show you this; the draft trick needs Advanced.)
4. Google Play autocomplete.The same technique on the other store — essential if you ship Android, since Play’s query patterns differ.
5. Communities where your users complain. Reddit threads, Discord servers, App Store reviews of competitors — the natural-language phrasings there become long-tail terms no tool surfaces.
6. Your own search-terms data, later. Once you run any Apple Ads at all, the search-terms report shows the actual queries that found you — the highest-grade keyword data that exists, as a side effect.

Paid trackers (AppTweak, App Radar, ASO.dev and friends) add daily rank history and cross-market dashboards. Worth it when your keyword set outgrows a spreadsheet; redundant before then — the comparison pages map the boundary.

Autocomplete mining, properly

1. Use the App Store app on a phone — the mobile surface is where your users search.
2. Type letter by letterand record suggestions at each prefix: “bud”, “budg”, “budge” each reveal different variants, and position is your volume proxy.
3. Mine modifiers with the space trick:type your seed plus a space plus each letter of the alphabet — “budget a”, “budget b”, “budget c” — and harvest the modifier suggestions (“budget app for couples”, “budget book”, “budget calendar”). This is where long-tail intent hides.
4. Log everything in a sheet with prefix, suggestion, and position — memory flattens exactly the rank differences that matter.
5. Repeat per storefront.Each country’s autocomplete is its own dataset; Japanese suggestions owe nothing to English ones, which is the entire argument of the localization guide.

Accelerating the workflow with AI

AI belongs at two points: widening step 2 and assembling step 5. Feed a model your one-sentence definition and it returns synonyms, adjacent jobs, and audience phrasings that triple your candidate pool — including the non-obvious ones manual brainstorms miss. Then, once verified terms exist, AI assembles them into a compliant field: Push My App’s AI metadata generator writes the 100-character string with no title/subtitle duplicates, no trademark bait, and per-locale regeneration across 14 languages.

What AI cannot do is step 3. No model knows current App Store query volume; it produces plausible terms, and plausible is precisely the failure mode — a keywords field full of words that read perfectly and rank for nothing. The pipeline that works: AI to generate, autocomplete to verify, you to choose.

Pitfalls that waste research time

Bidding on category giants. Photo editor and todo listare owned by apps with decade-deep review moats. Volume you can’t rank for is decoration.

Confusing browse intent with install intent. Broad terms attract comparison shoppers; specific terms attract installers. The follow-up-query test: if users who type it usually refine it, it’s a browse term.

Skipping verification because the term is “obviously” right. The store will tell you for free whether anyone types it. Let it.

Re-researching instead of iterating.Burning each quarter’s cycle on a fresh keyword set means no bet ever gets the ninety days it needs to resolve. Hold the three, read the data, swap the proven losers only.

Translating instead of re-mining for locales. Every storefront’s autocomplete is a separate demand dataset. The research transfers as a method, never as a word list — and before any of it ships, the 9-minute pre-submission checklist catches what the excitement missed.

Frequently asked questions