P17 min read

Safari vs Chrome: where journeys diverge

Most teams assume parity until a stakeholder opens a Mac. These are the functional differences between Safari and Chrome that I check before UAT.

Evidence pack✓ journey A! deferred riskdecision: ship / wait

Most teams assume Safari and Chrome behave the same until a stakeholder opens a Mac. The build was developed in Chrome and the automated checks ran in Chrome, so the first person to try the journey in Safari is a business owner in a UAT session, on their own iPhone, with their own saved passwords.

The two browsers diverge in places unit tests never see, and nothing throws an error when they do. A session quietly ends. A date comes back empty. A new tab never opens. Functional QA catches these because we follow what the user is trying to do, not a snapshot of the DOM.

Why Safari and Chrome behave differently

They are different engines with different defaults. Chrome is built on Chromium and Safari on WebKit, and where the web standards leave room each has made its own choices. Safari’s lean towards privacy: it restricts cross-site cookies and limits what a site can keep in the browser. Chrome has generally been more permissive, and because most development happens in Chrome, the permissive behaviour gets mistaken for how the web works.

Two things follow. Browsers on iPhone have traditionally been built on WebKit, so Chrome on an iPhone is closer to Safari than to Chrome on a desktop. And Safari on a Mac and Safari on an iPhone share an engine but not a keyboard, a file picker or a screen, so each needs its own look. Which of them to test, and how often, belongs in the cross-browser matrix. This piece is about what to look for once you are there.

Where Safari and Chrome differ in ways that break journeys

These are the differences I check, roughly in the order they hurt.

AreaWhat Safari doesWhat I check
Cookies in embedded contentBlocks third-party cookies by defaultAnything served in a frame from another domain: payment widget, embedded form, chat. Does it keep its session?
Single sign-onBlocks the cookie that a hidden frame to an identity provider on another domain relies on for silent renewalStay signed in past the token lifetime. Signed out early? Redirect loop?
Browser storageCaps the life of cookies set by script and clears local storage for sites the user has not returned toWhat does the journey keep only in the browser? Clear it and return. Do the draft, the consent choice and the remembered device recover?
DatesShows its own native date picker and parses non-standard date strings by its own rulesPick, type and edit dates. Any empty field, “Invalid Date” or a day that shifts?
AutofillDecides for itself which field is which and offers its own strong passwordLet the browser fill the form. Right fields? Does the suggested password pass the rules? Does the button enable?
New tabs and pop-upsIs stricter about windows that are not the direct result of a tap or clickSteps that open a tab after a server call: bank verification, document preview, export
Uploads on iPhoneOffers camera, photo library and files. iPhone photos are HEIC by defaultA photo straight from the camera. Accepted, within the size limit, readable at the other end?

Layout is the other family: the subtle CSS that breaks a form without an error. On an iPhone the browser’s toolbars and the on-screen keyboard take space a desktop layout never allowed for, and a submit button pinned to the bottom of the screen can end up underneath them. Safari on iPhone also zooms in when the user taps a field with small text, which can push the rest of the form out of view. On a Mac, Safari does not move focus to a button when it is clicked, as Chrome does, so a menu that closes when focus leaves it can close before the click lands.

Which journeys to test on both browsers first

I trace the revenue and compliance journeys on both engines first: sign in, payment, document upload, session timeout. If those pass with honest oracles, I expand to secondary flows. Honest means the oracle is the outcome and not the absence of an error. The payment is taken once. The document arrives and can be opened. The user who timed out lands somewhere sensible, with their work intact or a clear message that it is gone.

Session timeout is as much a mobile difference as a Safari one. A warning driven by a timer in the page cannot fire while the phone is locked or the browser is in the background, so the user comes back to an expired session with no notice. I check it on an iPhone because that is where stakeholders tend to meet it.

I reproduce with the data Product will use in the UAT session, not with sanitised dev fixtures. That means the single sign-on route the business uses and not the test login that bypasses it, accounts with saved credentials, a photo from the phone’s camera in place of a small image from a fixtures folder, and privacy settings left at their defaults. Most Safari-only failures need one of those to show themselves.

How to report a Safari-only defect

When I find a gap, I write it as user impact: who cannot complete what, and what workaround exists. A title that only names the browser invites an argument about market share. A title that names the user and the blocked outcome is the kind of defect Product acts on.

A Safari-only defect written two ways (illustrative)
  • As a browser complaint: “Upload broken in Safari.”
  • As user impact: “Customers on iPhone cannot upload proof of address taken with the camera. The file is rejected as an unsupported format.”
  • Who is affected: anyone completing onboarding on an iPhone.
  • Workaround: choose an existing JPEG or PDF from Files. Not obvious to the user.
  • Seen on: Safari, iPhone, physical device, build rc2. Not seen on: Chrome on Android, Chrome and Safari on desktop.

That framing keeps Engineering on behaviour and away from browser religion. It also gives Product what it needs to decide whether the gap blocks UAT or is deferred with a named risk.

The objection: it works in Chrome

The usual response is that Safari is the odd one out and the user should switch browser. On an iPhone that changes little, and for a business owner in a UAT session it is not an answer. The second response is that the difference is a Safari bug. Sometimes it is. More often Safari is applying a stricter default and the product was leaning on the looser one. Either way the user is on the browser they are on.

What goes wrong on the QA side is treating a list like this as a script. It dates. Browsers change their defaults, and Chrome has moved towards some of Safari’s restrictions over time. The part that lasts is the method: know which behaviours your journeys lean on, whether that is a cookie in a frame, a value in local storage, a new tab or a timer, and try those on the least forgiving engine.

Cross-browser confidence is not equal coverage everywhere. It is knowing where your users browse and having evidence that the journeys that matter work there. With that done before UAT, the Mac in the stakeholder’s hands confirms what the readiness note already said, and the session is spent on acceptance.

Questions I get asked

Why does my site work in Chrome but not in Safari?

Usually because the site relies on something Chrome allows and Safari restricts or handles differently. The common causes are third-party cookies in embedded frames, sign-in flows that renew the session in a hidden frame, data kept only in browser storage, date strings in a non-standard format, and new tabs opened after a server call. Walk the journey step by step in Safari and note where the outcome differs.

Do I need to test Chrome on iPhone separately from Safari?

Mostly no. Browsers on iPhone have traditionally been built on Apple’s WebKit engine, so Chrome on an iPhone behaves much more like Safari than like Chrome on Android or desktop. If Safari on iPhone is covered, a short check of sign-in and autofill in Chrome on iPhone is usually enough. It does not stand in for testing Chrome on Android.

What should I test in Safari before UAT?

The journeys that carry revenue or compliance, end to end, on a Mac and on a physical iPhone: sign in through the real single sign-on route, a payment including the bank verification step, a document upload using a photo from the camera, and what happens when the session times out. Use the accounts and data the UAT participants will use.

Filed under Release readiness and regression. Terms: Cross-browser testing.