In March 2026, Apple did something it almost never does: it published research describing how its own search algorithm decides what to rank. The paper, Scaling Search Relevance: Augmenting App Store Ranking with LLM-Generated Judgments, names two signals working side by side — textual relevance and behavioral relevance — and treats a behavioral metric like retention as a first-class input, not a tiebreaker (Apple Machine Learning Research, 2026). It was not the only 2026 shake-up: WWDC 2026 put video into search results and made Creative Assets editable live.
That confirmation matters because most iOS teams still optimize the title and keyword field, then stop — and can't explain why a perfectly keyworded app keeps slipping. The reason is that metadata is only one of two engines now — and ranking itself is just one channel inside your wider iOS app marketing. This is the full list of App Store ranking factors in 2026, grouped by how Apple actually weighs them, and ordered by what to fix first.
Key Takeaways
- The App Store ranks on two engines: textual relevance (title, keyword field, subtitle, and newly OCR-indexed screenshot captions) and behavioral relevance (velocity, conversion, ratings, retention).
- Apple's March 2026 LLM research confirms behavioral signals like retention sit beside your keywords — its redownload-to-download ratio hit 2:1 (ASO World, 2026).
- Since mid-2025, Apple OCR-indexes screenshot caption text, so captions can rank you for terms that aren't in your title or keyword field.
- Ratings gate everything: drop below 4.0 stars and tap-through falls sharply. Fix behavioral problems before you scale user acquisition.
How does the App Store algorithm actually work in 2026?
Apple ranks apps with two engines. The first is textual relevance: how well your metadata semantically matches a user's query. The second is behavioral relevance: what users do with your result — taps, downloads, and post-install retention. Apple's March 2026 research fine-tuned a three-billion-parameter language model on human relevance labels, generated millions of new judgments, and retrained App Store ranking on both — and it names those two engines explicitly (Apple Machine Learning Research, 2026).
Why does this framing matter so much? Because it's the master key to every factor below. Textual relevance gets you considered: it decides which searches you can appear for at all. Behavioral relevance decides whether you stay: it sorts you among the apps that are already relevant. Keywords open the door; behavior keeps you in the room. Run only one engine — great metadata with weak retention, or strong retention with thin keywords — and you leave most of your organic growth on the table.
This is high-stakes because search is where iOS discovery happens. In 2026, search drives roughly 65% of all App Store app discovery, far more than browse or referral (Digital Applied, 2026). Win the two engines in search and you've won most of the store.
Share of iOS app discovery that starts in search
How much do the title, subtitle, and keyword field matter?
Metadata is the textual-relevance engine, and on iOS it runs on three controllable fields. The 30-character title is the highest-weight field. The 30-character subtitle is a strong secondary signal. And the 100-character keyword field — invisible to users and unique to Apple — sets which searches you're eligible for at all (AppFollow, 2026). Crucially, Apple does not index your description or your review text for ranking, which is the opposite of Google Play.
The keyword field has its own mechanics: separate terms with commas and no spaces, skip plurals and the word "app," and never spend characters on a competitor's brand name. Because the title carries the most weight, the keyword-versus-brand decision there is the highest-leverage call in your metadata; our guide to how to spend your 30-character title walks through it. Don't repeat words across fields, either — Apple de-duplicates, so a term wasted twice is a keyword you didn't claim.
If you run both stores, this is where they diverge most. Apple reads three short, partly hidden fields; Google reads everything you publish. Our breakdown of how Apple's keyword field differs from Google's full-text indexing covers why one set of metadata can't serve both, and it's also why keyword field and metadata optimization is a platform-specific job rather than a copy-paste.
Does the App Store really index screenshot text now?
Yes — and it's the headline change of the last two years. Starting in mid-2025, Apple's algorithm began OCR-indexing the caption text overlaid on your screenshots, so keyword-rich captions can rank you for terms that appear nowhere in your title or keyword field (AppFollow, 2026). Practitioners first caught it through ranking anomalies across English-speaking markets; it's now treated as a confirmed, live signal.
That changes how you treat your first three frames. On iOS those frames render above the fold, so they already carried most of your conversion weight — now they carry keyword weight too. The same caption has to read cleanly to a human and to Apple's OCR at once. Two related surfaces also rank organically in 2026: Custom Product Pages, of which you can run up to about 70 per app, can be assigned keywords and surface in search; and In-App Events appear in results as well.
That dual job — convert and rank from one caption — is exactly why we fold keyword intent into screenshot and creative testing rather than treating creative and metadata as separate workstreams. Get it right and one screenshot earns its place twice.
Why do ratings and reviews carry so much weight?
Your star rating is a store-wide quality signal applied across every search, and 4.0 stars is the cliff. Drop below it and tap-through and download rates fall sharply (Strataigize, 2026). The upside is just as steep: apps that climbed from 3.6 to 4.2 stars saw roughly 60% higher conversion (Adapty, 2026). Apple also down-weights ratings older than 12 months, so your visible score reflects how the app feels now, not its lifetime average.
Review velocity matters alongside the raw number. A steady stream of recent, positive reviews reads as a freshness signal, while a stale review wall slowly loses weight. And the cost of a slip isn't only organic — when conversion drops, download velocity drops with it, and your paid cost-per-install climbs to hold the same volume.
Relative conversion by star rating (5★ = 1.00)
Because the rating sits at the top of both engines — it suppresses conversion and the velocity that conversion feeds — protecting it is one of the highest-leverage things you can do, which is why we treat ratings and reviews management as a core ASO workstream, not an afterthought. For the full playbook on getting more reviews, replying for a rating lift, and recovering a sagging score, see the complete guide to App Store reviews.
How much do downloads, velocity, and conversion move rankings?
Download velocity and conversion from search are the core behavioral signals, and they're tightly coupled. Apple weighs recent install speed more heavily than lifetime totals, and the conversion rate that produces those installs is itself a relevance signal — a high tap-to-install rate on a query tells Apple you're a good answer for it (AppFollow, 2026).
Creative is what feeds that loop on iOS. Average App Store conversion sits around 25%, and iOS runs about 5.7 percentage points higher than Google Play, largely because the first three screenshots render above the fold (AppTweak, 2026). Screenshot tests on those frames lift conversion 10–25%, the highest-leverage creative test on the platform (Digital Applied, 2026). Better creative raises conversion, conversion drives velocity, and velocity lifts rank — which is also why a paid burst without good creative decays fast. If this is the engine holding your app back, our guide to how to actually increase your app conversion rate walks through every lever in order.
How much does retention and engagement count now?
Retention is the breakout factor of 2026. Apple's redownload-to-download ratio reached 2:1, making the weight on retention empirically visible, and the March 2026 LLM paper treats post-install behavior as a first-class objective rather than a footnote (ASO World, 2026). Session frequency, daily actives, and churn all feed a quality score that decides how high you place among relevant apps.
The baseline makes retention a steep differentiator: the average app loses roughly 70% of its users within the first 30 days (UXCam, 2026). Here's the part teams miss — uninstall spikes from low-quality installs don't just hurt in general; they can suppress your rankings for the specific keywords that drove those installs. Buy untargeted traffic and you can demote yourself on the very terms you paid to win. You can't keyword your way past a retention problem, and on iOS you can't buy your way past one either.
Hold one mental model and the whole list gets simpler: every App Store ranking factor is either a textual-relevance lever (gets you considered) or a behavioral-relevance lever (keeps you ranked). Title, subtitle, keyword field, and screenshot captions are textual. Velocity, conversion, ratings, and retention are behavioral. It's not our framing — it's the two objectives Apple named in its own 2026 research. When a fix doesn't move your rankings, it's almost always because you pulled a textual lever on a behavioral problem, or the reverse.
Do localization, updates, and Apple Search Ads matter?
They do — these are the supporting factors that round out both engines. Localization unlocks a separate keyword pool in each market and lifts in-market conversion, but only if you localize the metadata and the first three screenshots rather than just translating the text. Update cadence signals active maintenance: a steady two-to-four-week rhythm keeps the listing fresh and re-triggers keyword-field reindexing, which takes about two to four weeks to stabilize after each version (AppFollow, 2026).
Apple Search Ads sit in a useful gray area. They don't directly buy organic rank, but the downloads they drive feed the velocity and conversion signals that do — a paid campaign aimed at a target keyword can warm up your organic position on it. The lift compounds when your creative and rating are already strong, and wastes spend when they aren't.
iOS conversion lift by lever
One factor sits outside ranking but feeds it: editorial featuring. A Today-tab placement won't change the algorithm's math, but the velocity and reviews it drives absolutely do — which is why editorial featuring is a separate visibility lever worth pursuing alongside ranking work. Mapping keyword pools and creative market by market is what localization across your priority markets is built for, and structuring paid to compound the organic side is the job of Apple Search Ads.
Which App Store ranking factors should you fix first?
Triage by engine. If a behavioral signal is failing — a rating under 4.0, weak retention, low search conversion — fix that first, because it caps everything else. No amount of keyword work outranks an app the algorithm is quietly demoting for poor behavior, and on iOS you can't buy past it with installs either. Stop the bleeding before you pour in traffic.
Once behavior is stable, sharpen the textual engine: the title, subtitle, keyword field, screenshot captions, and per-market localization. Only then scale user acquisition, because velocity now amplifies a healthy app and wastes spend on a leaky one. This fix-first triage is one spoke of a larger loop — our guide to building an effective ASO strategy shows where ranking work sits in the full framework. The same discipline maps cleanly onto Android — if you also ship there, the Android side of the same algorithm shift shows how Google's two-gate model lines up with Apple's two engines. If you'd rather have someone tell you which engine is holding your app back, that's the first thing a free ASO audit is built to find.
Frequently asked questions
What is the most important App Store ranking factor?
There isn't a single one. Apple blends textual relevance (title, keyword field, subtitle, screenshot captions) with behavioral relevance (velocity, conversion, ratings, retention). Its March 2026 research names both as first-class objectives (Apple ML Research, 2026), and your rating acts as a store-wide quality gate.
Does the App Store index screenshot text?
Yes. Since mid-2025, Apple OCR-indexes the caption text on your screenshots, so keyword-rich captions can rank you for terms that aren't in your title or keyword field (AppFollow, 2026). It surfaced through ranking anomalies across English markets and is now treated as a live signal.
Do reviews affect ranking, or just conversion?
Both, indirectly. Apple doesn't index review text for relevance the way Google Play does, but rating level and review velocity feed a quality signal, and the 4.0 threshold strongly affects conversion and velocity. Apps that climbed from 3.6 to 4.2 stars saw roughly 60% higher conversion (Adapty, 2026).
Does retention affect App Store ranking?
Yes. Post-install behavior — retention, sessions, churn — is a first-class signal in 2026; Apple's redownload-to-download ratio hit 2:1 (ASO World, 2026). Uninstall spikes can demote you for the keywords that drove those installs.
How often should I update my app for ASO?
A steady two-to-four-week cadence signals active maintenance and re-triggers keyword-field reindexing, which takes about two to four weeks to stabilize after each version (AppFollow, 2026). Going dormant risks a slow ranking decay.
The bottom line
App Store ranking in 2026 is a two-engine system, and for once that's not an analyst's metaphor — it's the model Apple described in its own research. Textual relevance gets you considered; behavioral relevance decides whether you stay. The factors that climbed the list — retention, ratings, conversion velocity, screenshot-caption indexing — are mostly behavioral, and they cap what keywords can do. The practical order of operations:
- Fix failing behavioral signals first — a rating under 4.0, weak retention, low search conversion — because they limit everything else.
- Then sharpen the textual engine: title, subtitle, keyword field, keyword-rich screenshot captions, and per-market localization.
- Scale user acquisition last, once velocity will amplify a healthy app instead of feeding a leaky one — and avoid untargeted installs that demote you on your own keywords.
