TL;DR

Answer
Most major AI crawlers fetch JavaScript files but do not run them. Googlebot does.
Bots
OpenAI, Claude and Perplexity crawlers did not execute JavaScript in one large study. Google and Bing render.
Check
Compare view-source with the rendered page, fetch with a bot user agent, then run the free audit.
Fix
Put key facts in server-rendered or static HTML. Keep client rendering for extras.
Gemini and ChatGPT
Gemini uses Googlebot’s rendering. ChatGPT search may use third-party providers such as Bing.
Caveat
The main data is one study, and Anthropic does not document rendering. Confirm in your own logs.

Most AI crawlers cannot render JavaScript. In the largest published study of their behavior, the crawlers behind ChatGPT, Claude and Perplexity downloaded JavaScript files but did not run them, while Googlebot did. If your key content only appears after scripts run, those bots may see an empty page.

This guide gives a per-bot table, explains what it means for React and Next.js sites, and shows a 10 minute check of what AI crawlers see on your page. It ends with the fixes and a checklist.

1Do AI crawlers render JavaScript?

Mostly no, with one big exception. Here is what each crawler does and where the claim comes from. The behavior data comes from a study by Vercel and MERJ (The rise of the AI crawler). The operators’ own documentation is quoted where it says anything.

Bot Used for Runs JavaScript? Source
GPTBot OpenAI model training No Vercel and MERJ study. OpenAI’s bot documentation does not address rendering
OAI-SearchBot ChatGPT search results No Vercel and MERJ study
ChatGPT-User Pages opened on a user’s request No Vercel and MERJ study
ClaudeBot Anthropic model training No, in the study. Not confirmed by Anthropic Vercel and MERJ study. Anthropic’s crawler page does not address rendering
PerplexityBot Perplexity search results No Vercel and MERJ study. Perplexity’s bot page does not address rendering
Googlebot Google Search, also used by Gemini Yes Google’s JavaScript guide, and the study for Gemini
Bingbot Bing Search Generally yes, with limits Bing webmaster blog
Applebot Apple search features Yes Vercel and MERJ study

Quotable stat

In Vercel and MERJ’s data, 11.50% of ChatGPT crawler requests and 23.84% of Claude crawler requests were for JavaScript files, yet neither crawler ran them (source).

Read the table with care. The study measured real traffic, mostly on one large site, so treat it as strong evidence and not a guarantee. Bot behavior changes, and the operators do not document rendering, so confirm what you can in your own logs. ClaudeBot is the least settled, because Anthropic says nothing about rendering and the study is the only evidence I could verify.

Two details help. First, a crawler that fetches a JavaScript file has not necessarily run it. Second, content that is already in the first HTML response is readable even by bots that never run a script, which is the whole basis of the fix below.

2Do ChatGPT, Gemini and Perplexity see JavaScript content?

ChatGPT and Perplexity crawlers did not execute JavaScript in the study, so client-rendered content is at risk with them. Gemini is different. The study reports that Gemini uses Googlebot’s rendering infrastructure, so a page Google can render can be read for Gemini too. Our guide to ranking in Gemini explains how Gemini finds sources.

ChatGPT search is less clear. OpenAI says ChatGPT may share search queries with third-party providers such as Bing (OpenAI help center), and it does not say whether it also runs its own index. Bing renders JavaScript in many cases, so check how your pages look in Bing as well. Our guide to ranking in ChatGPT covers the rest.

3Which sites are at risk: client-side, server-side or static?

The risk depends on where your content is built. If the facts are in the first HTML the server sends, a bot that cannot run scripts still reads them.

How the page is built Key text in the first HTML? Risk with AI crawlers
Client-side rendering (a single page app that fills an empty shell) No High. A bot that does not run scripts sees little or nothing
Server-side rendering (SSR) Yes Low
Static generation or incremental regeneration Yes Low
Server-rendered text with client-side extras Yes Low, if the extras are not key facts
Dynamic rendering (a prerendered copy for bots) Yes, for bots A workaround, so keep the content identical for bots and users

Google says server-side or pre-rendering is still a great idea because it helps users and crawlers, and not all bots can run JavaScript (Google Search Central). Bing recommends dynamic rendering for JavaScript heavy sites, provided the content is the same for every visitor (Bing webmaster blog). Vercel and MERJ recommend server-rendering critical content, including main text, meta information and navigation.

4How do I check what AI crawlers see on my page?

You can check a page in about 10 minutes with three steps. Do them on the pages that matter most, such as your home, pricing, product and top articles.

1 Compare view-source with the rendered page. Open the page, choose View Page Source, and search for a sentence that matters, such as a product name, a price or a claim. Then search the same sentence in the Elements panel of DevTools. If it is in the Elements panel but not in the source, it depends on JavaScript.
2 Fetch the page like a bot. Request the page without running scripts and search the response, as in the command below. It returns the same raw HTML that a non-rendering bot receives.
3 Run the free audit. The Wellows technical audit fetches each page as raw HTML and again after JavaScript runs, so you can compare raw and rendered HTML and see the gap for every page.
curl -s -A "Mozilla/5.0 (compatible; GPTBot/1.1; +https://openai.com/gptbot)" https://example.com/pricing | grep -i "your key sentence"

If the command prints nothing, the sentence is not in the first HTML. Use the exact user agent string from each operator’s documentation. Some firewalls verify bots by IP address, so a request from your own machine can get a different response from a real crawler. Treat the result as a strong hint, and confirm it in your server logs.

Also check robots.txt

A bot that is blocked sees nothing, whatever the rendering. OpenAI treats its bots separately: OAI-SearchBot is for ChatGPT search, GPTBot is for training, and ChatGPT-User handles actions a user starts, where robots.txt rules may not apply (OpenAI bot documentation). Anthropic lists ClaudeBot, Claude-User and Claude-SearchBot, and says its bots honor robots.txt (Anthropic crawler page).

An illustrative example

A SaaS pricing page is built as a single page app. View Page Source shows an empty div and a script tag. The Elements panel shows three plan names with prices. The curl command finds none of them. For a bot that does not run scripts, this page has no pricing. Moving the pricing text into the server-rendered HTML fixes it, and the same curl command then finds all three plans. This example is invented to show the check, and it is not Wellows data.

5How do I fix a JavaScript site so AI crawlers can read it?

Put every fact you want cited in the first HTML response, and keep client-side code for features that are not key content. The choice between server-side rendering and static generation matters less than the result: the text is there before any script runs.

1 List the facts AI answers should get right: product names, plans and prices, claims, definitions and contact details.
2 Render that text on the server, or build it statically, for every page that carries it.
3 Keep headings, the title, the meta description, the canonical tag and main navigation in the first HTML too.
4 Use real links, meaning anchor tags with an href. Google says only these are reliably discovered, and links built in script may be missed.
5 Return correct status codes and keep redirects, sitemaps and URLs consistent. Vercel and MERJ saw high 404 rates from AI crawlers on one site.
6 Recheck with the three steps above, then watch whether engines start citing the page.

Google also notes that a canonical set by JavaScript must match the one in the original HTML (Google Search Central). For wider technical problems, see our guides to technical SEO issues and the technical SEO checklist for agencies.

What only Wellows adds

The free Wellows technical audit takes a domain and an email address, and checks up to 10 pages that you choose or that it picks. It fetches each page as raw HTML and again after JavaScript runs, and treats the gap between the two as the finding. It also checks whether robots.txt allows GPTBot, ClaudeBot and PerplexityBot, and whether content survives with JavaScript off. You get a score, page by page findings with evidence and fixes.

After you fix a page, Wellows LLM Citations tracks the URLs that ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode cite for your prompts, so you can see whether the page starts to appear. It also tracks implicit citations, the times you are not cited but another page about your topic is. This is your shadow reputation, and it is the narrative that AI engines create based on pages you have not written.

Start by checking what your key pages serve to a bot that cannot run scripts.

Run the free technical audit

A JavaScript and AI crawler checklist

  • Search view-source for each key sentence, price and claim.
  • Fetch key pages with a bot user agent and confirm the text appears.
  • Render key facts on the server, or build them statically.
  • Keep title, headings, canonical and navigation in the first HTML.
  • Link with anchor tags that have an href.
  • Allow the bots you want in robots.txt, and decide on training bots on purpose.
  • Return correct status codes, and fix redirects and broken URLs.
  • Review server logs for AI bot requests, and recheck after each release.

JavaScript and AI crawlers: do and do not

Do Do not
Put key facts in the first HTML response Hide pricing, claims or definitions behind scripts
Test with view-source and a bot user agent Trust a browser view as proof a bot sees the page
Check Google and Bing as well as AI bots Assume one bot’s behavior applies to all
Confirm the claims in your own logs Treat one study as permanent fact
Keep bot and user content identical Serve bots different content from users

Conclusion

Most AI crawlers do not run JavaScript, and Google and Bing are the main exceptions. If your key facts are in the first HTML, every bot can read them, so render them on the server or build them statically. Run the three step check on your most important pages this week, and confirm the result in your server logs.


Mostly no. In the Vercel and MERJ study of real crawler traffic, the bots behind ChatGPT, Claude and Perplexity downloaded JavaScript files but did not execute them, while Googlebot rendered pages fully. Because Gemini uses Googlebot’s infrastructure, a page Google can render can be read for Gemini too.

In the study, GPTBot, OAI-SearchBot and ChatGPT-User did not run JavaScript, and OpenAI’s bot documentation does not address rendering. OpenAI also says ChatGPT may share search queries with third-party providers such as Bing, so check how Bing sees your pages as well as how a non-rendering bot does.

Search the page source for a key sentence, then search the same sentence in the rendered Elements panel in DevTools. If it only appears in the Elements panel, it depends on JavaScript. Next, fetch the page with curl and a bot user agent to see the raw HTML a non-rendering bot receives, then run the free Wellows technical audit to compare raw and rendered HTML across up to 10 pages.

Put every fact you want cited in the first HTML response, using server-side rendering or static generation. Keep the title, headings, canonical tag, main navigation and real anchor links in that HTML too, and leave client-side code for features that are not key content. Then recheck with view-source and a bot user agent.

No. OpenAI treats its bots separately: OAI-SearchBot is used for ChatGPT search results, GPTBot is used for model training, and ChatGPT-User handles actions a user starts. You can allow OAI-SearchBot and block GPTBot, so decide on each bot on purpose in robots.txt.

ai crawlers render javascript