Semantic Navigator
Navigate the web the way your users have to.
Most accessibility problems don't look like problems — until you remove the visual layer. Semantic Navigator replaces the page with its accessibility tree. You can still click, navigate, fill forms, and complete tasks. The structure is all you get.

The test
If you can complete the task through the accessibility tree, your users probably can too. If you can't, that's the bug.
- Open the side panel. Click the extension's icon in the Chrome toolbar. The panel opens next to the page and shows its tree.
- Turn on the curtain. The Curtain button in the panel's toolbar blacks out the page. The tree and every interaction stay.
- Do a real task through the tree. Check out, book an appointment, find the error after a failed submit, close the dialog that just opened. Select a row and press
Enterto act on it. - Stop where you get stuck. A button you can't find, a control you can't tell the purpose of, an action you can't tell worked: that's the bug your screen reader users hit.
The panel shows the tree screen readers read, not what NVDA, JAWS, VoiceOver or TalkBack say or how they move through a page. Use it to catch missing names, wrong roles and broken tab order early, and test with real screen readers before you ship.
Chrome's tree
The panel reads Chromium's own accessibility tree, the one Chrome gives assistive technology, over chrome.debugger. It sees what a reading of the page from inside it can't, such as a video player's built-in controls. Where Chrome won't let it read a page, with DevTools open on the tab, say, the panel says why.
While it reads, Chrome shows a bar across every window: “Semantic Navigator” started debugging this browser. That's expected. Pressing Cancel on the bar stops the panel reading that tab until you press ↻.
Chrome's tree is read when the panel opens, when the page navigates, and for a few seconds after you act. For any other change press ↻, or turn on Follow page changes (⟳). It has no tab order of its own, because tabindex never reaches a native node, so its TAB view, next to A11Y in its toolbar, lists the tab stops of the extension's own reading of the page, in tab order, as the TAB view of the other panels does: a positive tabindex first, then document order. Neither models shadow roots, iframes or focus traps. Showing it reads nothing through Chrome, so the bar doesn't come up for it, and it works on a page Chrome's tree can't read. It takes a moment to read the page as it comes up. Chrome's tree and the other panels' trees never match exactly, by design: see Two producers.
Up to version 0.1.17, Settings ▾ could switch the panel to the extension's own in-page tree, with its DOM view. Chrome's tree is the only one now.
Using the panel
Chrome's tree works the same way as the panel of @real-a11y-dev/inspector, <SemanticNavigator /> and the Storybook addon. For what each view shows, see Reading the tree.
- Search and role filters. The search box (
/) keeps the rows whose role, name, tag, attributes or text match, with the rows above them. Headings, Links, Buttons, Forms, Landmarks and Images each list just those nodes. - Selecting a row outlines its element on the page, scrolls it into view and focuses it if it can take focus. Hovering a row outlines it. The inspector and
<SemanticNavigator />do these only if you opt in, withhighlightOnHover,scrollHostOnSelectandfocusHostOnActivate. - Jump chips link a control to what it controls, such as a menu button and its menu:
→ menu "Options"and← button "Options".Alt+Jfollows one. - Modal dialogs. When the page opens a
<dialog>withshowModal(), the tree re-roots to it, as a screen reader gets it. An element withrole="dialog"stays in place among the rest of the page. The panel doesn't check that focus stays in the dialog or that Escape closes it: try both. - Long names end in an ellipsis. Hover one, or turn on Settings ▾ → Show full names.
Only the extension has these:
- Focus sync. Focus moving on the page, by Tab, a click or the page's own script, selects that element's row.
- Scoping. Double-click a container row, or press
Ctrl/Cmd+Enter, to show only that part of the page. Search, filters and Copy follow the scope. The breadcrumb above the tree, orEsc, takes you back. In the other panels, scope with the inspector'ssetRoot()or a differentrootfor<SemanticNavigator />. - The live log. When a live region (
role="status","alert"or"log", oraria-live) changes, its text appears at the bottom of the panel, assertive in red and polite in blue. Use it to check that toasts, validation errors and loading messages are announced. It approximates a screen reader: it shows the region's whole text, and doesn't watch iframes. - The key bar.
Esc,Tab,Shift+Tab,Enter,Space,↓and↑go to whatever has focus on the page, for example to open a combobox and pick an option. - Copy ▾ puts the tree on the clipboard as a Markdown report for a bug tracker, headed by the URL, which tree it came from and the scope, with the tab sequence too while the TAB view shows. Field values are left out.
- Frames. Each iframe's tree sits under its
<iframe>row, cross-origin frames included.
Keyboard
| Key | In the tree |
|---|---|
↑ / ↓ | Previous / next row |
→ / ← | Expand / collapse the row, or move into it / to its parent |
Home / End | First / last row |
Enter | Run the row's action: click, toggle, step a slider up (Shift+Enter steps down) |
Space | Expand or collapse the row |
* | Expand every sibling of the row |
/ | Move to the search box |
| Letters | Jump to the next row whose name starts with what you type |
Alt+J / Alt+Shift+J | Follow the row's jump chip / go back |
Ctrl/Cmd+Enter / Esc | Extension only: scope to the row / leave the scope |
Privacy
Everything happens in your browser. The extension makes no network requests and has no account or tracking. It stores its settings and, while its debugger is attached, which tab it's reading. Password, one-time-code and card fields show [redacted], and Copy leaves field values out. The privacy policy lists every permission it asks for and why.
Troubleshooting
- "This page can't be inspected." Chrome doesn't let extensions run on
chrome://pages, the Chrome Web Store or the PDF viewer. On an ordinary page, reload and press Try again. - The panel is empty after switching tabs. That's deliberate. Press Load tree or ↻.
- Chrome's tree can't read the page. Usually DevTools is open on that tab: close it, then press ↻. The TAB view still lists the tab order meanwhile.
- Chrome's tree stopped updating. Cancel on Chrome's bar, or a failed read, pauses it. Press ↻.
Install
Install Semantic Navigator from the Chrome Web Store. Chrome may warn "Proceed with caution — not trusted by Enhanced Safe Browsing" on first install. That's its default for any newly listed extension, not a finding about this one: click Continue to install.
To build it from source, see the extension's README.