Which AI Crawlers Can Read JavaScript Websites?

  • Googlebot and Bingbot have established JavaScript-rendering capabilities.
  • Apple documents browser-based rendering for Applebot.
  • Recent testing shows GPTBot can fetch JavaScript without necessarily executing it.
  • Recent studies have not demonstrated JavaScript execution by the major Anthropic and Perplexity crawlers.

For a long time, the common understanding was that Googlebot could render JavaScript, while most AI crawlers processed only the HTML returned by a server. However, new research published in 2026, along with recent server-log observations, has added some interesting details.

For example, a November 14, 2026 investigation by JS SEO Lab found that GPTBot downloaded JavaScript files but did not execute the React code on its test website. The same investigation found no JavaScript or CSS requests from OAI-SearchBot during its observation period. Meanwhile, other developers have reported different behaviour from OpenAI crawlers on their own websites.

So, can AI crawlers actually read JavaScript websites?

The answer depends on the crawler, the type of request, and what we mean by reading JavaScript. Downloading a JavaScript file, executing it, and retrieving the content generated by it are three different things.

This article examines recent crawler experiments, real server-log observations, and website-rendering studies to understand what crawlers can actually access.

Check this guide to learn how AI Crawlers see your website.

First, understand the three levels of JavaScript access

Before comparing crawlers, we need to separate three different activities.

A crawler can access a JavaScript website without necessarily understanding the content generated by JavaScript.

Level 1: Fetching JavaScript files

A crawler requests a JavaScript file from the server.

For example:

GET /assets/app.js
GET /assets/main.bundle.js
GET /js/product-page.js

This confirms that the crawler downloaded the files. It does not confirm that the crawler executed them.

Level 2: Executing JavaScript

A crawler runs the downloaded JavaScript using a browser engine or another JavaScript runtime.

For example, the JavaScript might:

  • Build the page content.
  • Load product information from an API.
  • Insert headings and paragraphs.
  • Generate internal links.
  • Add structured data to the document.

A crawler that executes this code may be able to access the resulting content.

Level 3: Retrieving JavaScript-generated content

This is the most important level for SEO.

Consider a React application that initially returns:

<div id="root"></div>
<script src="/assets/app.js"></script>

After JavaScript executes, the browser displays:

<div id="root">
  <h1>Best Technical SEO Tools</h1>
  <p>Compare technical SEO tools for website auditing.</p>
  <a href="/tools/site-audit/">Site Audit Tool</a>
</div>

The question is whether the crawler can access the heading, paragraph, and link.

A JavaScript file request alone cannot answer that question.

My recommendation: When analysing crawler behaviour, look for evidence of execution and content retrieval, not simply requests for .js files.

What does the latest research say about AI crawlers?

Several studies published in 2026 provide useful evidence.

Here are the most relevant investigations.

Case Study 1: JS SEO Lab’s November 2026 crawler investigation

Published November 14, 2026. Research by Jakub Sawa.

This is one of the most interesting recent investigations because it examines actual server requests from verified OpenAI crawler IP addresses.

The researcher monitored two websites, including a Next.js test environment, between October 23 and November 14, 2026.

The investigation separated OpenAI’s crawler identities rather than treating all OpenAI traffic as a single group.

Here are the reported observations:

OpenAI crawlerObserved behaviour
GPTBotDownloaded JavaScript files
OAI-SearchBotDid not request JS or CSS in this test
ChatGPT-UserMostly fetched HTML

The researcher also installed execution beacons in the test environment. These beacons were designed to report when inline JavaScript and React client-side code actually executed.

The important finding was that GPTBot downloaded JavaScript bundles, but the React code did not execute.

This provides a useful real-world example of why asset requests should not be confused with rendering.

What about OAI-SearchBot?

The researcher observed 560 page requests from OAI-SearchBot but no JavaScript, CSS, or API requests on the monitored site.

However, OAI-SearchBot did not visit the JavaScript test pages during the observation period.

Therefore, this investigation does not prove that OAI-SearchBot cannot render JavaScript. It shows that rendering was not demonstrated by the observed requests.

Read the investigation: GPTBot downloads your JavaScript. It does not run it.

Case Study 2: The 41-day AI crawler experiment

Published August 19, 2026 by Vinicius Stanula

This experiment examined whether AI crawlers could discover pages linked only through JavaScript.

The researcher monitored a website containing approximately 2,400 pages.

The test included different page-discovery methods:

  • Pages linked through normal HTML anchors.
  • Pages linked through JavaScript-generated navigation.
  • Pages accessible through different internal linking structures.

The most important observation was that six crawlers failed to discover 293 pages during the test when those pages were accessible only through JavaScript-injected links.

After the researcher replaced the JavaScript-only links with ordinary HTML links, GPTBot discovered 250 previously inaccessible pages in less than three hours.

However, remember that this experiment tests link discovery. It does not establish that every crawler is incapable of executing every type of JavaScript.

Read the experiment: JavaScript links can make your pages invisible to AI search.

Study 3: QuanticData’s analysis of 800 websites

QuanticData compared the initial HTML responses and fully rendered browser output of 800 popular domains.

Of those, 374 websites returned usable HTTP 200 responses in both tests and were included in the analysis.

The findings included:

FindingResult
Websites with readable content only after JavaScript6.4%
Websites losing more than half their words without JavaScript17.4%
Websites losing more than 80% of their words11.5%
Websites serving a different page title13.4%

One particular example was Reddit, which returned no readable words in the non-JavaScript capture but displayed 376 words after browser rendering.

The study compared website responses. It did not independently verify the execution capabilities of every AI crawler.

Read the study: Do AI Crawlers Render JavaScript? What 800 Top Sites Reveal.

Case Study 4: Suff Digital’s analysis of 384,588 websites

Published September 24, 2026 | Based on HTTP Archive data

This study examined a much larger sample of websites.

Researchers compared the initial HTML response with the browser-rendered version of 384,588 homepages.

The findings were particularly interesting for websites built using JavaScript frameworks.

FindingJavaScript framework websitesOther websites
Fewer than 200 readable characters before JS29.3%14.4%
H1 appears only after JavaScript17.5%6.1%
Median characters available before JS2,9663,829
Structured data added only through JS7.2%3%

This suggests that using a JavaScript framework does not automatically make a website inaccessible. However, framework-based websites were more likely to hide important information until JavaScript executed.

The study used HTTP Archive data and compared page representations. It was not a direct test of individual AI crawler execution.

Read the research: JavaScript Home Pages vs AI Crawlers.

What do these studies collectively tell us?

Three findings:

  1. Downloading JavaScript does not prove that a crawler executes it.
  2. JavaScript-generated links can create URL discovery problems.
  3. A website can return a successful HTTP response while delivering very little readable content before JavaScript runs.

Another important development: some recent observations suggest OpenAI’s crawling behaviour may be changing. We need to examine that separately rather than relying on older assumptions.

Which AI crawlers can read JavaScript?

Only Googlebot, Bingbot, and Applebot reliably execute and read JavaScript. AI crawlers from OpenAI, Anthropic, and Perplexity only download raw static HTML and cannot see content generated dynamically by client-side JavaScript.

CrawlerJavaScript execution evidenceWhat we can conclude
GooglebotOfficially documentedSupports rendering
BingbotSupported by Microsoft’s rendering infrastructureSupports rendering
ApplebotApple documents browser renderingMay render
GPTBotRecent execution-beacon testDownloaded JS, execution not observed
OAI-SearchBotConflicting recent observationsNot conclusively established
ChatGPT-UserRecent log observationsJS execution not demonstrated
ClaudeBotRecent studies report no executionNot demonstrated
Claude-SearchBotLimited direct testingUnknown
Claude-UserLimited direct testingUnknown
PerplexityBotRecent studies report no executionNot demonstrated
Perplexity-UserLimited direct testingUnknown
Meta-ExternalAgentRecent studies report no executionNot demonstrated
CCBotRecent studies report no executionNot demonstrated

Note: This table is deliberately conservative. A crawler should not be labelled JavaScript-capable simply because it requests JavaScript files.

Googlebot

Googlebot supports JavaScript rendering through Google’s Web Rendering Service. Google uses an evergreen Chromium-based rendering environment to process JavaScript pages. Its process includes crawling, rendering, and indexing.

This is different from a crawler that only downloads the original HTML response.

Official documentation: Google JavaScript SEO basics.

Bingbot

Microsoft has documented its use of the evergreen Microsoft Edge rendering engine for Bingbot.

This means Bing can process modern JavaScript websites.

However, successful rendering still depends on accessible resources, working scripts, and the website’s implementation.

Applebot

Apple’s documentation states that Applebot may render webpage content.

It can process resources such as JavaScript, CSS, and AJAX requests.

This makes Applebot another important crawler to consider when comparing browser-rendering capabilities.

Official documentation: About Applebot.

GPTBot

GPTBot is used by OpenAI to collect web content that may be used for model training.

The November 2026 JS SEO Lab investigation observed GPTBot downloading JavaScript files without executing the React code in its test environment.

However, it is still one monitored environment, not a universal test of every GPTBot request.

OAI-SearchBot

OAI-SearchBot is particularly important because it helps surface websites in ChatGPT Search.

Its behaviour deserves separate attention.

In September 2026, some developers reported increased JavaScript-related requests from OpenAI crawlers, including requests involving JavaScript assets and Next.js related resources.

In November, JS SEO Lab reported a different result on its monitored properties: OAI-SearchBot requested pages but did not fetch JavaScript, CSS, or API resources.

Note: These observations are not necessarily contradictory. They come from different websites, test environments, and observation periods. They also do not establish whether OAI-SearchBot executes JavaScript when it encounters a different application.

My conclusion: OAI-SearchBot’s JavaScript-rendering capability remains an open question based on the publicly available evidence reviewed here. Do not assume it behaves exactly like ChatGPT-User or GPTBot.

Official documentation: OpenAI crawler documentation.

ClaudeBot and Claude-SearchBot

Anthropic uses separate crawlers for different purposes:

  • ClaudeBot: Model training.
  • Claude-SearchBot: Search-related crawling.
  • Claude-User: User-requested page retrieval.

Recent research has reported no JavaScript execution by ClaudeBot.

However, this does not establish the rendering behaviour of every Anthropic crawler.

Anthropic’s documentation explains the purpose of these crawler identities but does not promise JavaScript rendering.

Official documentation: Anthropic’s crawler information.

PerplexityBot

PerplexityBot is used for crawling and discovering web content for Perplexity’s search experience.

Recent studies report that it processes initial HTML rather than executing JavaScript.

However, its user-requested retrieval process should not automatically be assumed to work identically.

Official documentation: Perplexity crawler documentation.

So, if I had to SEO a JavaScript-heavy website, I would not start by telling the development team to remove JavaScript. I would work on JavaScript SEO.

FAQs

Do AI crawlers execute JavaScript?

It depends on the crawler. Googlebot and Bingbot support JavaScript rendering. Recent tests have not demonstrated JavaScript execution by several major AI crawlers, while OpenAI’s newer crawler behaviour remains under investigation.

Does GPTBot download JavaScript files?

Yes. A November 2026 JS SEO Lab investigation observed GPTBot downloading JavaScript bundles. However, its execution beacon did not detect React code execution.

Is OAI-SearchBot now rendering JavaScript?

There are reports of increased JavaScript-related requests from OpenAI crawlers, but publicly available tests do not yet establish consistent JavaScript rendering by OAI-SearchBot.

Can ChatGPT read React websites?

ChatGPT may access content from React websites through its search index or user-requested retrieval. Whether client-side content is available depends on the retrieval process and what content the relevant crawler or system receives.

Does fetching JavaScript mean a crawler can render a website?

No. Fetching downloads the JavaScript file. Rendering requires executing the code and processing the resulting page.

How can I check whether GPTBot executes JavaScript?

Monitor verified GPTBot requests in your server logs. Use a test page with JavaScript execution beacons and check whether the crawler triggers them.

Can AI crawlers discover JavaScript-generated links?

This has not been demonstrated consistently. A 2026 experiment found that six tested crawlers failed to discover 293 pages linked only through JavaScript during the observation period.

Does Googlebot render JavaScript?

Yes. Googlebot uses Google’s Web Rendering Service to execute JavaScript and process the rendered page.

If you have any questions, feel free to comment or connect with me on Instagram and LinkedIn.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top