ATT Opt-In Rate Benchmarks by Category, and How to Read Them
A good ATT opt-in rate in 2025 is roughly 35% of the users who actually saw the prompt, per Adjust's Q2 2025 benchmark data. Gaming pulls well above that, social sits below it, and everyone else clusters in the low-to-mid thirties. But that headline number is the single most misread figure in mobile measurement, and if you're using it to size your addressable audience you're probably off by a factor of two or three.
I'll get to the table. First I want to show you why the table is a trap.
The benchmark most people quote is a per-prompt rate
Here's the thing every dashboard buries in a tooltip: the 35% is a share of users who were presented the prompt, not a share of your installs. Those are wildly different denominators. Adjust, Singular, AppsFlyer, and the rest each compute opt-in on the population that saw the OS dialog, because that's the only population the SDK can observe cleanly.
That framing hides two leaks. Some users never reach the prompt at all (they bounce before onboarding fires it), and Apple's own restricted-tracking settings mean a chunk of your base is silently pre-declined at the OS level. Singular's 2024 write-up spent most of its length untangling exactly this, because the "rate went down" panic that year was partly a denominator artifact, not a behavior change.
So when a vendor tells you the average is 35%, mentally append: "of the people who got asked, in the geos where the OS lets the question through." That's a much smaller slice of reality than it sounds.
Opt-in rate benchmarks by app category, Q2 2025
With that caveat stapled to the top, here's where the verticals landed. All figures are per-prompt opt-in from Adjust's Q2 2025 category benchmarks unless noted, and I've rounded to whole points because nobody should be defending a decimal place on this metric.
| App category | Opt-in rate (per prompt) | Read it as |
|---|---|---|
| Sports | ~50% | Fans expect personalization; the value trade is obvious to them |
| Hyper casual games | ~43% | High volume, low friction, ad-funded, so users get the deal |
| Action games | ~40% | Engaged sessions, willing to trade for "better experience" framing |
| Board games | ~30% | Jumped notably in 2025; casual-but-loyal audience |
| Commerce / e-commerce | ~34% | Roughly the industry average |
| Education | ~14% | Up from just 7% in 2023, the biggest relative climb of any vertical |
| Social | ~26% | Below average; privacy-sensitive context, personal data on show |
| Industry average | ~35% | Up from 34.5% (Q2 2024) and 34% (Q2 2023) |
A few things I'd flag before anyone screenshots this.
The gaming spread is real and it's structural, not clever. Game users have been trained for a decade that "allow tracking" buys them a smoother, more relevant experience, and the ad-funded model makes the exchange legible. Education climbing from 7% to 14% is the most interesting line in the whole set. That's a doubling in two years, and it almost certainly reflects better prompt copy and onboarding rather than any shift in what education apps do with the ID.
And notice the industry average has crept up about a point a year. That's not a rebound. That's slow, grinding improvement in prompt design across the board. Anyone selling you a "recover your opt-in rate" playbook that promises 15-point swings is selling you the pre-prompt, which brings me to the one lever that actually moves the number.
Pre-prompt priming: the only reliable lift
You cannot A/B test Apple's system dialog. The copy is fixed, the buttons are fixed, and you get exactly one shot per install. What you can control is the screen you show immediately before it, the pre-prompt (or priming screen), where you explain in your own words why saying yes helps the user.
The lift here is genuine and it's large. Jampp's pre-prompt guide documents Voodoo's Helix Jump reaching a 59% opt-in rate using an education screen that framed tracking as better gameplay and relevant ads before the OS ever asked. Across MMP guidance the pattern is consistent: a well-built priming screen adds double-digit percentage points versus firing the raw system prompt cold.
Why it works is behavioral, not technical. The system prompt is a yes/no with a scary verb ("track"). The pre-prompt lets you pre-load context and, critically, lets users who'll say no bail on your screen, where a decline costs you nothing, instead of on Apple's, where a decline is permanent and you can never ask again. That second part is the underrated half. A pre-prompt is as much a filter as a persuader.
The don'ts are where teams torch their rate. Do not make the pre-prompt look like the system dialog (Apple's review team treats mimicry as a rejection risk). Do not bribe for consent. And do not fire the prompt at cold-start before the user has done anything, which is the single most common self-inflicted wound I see. You're asking for tracking permission from someone who doesn't yet know what your app is. Earn one moment of value first.
The math nobody puts on the slide: addressable reach
Now the part that made me want to write this. A high opt-in rate does not give you a high match rate, because attribution needs consent on both sides of the transaction.
Walk through what actually has to happen for a deterministic IDFA match. The publisher app (where the ad was shown) needs the user opted in, so it can pass a usable IDFA on the click. And the advertiser app (the one that got installed) needs that same user opted in, so it can read an IDFA on the install and match the two. One consent is not enough. You need the intersection.
So the addressable rate isn't your opt-in rate. It's roughly your opt-in rate times the other side's opt-in rate:
addressable ≈ P(source opt-in) × P(destination opt-in)
Plug in the industry average both ways. 0.35 × 0.35 ≈ 0.12. Twelve percent. That's the fraction of your install traffic that's deterministically matchable on the IDFA path, and it lines up uncomfortably well with the low-to-mid-teens global figures Business of Apps tracks for actual deterministic coverage. The two-sided multiplication is why your "35% opt-in" quietly becomes a ~12% addressable reality before anyone touches your funnel.
This is also the resolution to a discrepancy AppsFlyer flagged directly: high ATT opt-in volumes alongside stubbornly low IDFA collection rates. In their analysis, the gap is exactly this cross-side dependency plus the OS-level restrictions that swallow a device even after a user taps "Allow." Consent granted is not the same as an ID collected, and an ID collected is not the same as an ID matched.
I'll put it in the units I actually think in. If you model a 35% opt-in as your addressable base, and the real deterministic base is 12%, you've overstated matchable reach by roughly 2,300 basis points. That's not a rounding error. That's the difference between a media plan that pencils out and one that quietly bleeds budget into unmatchable traffic every day.
One caveat, because I don't want to oversell the tidy multiplication. The two rates aren't perfectly independent. Opt-in correlates with geography, device age, and app category, so a games-to-games install path will match richer than a finance-to-social one. The 0.35 × 0.35 model is a planning heuristic, not a physics equation. Use it to sanity-check a reach claim that smells too good, not to forecast to two decimals. If a partner tells you they can retarget 30% of your iOS base deterministically, this is the back-of-envelope that tells you to ask harder questions.
What to actually track instead
Opt-in rate is a fine health metric for your onboarding. It's a terrible planning metric for reach. Track it, but pair it with numbers that reflect what you can measure downstream.
Watch your IDFA collection rate, not just opt-in: the share of installs where you actually obtained a usable device ID after all the OS filtering. It'll sit well under your opt-in rate, and the gap is your OS-tax. Watch your deterministic match rate on the campaigns that matter, because that's the two-sided number that governs retargeting. And build your iOS plan assuming SKAN (and its aggregated, delayed, privacy-thresholded postbacks) is the primary signal for the ~65% that never opts in, with IDFA as the bonus layer on top.
If you want the wider frame on measuring accurately once the IDFA is gone, I laid out how the privacy-safe pieces fit together in our privacy-first attribution reference architecture. And if you're benchmarking opt-in against your broader product funnel (activation, retention, the stuff opt-in is supposed to eventually feed), the 2026 product analytics benchmarks give you the yardsticks for the metrics that survive the privacy tax.
Frequently asked
What's a good ATT opt-in rate? Around 35% of prompted users is the 2025 average per Adjust. Gaming apps run 40-50%, social sits near 26%, and most other verticals cluster in the low thirties. Judge yourself against your category, not the blended number.
Why is my opt-in rate so much higher than my match rate? Because matching needs consent on both the source and destination app. Your rate times the other side's rate is roughly your addressable base, often around 12% when both sit at the industry average.
Do pre-prompts actually work? Yes, and they're the only lever that reliably moves the number. Documented cases like Voodoo's Helix Jump hit 59% with a good priming screen. Just don't mimic the system dialog or fire it cold at launch.
Should I optimize for opt-in rate at all? Optimize the pre-prompt and prompt timing, sure. But plan your media and measurement around SKAN and match rates, because opt-in rate overstates how much traffic you can deterministically measure.
The number on the dashboard isn't lying to you exactly. It's just answering a narrower question than the one you're asking. Read it as "share of prompted users who agreed," multiply by the other side, and plan for the twelve percent, not the thirty-five.