Original research

The iOS Attention Census 2026

By Juhyun NamData collected 30 August 2026US App Store

Of 692 iOS apps built to fight phone overuse, 511 (73.8%) work by blocking whole apps and only 49 (7.1%) act on the feed inside an app. Add the 84 apps (12.1%) that put friction in front of an app and then let you through, and 595 of the 692, 86.0% of the category, work by placing a wall or a speed bump in front of the whole app rather than changing what is inside it. That is the result of a census of the US App Store's screen-time category, run on 30 August 2026 against twelve search terms frozen before any data was collected.

Every number on this page is computed from a published 692-row dataset, which you can download at the bottom and check. The method is written out before the results, on purpose. We also publish the findings that cut against us, because a page that only carried the flattering half would not be worth citing.

The method, before the results

Research about an industry, published by a company in that industry, is worth exactly as much as its method. So here is the method, first and in full.

The twelve search terms, frozen before collection

These twelve terms were written to a file before a single API call was made, and were not changed afterwards. That file is the pre-registration. Nothing was added once the results were visible, which is the usual way a category study quietly becomes an advert.

• screen time
• app blocker
• block apps
• stop scrolling
• doomscrolling
• focus app
• digital wellbeing
• phone addiction
• block social media
• block reels
• block shorts
• distraction blocker

The query

Each term was queried once against Apple's public iTunes Search API, on the US storefront, on 30 August 2026:

Every raw response was saved verbatim before anything was filtered, so the run can be re-read and re-run. No term came back at the API's per-term cap, so nothing was truncated by the limit. Apple's own ranking still decides which apps a term returns, which is a real constraint and is listed in the limitations below.

From 2,149 rows to 692 apps

The four census genres are Productivity, Health & Fitness, Lifestyle and Utilities. The topical screen kept apps whose listing is about phone or app use at all. Everything that survived both filters was then read individually.

What was excluded, and why

35 apps passed the automatic screens and were still thrown out after a person read the listing.

How each app was classified

All 692 were read individually and given exactly one primary intervention type, with a verbatim sentence from the app's own App Store description recorded as the evidence for the call. Quotes are pulled from the description by sentence index when the dataset is built, so no quote was retyped and none can drift from its source.

• app-level blocks, locks or removes whole apps, or the device.
• feed-level removes or hides specific feeds or surfaces while the rest of the app keeps working.
• friction-level interposes a delay, a breath, a question or a reminder, and then you continue.
• tracking-only measures and reports usage without intervening.
• replacement substitutes a different activity, feed or app.
• unclear the description does not settle the mechanism.

Where an app both blocks and interposes, and most do, it was classified on the mechanism its own copy leads with. Two secondary flags were also recorded: audience (self-directed or parental) and purpose (core, meaning the app exists to manage phone use, or adjacent, meaning a focus timer, app lock, faith app or fitness app that adds blocking as a feature).

A correction we made to our own method

The first version of the topical screen was wrong, and the way it was wrong matters. It was built around screen-time vocabulary, and a separate sweep for feed language showed that it was missing feed-level apps for exactly that reason: many of them never say "screen time" or "addiction" at all. They talk about Reels and Shorts. We widened the screen, re-ran it, and reviewed the additional in-genre apps it surfaced.

The correction moved the headline number against the story we would have preferred to tell. A narrower screen would have found fewer feed-level apps and made "almost nobody does this" look stronger than it is. The 49 apps and 7.1% published here are the figures after the fix, not before it. The earlier version of the screen is kept alongside the current one so the two can be compared. Only the term list is pre-registered; the screen is not, and we would rather say that plainly than let you assume otherwise.

What the category actually does

The category has converged on one answer: take the app away. That is what 511 of the 692 do outright, and another 84 slow you down on the way in. Only 15 apps, 2.2%, are content to measure and tell you, which suggests that pure tracking has gone out of fashion.

The obvious objection is that the census is diluted by parental controls and by productivity tools that merely include a blocker. It is not. Restricting to the 582 self-directed apps that exist for this purpose barely moves anything.

For scale: parental-control products are 30 apps, 4.3% of the census, and 89 apps are adjacent tools that added blocking to something else.

The feed-level group, named in full

49 is a small enough number to print, so here is the whole feed-level group, with the rating counts Apple reported on 30 August 2026. Ours is in there, near the bottom.

Those 49 apps hold 8,425 ratings between them. The single largest app-level app in the census holds 289,248 on its own. Median rating count is 4 for feed-level apps and 8 for app-level ones, so the niche has not translated into more traction per app; it has translated into less.

The group is also young and lopsided. 77.6% of feed-level apps first shipped in the last 12 months, against 58.3% of app-level apps. 9 ship as Safari extensions and 17 describe themselves as a browser. And they are not evenly spread across platforms.

10 of the 49, 20.4%, handle YouTube and nothing else, and only 17, 34.7%, name three or more platforms. TikTok, an app that is almost entirely feed, is named by 14.3%. Whatever else feed-level tooling is, it is mostly a YouTube story so far.

Price

681 of the 692 apps, 98.4%, were free to download on 30 August 2026, and the median price across the census is $0.00. Only 11 apps, 1.6%, charge anything upfront, and the median price of those is $3.99, in a range from $0.99 to $4.99.

What this does not tell you is what these apps cost to use. The iTunes Search API exposes no in-app-purchase or subscription field, so free versus free-with-subscription is simply not measured here, and we are not going to pretend otherwise. As a text proxy only, 355 listings, 51.3%, mention a subscription or a premium tier somewhere in the description. That is a proxy, not a measurement.

Ratings, and how concentrated they are

snowscroll is excluded from every figure in this section, because 2 ratings is not a number you can compare anything to. That leaves 691 apps.

The median app has 6 ratings. The mean is 1,539, which is a meaningless figure and is included only to show how skewed the distribution is: the five largest apps hold 546,561 ratings, 51.4% of every rating in the census. Among apps with at least one rating, the median average star rating is 4.72, so the visible scores tell you very little.

This category is eight months old

The median app in the census first shipped 8.4 months before the data was collected. 61.6% first shipped in the last 12 months and 78.3% in the last 24.

The category is also being maintained: 67.9% were updated in the last 3 months and 90.8% in the last 12. Only 9.2% have gone more than 12 months without an update, and 2.6% more than 24. This is not a graveyard of abandoned side projects. It is a gold rush.

What people are being asked to do

The most interesting thing in the descriptions is not the blocking. It is what the blocking is wrapped in.

96 apps, 13.9% of the census, make you earn access: camera-counted push-ups, steps, quizzes, chores, flashcards, a selfie with a friend, a handwritten sentence. 23 apps, 3.3%, require a physical object to unlock, an NFC tag or card, a barcode or a QR code. Meanwhile 181 listings, 26.2%, name Apple's own Screen Time or Family Controls, including 153 of the 511 app-level apps, 29.9% of them.

Put those together and a pattern falls out. When 26.2% of these listings sit on the same Apple API, a large part of the category is not competing on capability. It is competing on ritual. The difference between one app and the next is what you have to do to get back in.

They are also built by very small teams. 676 distinct sellers ship the 692 apps, and 664 of them, 98.2%, ship exactly one. No seller in the census ships more than three.

Feed-level apps outside the census genres

The genre filter was frozen before the data was seen, and it happens to be biased against the finding we care about most: several feed-level apps file under Social Networking or Entertainment rather than the four census genres. Rather than quietly repair the filter, we measured the gap. An exhaustive feed-language sweep of all 1,516 unique apps found 7 more: TimeCap, YT Shorts Blocker, Icare, Block Reels - Less Is More, FeedLite, ShortVidsBlocker and yarrOS.

Feed-level apps found anywhere in the corpus: 56, which is 3.7% of the 1,516 unique apps the frozen terms returned. That figure is reported separately from the census on purpose, so that the pre-registered method stays intact and you can see both numbers.

What this does not show

The limits of this study are as load-bearing as the results, so they are not going at the bottom in small type.

• It is a term-based sample, not an enumeration of the App Store. It is what these twelve terms returned. The clearest illustration is RepsForReels, a real app in this category that we verified was live on 30 August 2026 and which appears nowhere in the 1,516-app corpus. If it can be missed, so can others.
• Apple's ranking chose the sample. No term hit the API's per-term cap, so nothing was cut off by the limit, but which apps a term surfaces is Apple's decision, not ours.
• One snapshot, one storefront. US App Store, one day. Ratings, prices, descriptions and listings all change, and some of these numbers will already be slightly wrong by the time you read them.
• The genre filter excludes some feed-level apps. Anything filed under Social Networking, Entertainment or Education is outside the census. The section above quantifies that gap rather than hiding it.
• Classification is from App Store description text only. No app was installed, bought or tested. An app is classified on what it says it does. Marketing copy is not behaviour.
• The app-level and friction-level boundary is a judgment call, and it is the softest part of the taxonomy, because most apps do both. Every row carries the exact sentence the call was based on, so you can disagree row by row.
• One reviewer, no second rater. There is no inter-rater reliability figure, because there was only one rater.
• Subscription pricing is not measured. The API does not expose it.

The findings that are inconvenient for us

We make a feed-level app. Here is everything in this dataset that argues against our own pitch, stated without softening, because you should not have to take our word for the rest of the page.

1. "Almost nobody does feed-level blocking" is false as a claim about counts. 49 apps in the census do it, and 56 across the whole corpus. The honest claim is about adoption, not existence.
2. snowscroll is neither the first nor the largest feed-level app. SocialLite alone holds 6,334 of the 8,425 ratings held by the entire feed-level group. BeTimeful, UnTrap, SocialFocus, Unhook, #blockit and ScrollGuard all shipped before we did. snowscroll has 2 ratings.
3. Feed-level apps have a lower median rating count than app-level apps, 4 against 8. Being in a narrower niche has not meant more traction per app. It has meant less.
4. The idea is being copied fast. 77.6% of feed-level apps launched in the last 12 months. Whatever advantage exists here is not a moat.
5. The category is not asleep. Only 9.2% of these apps have gone a year without an update. Any pitch that rests on incumbents being stale is not supported by this data, including ours.
6. Our own rating is excluded from every comparative figure on this page, and should stay excluded. 2 ratings is not a number.

Where snowscroll fits

snowscroll is one of the 49 feed-level apps in this census, and one of the smallest, with 2 ratings against the 6,334 held by SocialLite, which we compare ourselves against without pretending we are the established one. It is a browser that opens Instagram, YouTube, Facebook, X, TikTok and LinkedIn with the feed surfaces not loaded, which puts it in the 7.1% column rather than the 73.8% one. If a whole-app block is what you actually need, this census contains 511 of them and 98.4% of the category is free to download.

The data

The full census is published as a CSV: ios-attention-census-2026.csv, 692 rows, one per app. Each row carries the App Store id, app name, seller, price, star rating, rating count, first release date, last update date, genre, our intervention type, the audience and purpose flags, the verbatim evidence sentence the classification was based on, the App Store link, which of the twelve terms returned the app, and a short reviewer note.