Are PWAs still relevant?

Short answer

Yes, but far less than the advocacy suggests. We opened 386 installable web apps in a real browser and cut the network: 151 kept working, 235 showed nothing. So the defining promise of the format — it works without a connection — holds for 39% of the apps that claim it. PWAs are relevant where someone did the work; the format is not the guarantee people assume it is.

What we measured, and how

Every app in this directory is opened in a headless Chromium. We wait for the service worker to activate, disconnect the network, reload, and record what is still served. That is the whole method, and it is worth stating plainly because most claims about PWAs are made without one.

On 12 August 2026 the results across 386 installable apps were: 151 work offline, 235 do not. Separately we read every manifest we could reach — 373 of them — and found 215 without a maskable icon, 329 without an id, 331 without screenshots, and 3 apps declaring a manifest that no longer responds at all.

Mentioned here: ✏️ Excalidraw, 🖼️ Photopea, 🔐 CryptPad

Why the number is lower than the discourse

Because the failing state is one nobody tests. A developer builds the app online, checks it online, ships it, and the install prompt appears — every signal says success. The blank screen only exists for a user who installed it, launched it later, and had no connection. That user does not file a bug; they delete the icon.

It compounds because the tooling reports the wrong thing. A registered service worker is treated as a pass by most audits, but a worker with an empty cache is indistinguishable from a working one until you actually cut the network. We wrote about that failure specifically — it is the single most common one we see.

Where they genuinely win

The apps in the 39% are not marginal. Photo editors that open PSDs, whiteboards, file transfer, audio editors, note-taking — real software that installs in one click, costs no app-store commission, updates without review queues, and runs the same code on iOS, Android, Windows, macOS and Linux.

The economics are the strongest argument and they have not changed: no 30% cut, no gatekeeper who can remove you, no separate codebase per platform. What has changed is the honest framing — the format enables this, it does not deliver it. Someone still has to cache the right things.

Where the format still loses

Installation on iOS remains the hard wall. There is no automatic prompt: the user must find Share, scroll, and tap Add to Home Screen. Anyone who has walked a non-technical person through it knows the completion rate is poor, and no amount of manifest work changes it.

That wall, and the discovery gap, are why the original promise did not land. Discovery is the other. An app store is a place people already go; a web app has to be found. That is the gap this directory exists to close, and it is also why measuring matters — a list nobody can trust is worse than no list.

Frequently asked

How many PWAs actually work offline?

Of 386 installable web apps we opened in a real browser with the network cut, 151 kept working — 39%. The remaining 235 install correctly and then show nothing, usually because the service worker is registered but its cache is empty.

Did PWAs fail?

No, but the story that they would replace app stores did. What exists is a format that works well for tools and utilities, badly for discovery, and unevenly in practice — 39% of installable apps deliver offline use, and most manifests are missing details that make an installed app feel finished.

Is it worth building a PWA in 2026?

For a tool, a utility, or anything where you want no gatekeeper and one codebase — yes, and the measured leaders show it works. For anything depending on store discovery or effortless iPhone installation, the constraints have not improved and you should plan around them rather than hope.

Also published on Medium. This page is the original.