---
title: "Works in incognito but not in a normal window: extension or stored data? | Proofline"
description: "If a page fails in a normal Chrome window and works in incognito, an extension or the site's stored data is usually the cause."
url: https://proofline.sh/learn/works-in-incognito/
---

[Learning center](/learn/)

# Works in incognito but not in a normal window: extension or stored data?

If a page fails in a normal Chrome window and works in incognito, an extension or the site's stored data is usually the cause. Turn on one difference at a time to find which.

Broken dependencies The Proofline team November 20, 2026 4 min read

## The two differences

An incognito window starts with an empty profile for the site, and Chrome runs an extension there only if the user allowed it to run in incognito. [Extensions in incognito](https://developer.chrome.com/docs/extensions/reference/manifest/incognito)

|  | Cause A: an extension changes the page or its requests | Cause B: state stored for the site is stale or wrong |
| --- | --- | --- |
| Examples | A content blocker blocks a request; a password manager or translator edits the DOM; a privacy tool strips headers | An old service worker serves stale files; a cookie holds a broken session; local storage holds data in an old format |
| Fails in a new normal profile? | Only if you install the same extension there | No, the new profile has no stored state |
| Fails in the normal window with extensions off? | No | Yes |

## Separate them in two steps

**Step 1: turn off extensions, keep the stored data.** Open `chrome://extensions`, switch off every extension and reload the page in your normal window. The site's cookies, storage and service worker stay as they were. If the failure stops, re-enable extensions one at a time until it returns. That extension is the cause.

**Step 2: keep extensions, clear the site's data.** If the failure continues with extensions off, the stored state is the suspect. Before clearing anything, record what is there, because clearing destroys the evidence. In DevTools, open **Application** and note:

-   **Service workers**: whether one is registered, and its status.
-   **Storage**: the usage per type, and the keys in Local storage, Session storage and IndexedDB that the app uses.
-   **Cookies**: the names, domains, paths and expiry of the site's cookies. Do not copy their values into the report.

Then use **Clear site data** in the Application panel's Storage section and reload. If the page works, state was the cause. To find which part, repeat with a fresh failure and clear one storage type at a time. [Application panel](https://developer.chrome.com/docs/devtools/application) · [Cookies in DevTools](https://developer.chrome.com/docs/devtools/application/cookies)

## Worked example: a dashboard that loads blank for long-time users

This example is fictional. After a release that renamed a saved-filter setting, some users see a blank dashboard. New users and anyone in incognito see it normally.

With extensions off, the failure continued, which ruled out Cause A. In the Application panel, Local storage held a key `dashboard.filters` with a value in the old format: `{"range": "7d"}` instead of the new `{"period": {"days": 7}}`. The Console showed `TypeError: Cannot read properties of undefined (reading 'days')` at page load. After deleting only that key, the dashboard loaded.

The report said: _users who saved filters before the release have `dashboard.filters` in the old shape in Local storage. The new code reads `period.days` without a fallback and throws at load. Deleting the key fixes it for one user; the code needs to migrate or ignore the old shape._ The receiver could reproduce it in seconds by setting that key.

## When it is an extension

Report the extension's name and version, the request or element it changes, and what you observed. Content blockers often block paths containing words like `ads`, `track` or `analytics`, which shows as `(blocked:other)` in the Network panel's Status column and `net::ERR_BLOCKED_BY_CLIENT` in the Console. See [requests with no status](/learn/failed-request-no-status/). Extensions that edit the DOM can break server-rendered pages during hydration; see the [hydration mismatch guide](/learn/hydration-mismatch-error/).

Whether to change the app is a product decision. Renaming a blocked path or tolerating extra DOM attributes may be cheap; supporting every extension is not. The report should give the receiver enough to decide.

## What to put in the report

-   The result in a normal window, a normal window with extensions off, and incognito.
-   The extension's name and version, if one is involved.
-   The stored keys, cookies and service worker state you saw before clearing, without values that identify a person or session.
-   Which single change made the failure stop.

## Questions

### Does incognito disable the HTTP cache?

An incognito window starts with an empty cache and keeps a separate one for its session. For a clean comparison in a normal window, tick **Disable cache** in the Network panel, which affects requests while DevTools is open. [Disable cache](https://developer.chrome.com/docs/devtools/network/reference)

### Why does it work in incognito and fail again later?

Incognito state builds up during its session, so the state the app wrote since the window opened becomes a suspect. It is not the only one: the server's response may have changed, or an extension allowed in incognito may now act on the page. A second incognito window is not a clean control, because Chrome's incognito windows open at the same time share one session's cookies and storage. Close every incognito window, which ends that session, or use a guest profile, and repeat the action. If it works there, compare it with the failing session's storage, which you recorded in the Application panel before closing. If it fails in the fresh session too, stored state is less likely, and the failing request's response and any extension allowed in incognito are the next suspects.

### Is a guest profile better than incognito?

A guest profile also starts empty and runs no extensions. Either works for this test. Use whichever lets you open DevTools and repeat the steps easily.

Return to the [method for browser bug reports](/learn/browser-bug-report-evidence/) for the other symptoms in this collection.

## One month of free code review.

New organizations get one calendar month of hosted code review for private and public GitHub repositories. No card or cloud connection required. [Eligibility and fair-use terms](/free-code-review/).

[Start free with GitHub](https://proofline.sh/orgs/plan?offer=free_month)

## Related reading

-   ### [Turn a browser symptom into a bug report another engineer can act on](/learn/browser-bug-report-evidence/)
    
    Network outages The Proofline team November 20, 2026 7 min read
    
-   ### [Users still see the old version after a deploy: find which cache answered](/learn/old-version-after-deploy/)
    
    Deployment failures The Proofline team November 20, 2026 5 min read
