App Store Keyword Research 2026: The Scores Broke. Do This
The short answer
App Store keyword research in 2026: a method that still works
Since late September 2025, most keywords that used to score above 5 in the popularity number most tools show now sit at 5, so treat a 5 as unknown, not as no searches. Every keyword has to pass three tests: people search it, your app answers it, and you can win it. Build a long list, score relevance first, place words by field, then test before you commit.
Key facts
- About 3 in 4US keywords that scored above 5 and fell to the floor of 5 from 29 September 2025, by one tracker's count (about 76% fewer).aso.dev, Apple Ads popularity massive drop (updated 22 Aug 2026)
- 1 to 100Apple's newer popularity score for the most popular search terms inside a genre, from the Apple Ads Platform API.Apple Ads Platform API, Insights endpoints
- 100 bytesSize of the keyword field, with commas between words and no spaces. Name and subtitle get 30 characters each.App Store Connect Help, Platform version information
Chapters
- 0:00App Store keyword research in 2026: what broke?
- 0:35What does a keyword popularity score measure?
- 1:43Where is Apple's popularity data now?
- 2:35Where do keyword ideas come from?
- 3:28How do you score a keyword?
- 4:26Which keyword should you drop?
- 5:04Where does each keyword go?
- 6:05How do you know you picked the right keywords?
- 6:56What's on the final short list?
Transcript
Edited for readingCoral is a fictional document scanner app; its keywords, scores and test results are illustrative. Every fact comes from a real source, listed under Sources.
App Store keyword research in 2026: what broke?Watch from 0:00
If you're doing App Store keyword research this year, the popularity score most tools show has stopped working. One tracker found most keywords that used to score above 5 now sit at 5. That's the bottom of the old 5 to 100 scale.
So here's the method that still works. Every keyword has to pass three tests. People search it, your app answers it, and you can win it.
We'll run it on Coral, a made-up scanner app. By the end you'll see its short list, and why it throws away its most popular keyword. Every fact is sourced, as of 10 October.
What does a keyword popularity score measure?Watch from 0:35
First, what that score actually measures. A lot of developers think it's how many advertisers bid on a word. Apple's definition says something else. Apple says a keyword's popularity is based on App Store searches.
And it's relative. A rank against other keywords, not a count of searches. Inside Apple Ads you see it as a number from 1 to 5. Five is the most popular. Most tools also show an older Apple Ads number, from 5 to 100. There, 5 is the floor.
Then, at the end of September 2025, those scores collapsed. One tracker counted the US keywords that used to score above 5. About three out of four fell to the floor.
One tool stopped updating those scores. Another now builds its own estimate instead. A year on, we found nobody reporting that the old scores came back. And Apple hasn't said anything publicly.
Our rule is simple: a 5 means unknown. It doesn't mean nobody searches. So don't delete a word because a tool says 5. Check it another way. That's what the four steps below do.
Where is Apple's popularity data now?Watch from 1:43
So where do you get a real number now? Apple has two places, and neither is App Store Connect.
The first is Apple's new Ads API, from August 2026. It scores the most popular search terms from 1 to 100 inside a genre. A second score covers the whole country, and there's a rank.
Our take: use the genre score inside your genre. Use the country score to compare across genres.
The second is a monthly ranking of search terms, in Apple Ads Insights, still in beta. Both show you the popular terms. Trackers say the monthly one only lists the more popular terms.
And the long tail is where it gets interesting. Apple's own researchers found their method improved ranking most on rare, specific searches.
Our take: Apple's data tells you which big words matter. The specific searches, the ones a new app can win, you find yourself.
Where do keyword ideas come from?Watch from 2:35
Step one is a long list. Here's where the words come from. A common mistake is starting from the words the team likes. Searchers use their own.
Start with what your app does, and the problem each feature solves. Coral scans, crops, reads text and signs.
Then read your reviews. People search in plain, everyday words, and reviews are full of them.
Read competitors' listings for words, never names. Other apps' names aren't allowed in your listing.
Type each idea into App Store search and note what it suggests. Apple calls those search hints. If you run Apple Ads, its search terms report shows the words people actually typed before finding you.
Coral's team started with 64 seed words. That grew to 180 terms. Your checkpoint: if your list is short, check which source you skipped.
How do you score a keyword?Watch from 3:28
Step two: three tests, in this order.
Relevance comes first. Would someone who typed this be happy to find your app? Score it from 1 to 5. Why first? Apple's researchers used AI relevance judgements, modelled on expert labels, to improve App Store ranking. And for ads, Apple says an app that isn't relevant to a search won't show, whatever you bid. So keep only fours and fives.
Popularity is second. Use Apple's genre score when you have it, and mark any tool's number as an estimate.
Difficulty is third. Tools score it, but check the top ten results by hand. Our rule: if every one has over 100,000 ratings and a famous name, skip it for now.
Then a rough order: popularity times relevance, divided by difficulty.
Coral's ten finalists look like this. Which popular one fails the first test?
Which keyword should you drop?Watch from 4:26
It's pdf editor. The most popular term on the whole list. Coral can scan a PDF, but it can't edit one. Relevance two.
Ranking for it would bring people looking for something Coral doesn't do. And relevance is exactly what Apple's researchers trained their labels to judge. So it goes, along with id card scanner and fax app.
The winner on Coral's list is receipt scanner for taxes. Tiny popularity, almost no competition, and relevance five.
In our earlier video, App Store Optimization 2026: How Search Ranking Works, Coral reached 41st for scan receipts. This list is how it picks the next words.
Where does each keyword go?Watch from 5:04
Step three is placement. Three fields, with different rules. Your name and subtitle get 30 characters each. The keyword field holds 100 bytes, with commas and no spaces.
Practitioners agree the name counts most, then the subtitle, then the keyword field. Apple has never published weights. So the best phrase you can realistically reach goes in the name. The next intent goes in the subtitle. The keyword field completes the rest.
Don't repeat words already in your name, subtitle or category. Apple counts plurals as repeats too.
Our take: words combine across those fields. So receipts in the subtitle and business in the keyword field can cover business receipts.
For Coral, the name and subtitle stay. The keyword field drops converter, print and photo, and adds signature and business. That's 95 bytes, with room for one short word.
These words ship with a new version of your app. So plan them with a release.
How do you know you picked the right keywords?Watch from 6:05
Step four: check before you commit, and measure after. If your app is already live, run an exact-match Apple Ads test on your short list.
Our rule: at least two weeks, and keep going until it reaches the impression and tap counts shown in the video. Impressions tell you people search it. Taps and downloads tell you they want your app.
No impressions at a decent bid? Either almost nobody searches it, or Apple doesn't see your app as relevant. Try a closer word.
Coral ran that test on two terms. Receipt scanner for taxes got steady impressions and taps. Business receipts got impressions, but very few. People rarely search it. So business stays a cheap word in the keyword field. It never goes near the subtitle.
After you ship, read your ranks at day 7, 14 and 28. Compare them with keywords you didn't touch.
What's on the final short list?Watch from 6:56
So here's Coral's short list. Seven keywords kept, three dropped.
pdf scanner holds the name. Document scanner and scan receipts come from the name and subtitle together. The long tail fills the keyword field.
And the most popular word Coral found isn't on it at all. Popularity told Coral where people search. Relevance told it where it belongs.
What this can't tell you: Coral's numbers are made up. And Apple's popularity scores are relative. They never count searches.
Sources
- Apple Ads Help, Reporting options and definitions (popularity "based on App Store searches")
- Apple Ads Help, Glossary (search popularity, a relative indicator)
- Apple Ads Help, Use recommendations (popularity 1 to 5)
- Apple Ads, Keyword best practices
- Apple Ads Platform API, Insights endpoints (popularity scores 1 to 100, rank, 1 to 5 tier)
- Adapty, Apple Ads Platform API (release date) (18 August 2026)
- aso.dev, Search term rank report in Apple Ads Insights (10 July 2026)
- aso.dev, Apple Ads popularity massive drop (US keywords above 5: 165,875 to 39,254 from 29 Sep 2025; updated 22 Aug 2026)
- MobileAction, Statement on the Apple Ads popularity metric issue (17 October 2025)
- AppTweak Help, AppTweak volume estimate (27 August 2026)
- RespectASO, Apple Search Ads popularity and ASO keyword data (25 April 2026)
- Apple Developer, App Store search
- App Store Connect Help, Platform version information (keywords, 100 bytes; other apps' names not allowed)
- App Store Connect Help, App information (name and subtitle)
- Apple Ads Help, Understand dashboards and settings (search terms report)
- Apple Ads Help, Search results ad placement (relevance before bid)
- Apple researchers, Scaling Search Relevance (arXiv 2602.23234, v5) (7 June 2026)
- App Radar on field weights (practitioner view; Apple publishes no weights)
Caveat: Popularity scores are relative ranks, never search counts; Apple's search term popularity data covers the most popular terms only; the 76% figure is one tracker's count of US keywords that scored above 5; no tracker we found reports that the old scores recovered, and Apple has made no public statement we could find. Words combining across fields, the test thresholds and the day 7, 14 and 28 reads are our own rules. Coral's numbers are made up.

