Skip to main content
Version: 21.3.4

Query Selectors

Queries are the primary mechanism for interacting with the DOM on your site. For example, a typical workflow goes like:

// Import puppeteer
import puppeteer from 'puppeteer';

(async () => {
// Launch the browser
const browser = await puppeteer.launch();

// Create a page
const page = await browser.newPage();

// Go to your site
await page.goto('YOUR_SITE');

// Query for an element handle.
const element = await page.waitForSelector('div > .class-name');

// Do something with element...
await; // Just an example.

// Dispose of handle
await element.dispose();

// Close browser.
await browser.close();

P Selectors

Puppeteer uses a superset of the CSS selector syntax for querying. We call this syntax P selectors and it's supercharged with extra capabilities such as deep combinators and text selection.


Although P selectors look like real CSS selectors (we intentionally designed it this way), they should not be used for actually CSS styling. They are designed only for Puppeteer.


P selectors only work on the first "depth" of selectors; for example, :is(div >>> a) will not work.

>>> and >>>> combinators

The >>> and >>>> are called deep descendent and deep combinators respectively. Both combinators have the effect of going into shadow hosts with >>> going into every shadow host under a node and >>>> going into the immediate one (if the node is a shadow host; otherwise, it's a no-op).


A common question is when should >>>> be chosen over >>> considering the flexibility of >>>. A similar question can be asked about > and a space; choose > if you do not need to query all elements under a given node and a space otherwise. This answer extends to >>>> (>) and >>> (space) naturally.


Suppose we have the markup

<template shadowrootmode="open">
<template shadowrootmode="open">
<template shadowrootmode="open">
<h2>Light content</h2>

Note: <template shadowrootmode="open"> is not supported on Firefox. You can read more about it here.

Then custom-element >>> h2 will return h2, but custom-element >>>> h2 will return nothing since the inner h2 is in a deeper shadow root.


P elements are pseudo-elements with a -p vendor prefix. It allows you to enhance your selectors with Puppeteer-specific query engines such as XPath, text queries, and ARIA.

Text selectors (-p-text)

Text selectors will select "minimal" elements containing the given text, even within (open) shadow roots. Here, "minimum" means the deepest elements that contain a given text, but not their parents (which technically will also contain the given text).

const element = await page.waitForSelector('div ::-p-text(My name is Jun)');
// You can also use escapes.
const element = await page.waitForSelector(
':scope >>> ::-p-text(My name is Jun \\(pronounced like "June"\\))'
// or quotes
const element = await page.waitForSelector(
'div >>>> ::-p-text("My name is Jun (pronounced like \\"June\\")"):hover'

XPath selectors (-p-xpath)

XPath selectors will use the browser's native Document.evaluate to query for elements.

const element = await page.waitForSelector('::-p-xpath(h2)');

ARIA selectors (-p-aria)

ARIA selectors can be used to find elements with a given ARIA label. These labels are computed using Chrome's internal representation.

const node = await page.waitForSelector('::-p-aria(Submit)');
const node = await page.waitForSelector(
'::-p-aria([name="Click me"][role="button"])'

Custom selectors

Puppeteer provides users the ability to add their own query selectors to Puppeteer using Puppeteer.registerCustomQueryHandler. This is useful for creating custom selectors based on framework objects or other vendor-specific objects.

Custom Selectors

You can register a custom query handler that allows you to create custom selectors. For example, define a query handler for getById selectors:

Puppeteer.registerCustomQueryHandler('getById', {
queryOne: (elementOrDocument, selector) => {
return elementOrDocument.querySelector(`[id="${CSS.escape(selector)}"]`);
// Note: for demonstation perpose only `id` should be page unique
queryAll: (elementOrDocument, selector) => {
return elementOrDocument.querySelectorAll(`[id="${CSS.escape(selector)}"]`);

You can now use it as following:

const node = await page.waitForSelector('::-p-getById(elementId)');
// OR used in conjunction with other selectors
const moreSpecificNode = await page.waitForSelector(
'.side-bar ::-p-getById(elementId)'

Custom framework components selector


Be careful when relying on internal APIs of libraries or frameworks. They can change at any time.

Find Vue components by name by using Vue internals for querying:

Puppeteer.registerCustomQueryHandler('vue', {
queryOne: (element, name) => {
const walker = document.createTreeWalker(element, NodeFilter.SHOW_ELEMENT);
do {
const currentNode = walker.currentNode;
if (
currentNode.__vnode?.ctx?.type?.name.toLowerCase() ===
) {
return currentNode;
} while (walker.nextNode());

return null;

Query the Vue component as following:

const element = await page.$('::-p-vue(MyComponent)');

Web Components

Web Components create their own tag so you can query them by the tag name:

const element = await page.$('my-web-component');

Extend HTMLElementTagNameMap to define types for custom tags. This allows Puppeteer to infer the return type for the ElementHandle:

declare global {
interface HTMLElementTagNameMap {
'my-web-component': MyWebComponent;