Eight criteria, published weights, a dated verdict, and a flag that tells you how deep the review went. Here is the system.
What a desk review is, and what it is not
Every score on this site today is a desk review. That means we score a tool against the rubric using what can be checked without owning it: the developer's App Store or Google Play listing, its help center, its pricing page, its privacy label, its version history and its published terms. We read those on a stated date and link them on the page.
A desk review can answer real questions. Does the app work offline? Can it import your files? Does the free tier time out? What does the privacy label say is used to track you? How many steps does the published flow take before you see a page? What does the trial turn into, and when?
It cannot answer others. We do not report how a voice sounds after an hour, how long Time to First Page takes on a stopwatch, or how cancellation feels in practice, because we have not measured those. Where the rubric names a hands-on measure (Time to First Page, cancelling a trial), a desk-reviewed score uses the published facts instead and the field says "Not yet measured".
Every score carries its flag. If we later use a tool hands-on, the flag changes, the date changes, and the changelog says what moved.
Why distraction surface counts most
A reading tool has one job: get out of the way. Every store hook, notification, streak badge, and infinite settings menu is surface area for the exact brain pattern this audience is managing, so distraction surface carries the heaviest weight. The question is how many escape hatches lead from the page to somewhere that isn't the page.
The rubric
| Criterion | Weight | What we measure |
|---|---|---|
| Distraction surface | 20 | Escape hatches from the reading surface; store hooks; notification defaults |
| Start friction | 15 | Time to First Page, protocol below |
| Quality | 15 | Voices for TTS and audio; screens for devices; parsing for read-it-later |
| Format and library fit | 15 | Library borrowing, file formats, import paths, ownership |
| Retention support | 10 | Highlights, notes, resurfacing, export |
| Price honesty | 10 | Trial terms as written vs as experienced; we cancel and document |
| Accessibility | 10 | Text scaling, spacing controls, screen reader behavior, contrast |
| Staying power | 5 | Update cadence, company signals, lock-in risk |
What is Time to First Page?
Time to First Page is the seconds from opening a tool for the first time to actually reading or hearing your own content, timed cold: fresh install or fresh account, default settings, a real book or article you brought. Three runs, median published. Account walls, upsell screens, and forced tutorials all count against the clock, because they are the experience. Unmeasured tools show "Not yet measured" instead of a guess, and today every tool on the site shows it.
Verdicts
SHELF IT: buy or install; it earns a permanent place. BORROW FIRST: real merit, real caveats; trial it, use the library, or buy one credit before subscribing. LEAVE IT: your attention deserves better, and the review says exactly why.
Confidence flags
Every score carries a flag: Hands-on tested, Desk review (provisional), or Community-sourced with the number of reports stated. Today every score on the site is a desk review. A desk-review flag is a receipt, not an apology; if we test a tool hands-on, the score updates and the changelog logs the change.
Prices, corrections, and the money
Subscription prices recheck monthly on high-traffic pages; every review shows the date. Every fix we make is logged in the changelog at the bottom of the page it changes, with the date. Affiliate links exist and never touch a verdict; the full policy is at editorial policy. Books link to Bookshop.org first, because independent bookstores should outlive this website.
Where the rubric is applied
Every score on the site comes from this page: the toolkit collects them, the category hubs rank them, and the reviews show the criterion-by-criterion math.