Four vendors publish Apple Ads benchmarks with 2026 on the page. Their headline cost-per-tap figures run from $0.92 to $2.25. None of them is wrong.

That range is not measurement error, and averaging it would be the worst thing you could do with it. The four datasets measure different populations over different periods with different denominators, and report different statistics. Once you know which is which, the disagreement becomes the most useful thing on the page: it tells you the width of the band you operate inside, which no single number can carry.

There is also a fact that applies to all of them equally and that none of them mentions prominently: every published Apple Ads benchmark describes a one-ad-per-query auction. Multi-slot search results began rolling out on 3 March 2026 and completed globally by the end of that month. Not one public dataset covers the period after it.

→ WHAT’S IN THIS ANALYSIS
  1. 01The four datasets, side by side
  2. 02Six reasons the numbers do not line up
  3. 03Denominators: download, install, or acquisition
  4. 04Category spread swamps every global average
  5. 05The stale-benchmark problem
  6. 06Why 2025 tap-through rates describe a dead auction
  7. 07How to actually use a benchmark
  8. 08What they all agree on
  9. 09What no public dataset covers

01The four datasets, side by side

Here is what each vendor publishes, methodology attached. Read the methodology column first; the numbers are meaningless without it.

SourceSample & vintageStatisticHeadline figures
Adapty
pub. 10 Jun 2026, mod. 13 Jul 2026		 1M+ ad groups, 90 countries, 59 niches, trailing year. Subscription apps. Medians and spend-weighted blends US: CPT $1.58, CPA $2.51, TTR 9.0%, CR 62.71%. Global median country CPA $0.51; blended CPA $1.34.
AppTweak
pub. 2 Jul 2025, updated 22 Jul 2026		 ~3,500 apps, 50,000 campaigns, $1B spend, calendar 2025. 38 countries, 14 categories, all four placements. Medians Global search results: CPT $0.92, CPI $1.80, CR 56%, TTR 7.4%. US: CPT $1.91, CPI $4.06, CR 55%, TTR 6.6%. Today tab CPM $3.53.
SplitMetrics
last updated 13 Aug 2026		 Its own platform campaign data aggregated across 2025. Global, top 15 categories and markets. Averages Global: TTR 9.7%, CR 66.2%, CPT $2.25, CPA $3.76. North America highest-cost; LatAm and AMEI most efficient.
Business of Apps
page updated 16 Jun 2026		 Republished SplitMetrics figures from first-half 2022. Averages CPT $1.56, CPA $2.40, TTR 9.99%, CR 64.96%. Also carries a 2021 crypto-boom CPT of $11.73.

A fifth reference point matters because it is the one most likely to reach a client deck. Singular’s 2026 ROI Index, published 19 April 2026, was the first edition in which Apple Ads entered the top tier of scaled ad partners alongside AppLovin, Google Ads, Meta, TikTok and Unity Ads. Its Apple Ads figures are SplitMetrics’ — the same 9.7% TTR, 66.2% CVR, $2.25 CPT, $3.76 CPA. Cite Singular and SplitMetrics as corroborating sources and you have cited one source twice.

→ THE FIRST TRAP

Corroboration that is actually circulation

Mobile marketing benchmarks propagate. A vendor publishes a figure, an index republishes it, an aggregator republishes the index, and an agency blog cites all three as independent confirmation. Before you treat two numbers as agreeing, trace both to the panel that generated them. In this set, Singular and SplitMetrics collapse into one, and Business of Apps is SplitMetrics four years earlier.

02Six reasons the numbers do not line up

They compound, which is why the spread is as wide as it is.

One: mean versus median. SplitMetrics reports averages; AppTweak reports medians. Advertising cost distributions are heavily right-skewed, so a small number of very expensive keywords drag a mean well above the middle. On the same data the mean sits above the median essentially always, and a good part of the distance between SplitMetrics’ $2.25 and AppTweak’s $0.92 is this alone.

Two: global versus US. AppTweak publishes both, which sizes the effect cleanly: median CPT of $0.92 globally against $1.91 in the US — a 2.1× geography multiplier, derived from AppTweak’s own two figures, larger than most vendor-to-vendor gaps people agonise over. SplitMetrics calls North America its highest-cost region and Latin America and AMEI its cheapest, so a “global average” weighted toward North American spend and a “global median” across 38 countries are not comparable quantities.

Three: panel composition. Adapty’s dataset is drawn from subscription apps — a monetisation-selected population that is not a random sample of the App Store and bids differently. AppTweak’s panel is its ASO and UA tool customers; SplitMetrics’ is its own platform’s advertisers. Each is internally consistent and none is a census.

Four: placement mix. AppTweak is the only source here that separates the four placements — Search Results, Search Tab, Today tab, Product Pages — and reports search results distinctly. Today tab is bought on CPM, at a reported median of $3.53. Blending CPM inventory into a cost-per-tap average produces a number describing no purchasable thing, and a source that hides its placement mix gives you no way to check.

Five: denominators. Covered in full in the next section, because it is the one people get wrong most often.

Six: time window. AppTweak and SplitMetrics both use calendar 2025. Adapty uses a rolling year to roughly mid-2026. Business of Apps uses first-half 2022. That is a four-year spread on pages all carrying 2026 dates — and between the earliest and latest, the platform was renamed, custom product pages doubled from 35 to 70, Maximize Conversions reached general availability, and the auction gained a second slot.

03Denominators: download, install, or acquisition

Adapty is unusually explicit, and the industry would be better off copying it. It defines its CPA as cost per download — the moment a user taps Get and the app downloads — and notes that roughly 10% of downloads never become installs because the app is never opened. AppTweak reports CPI, cost per install.

So Adapty’s US figure of $2.51 and AppTweak’s US figure of $4.06 are not two estimates of the same quantity. Convert the first at Adapty’s own 10% rate and you get approximately $2.79 per install — a figure derived here, not published by either vendor. That closes some of the gap and leaves most of it open, which is the honest finding: the denominator difference is real, it is worth about 11%, and it explains a minority of a 62% discrepancy. The rest is panel, statistic and geography.

Two vendors reporting "cost per acquisition" for the same market in the same year can be measuring events that happen at different moments in the user’s life. Read the definition before you read the number.

Three consequences. If your stack counts installs at first open — most do, because that is when the SDK fires — a cost-per-download benchmark will always look cheaper than your reality and you will conclude you are overpaying when you are not. Apple’s own AdServices attributes on app open, the install event, not the download. And a network cost may be a SKAdNetwork-attributed install, a third thing again with its own delay and privacy thresholds.

04Category spread swamps every global average

This is the reason a headline benchmark is close to useless for setting a bid. AppTweak’s US search-results breakdown, all medians:

Category (US)CPTCPICR
Sports$12.60$26.8152%
Games$4.21$12.2857%
Finance$3.55$8.4446%
Shopping$2.05$6.2041%
Photo & Video$1.69$3.1363%
Music$1.22$2.1161%

Sports against Music is a 10.3× spread in cost per tap — derived from the two rows above — inside a single country, a single placement and a single dataset. The US median of $1.91 sits below every category in the top half of that table. It is not the centre of anything an actual advertiser experiences.

Adapty measures the spread differently and finds it wider still: cost per download ranges roughly 30× across niches, from Screen Recorder apps at $0.34 to Exam Prep at $10.34. SplitMetrics makes Sports its most expensive category at a $14.41 average CPT and $19.74 CPA — while also giving Sports the highest conversion rate in its set at 73.0%.

That combination is the clearest illustration of the high-ARPU dynamic in the whole dataset: the most expensive category to buy is the one that converts best. Expensive traffic is not being punished for being bad — it is expensive because well-informed bidders concluded those users are worth it, which is the expensive-user paradox showing up in third-party data rather than an account audit.

→ FREE 10-POINT AUDIT

Find out where your account sits against its real cohort

We open your Apple Ads account, match your keywords and geographies to the right benchmark rows rather than a global average, and show you which of your costs are genuinely out of band. One hour. Senior strategist. No pitch.

Book my free audit →

05The stale-benchmark problem

Business of Apps carries a “last updated” date of 16 June 2026 and serves first-half 2022 numbers underneath it. That is less a criticism of the page than a warning about how these figures travel: the CMS date is when the page was touched, not when the data was collected, and search engines surface the former.

The 2022 set — CPT $1.56, CPA $2.40, TTR 9.99%, CR 64.96% — was accurate when gathered. The same page carries a 2021 average CPT of $11.73, inflated by that year’s crypto-app bidding war: a reminder that one vertical’s speculative behaviour can distort a category average for a full year.

The practical rule is to require a data vintage alongside every benchmark you cite, and to state it in the same sentence as the number. "Average CPT is $1.56" is a claim about 2026. "SplitMetrics measured an average CPT of $1.56 in the first half of 2022" is a fact. The second is less impressive in a deck and considerably harder to be wrong about.

06Why 2025 tap-through rates describe a dead auction

Of the four metrics in every benchmark set, tap-through rate is the one most exposed to platform change, because TTR is a function of where an ad sits on the screen and what shares it — not just creative quality.

Multi-slot search results began rolling out in the UK on 3 March 2026, reached Japan next, and completed globally by the end of March, on iOS and iPadOS 26.2 and later. Up to two ads can now appear per query. Active search-results campaigns qualified for all positions automatically, with no modification and no way to select or bid for a specific slot — the auction decides on bid and relevance. Pricing is unchanged: cost per tap or cost per install.

Every figure in the table above predates that. The published TTR benchmarks — 6.6% to 9.99% depending on source — describe a screen where one ad competed with organic results. On a two-ad screen the second position is a different product with different expected engagement, and a campaign serving across both reports a blended TTR with no 2025 equivalent.

The measurement problem is real and unsolved: Apple does not break impression share out by slot position, so you cannot tell from reporting whether a TTR decline reflects worse creative or more second-position deliveries. Industry commentary on the rollout flagged exactly this, alongside an expectation that organic traffic would fall. Until someone publishes a post-March-2026 sample, treat every TTR number as historical and be sceptical of TTR improvement tactics benchmarked against the old auction.

07How to actually use a benchmark

Four rules, in order of how much money they save.

Match the denominator to your measurement. If you count installs at first open, use a CPI source. If you report on downloads, use Adapty. Mixing them produces a systematic error in one direction, which is worse than noise because it looks like a trend.

Read the row, never the headline. Your category and your country, or nothing. Given a 10× category spread and a 2.1× US-versus-global multiplier, a global average is wrong by a factor that dwarfs any optimisation you will make this quarter.

Use it as a variance check, not a target. The correct question is "am I strange?" — not "am I above or below average?" If your CPT sits at three times your category row, that is a prompt to look at match types, brand-versus-generic mix and what you are paying for your own name. If it sits inside the band, the benchmark has told you nothing actionable and you should stop looking at it.

Set the bid from your economics, not the table. Your ceiling is net revenue per download × tap-to-download conversion rate ÷ payback multiple, and every term in it is yours. A bid set from someone else’s average is a bid set from someone else’s business model — the same discipline that makes ROAS a better optimisation target than CPI.

08What they all agree on

Strip out the disagreements and a consistent picture survives, which is worth stating because it is the part you can rely on.

All four put Apple Ads conversion rates in a band no other paid channel approaches: 55% to 66.2% tap-to-install, across four panels and four years. Apple’s own figure, on data spanning November 2024 to October 2025, is more than a 60% average conversion rate for ads at the top of search results, and it reports that nearly 65% of downloads happen directly after a search.

That is the mechanism, not a marketing claim. A user typing a category term into App Store search has declared intent before you pay anything, and the ad sits on the exact surface where the download decision happens — no landing page, no app-switch, no re-engagement window. Search intercepts declared intent, which is why conversion is high and, unavoidably, why taps are expensive.

And the underlying user is worth more. Across 2025 aggregates, App Store consumer spend has been reported at roughly $117 billion against Google Play’s $49 billion, on about 142 billion combined downloads of which Play carries roughly three times more. Dividing gives approximately $3.30 versus $0.46 per download — derived, not cited, because the aggregates come from different source sets and Sensor Tower’s 2026 report gives a combined $167 billion across 149 billion downloads with no clean platform split. Order of magnitude, not a decimal.

Put the two together and the benchmark spread reads differently. A US median CPT of $1.91 against an Android CPI of a fraction of that is not evidence that Apple Ads is expensive. It is the market pricing a materially more valuable user, on the surface where that user has already told you what they want.

09What no public dataset covers

  • No post-multi-slot data of any kind. The auction changed in March 2026 and every published benchmark predates it. Any 2026 TTR figure you see describes a one-ad screen.
  • No per-cell sample sizes. None of the four vendors publishes how many campaigns sit behind a given category-and-country cell. A Sports-in-the-US median could rest on thousands of campaigns or on a dozen, and you cannot tell.
  • No creative-assets data. The creative assets format ships with iOS 27 in autumn 2026 and carries a 4+ age-rating ceiling on every asset. It has not shipped, so any claimed performance lift for it is extrapolated or invented.
  • No incrementality. Every figure here is a platform-reported cost or rate; none answers whether the install would have happened anyway. Singular’s 2026 index places Apple Search Ads on its Exclusive Reach multi-touch leaderboard, but that is a reach claim, not an incrementality measurement, and the two are routinely confused.
  • No controlled study of Maximize Conversions. It reached general availability in February 2026; no independent evaluation exists.
  • No EU-versus-US split. With EU App Store commission moving to 26% on 1 October 2026, the economics of an EU install are about to change — and there is no baseline EU benchmark against which to measure it.
  • No Apple-published benchmark. Apple releases conversion and search-behaviour statistics selectively and discloses no advertising revenue line item at all. There is no first-party cost benchmark to check the vendors against.

Two standing corrections while we are here, because both keep surfacing in benchmark write-ups: the platform has been Apple Ads, not Apple Search Ads, since April 2025, and the Custom Product Page limit is 70, not 35, since October 2025. SKAN 5 never shipped, so any benchmark methodology referencing it is describing something that does not exist.

The most useful thing on this page is not a number. It is the range — and the knowledge that a 10× category spread and a 2.1× geography multiplier sit inside it, which makes the right response to “what is a good CPT?” another question: which app, which country, measured how.

→ SOURCES
  1. Adapty, Apple Ads benchmarks 2026: CPT, TTR, CPA and more across 90 countries, published 10 June 2026.
  2. AppTweak, Apple Ads benchmarks 2026: CPT, CPI, CR & TTR by category, updated 22 July 2026.
  3. SplitMetrics, Apple Ads cost: how to evaluate CPT, CPA, CR, TTR, updated 13 August 2026.
  4. Business of Apps, Apple Search Ads costs (2026).
  5. PPC Land, Singular’s 2026 ROI Index: Apple Ads enters the top tier of mobile ad networks, 19 April 2026.
  6. PPC Land, Apple will squeeze more ads into App Store search — here’s what changes.
  7. 9to5Mac, The App Store will start showing multiple ads in search results beginning in March, 22 January 2026.
  8. Search Engine Land, Apple will add more App Store search ads.
  9. Search Engine Land, Apple Ads: what you need to know in 2026.
  10. PPC Land, Apple forces 4+ rating on App Store creative assets arriving this fall, 5 August 2026.
  11. Apple Newsroom, Apple announces changes for apps in the European Union, 18 August 2026.
  12. Phiture, WWDC26 recap: App Store and platform updates for mobile growth teams.
  13. eMarketer, Apple expands App Store search ads; advertisers will soon reach high-intent users.
  14. AppleInsider, More ads are coming to the App Store in 2026, 18 December 2025.
  15. Search Engine Journal, PPC Pulse: more Apple search inventory, exact match limits in AI Overviews.
  16. PPC Land, Apple’s record $111bn quarter: Services, ads, and a CEO farewell.
  17. Apple Developer, Apple Ads Platform API documentation.