WEIR
Get a key
Theme

Can ChatGPT Read Tweets? What Works in 2026

ChatGPT, Claude and Perplexity mostly can't open x.com links: X sent our test fetchers a 402. What each can do, and how to make ChatGPT read tweets.


Published · Updated · 8 min read

Cover image: Why AI can't open x.com links

ChatGPT can read tweets reliably only when you hand them over, because X blocks the web fetchers that ChatGPT, Claude and Perplexity use.

This guide is for you if you want an AI assistant to read X (Twitter) posts. You get what each assistant can do today, a test we ran, and the two routes that work.

We run Weir, the connector in our example. Paste this URL into ChatGPT or Claude as a custom connector:

https://api.weirapp.io/mcp
402the error x.com sent each AI fetcher
3crawlers X's robots.txt lets in
2Weir credits to read one post

Can ChatGPT read tweets?

Not from a link, most of the time. On its own, ChatGPT reaches the web in three ways, and none gets you X reliably:

  • Search can't use X pages. OpenAI says a site must let its crawler, OAI-SearchBot, in to appear in answers, and X's robots.txt doesn't.
  • Pasted links hit an error. When you paste a URL, an agent called ChatGPT-User fetches it, and x.com answered that name with a 402.
  • The cloud browser is a maybe. On paid plans except Go, ChatGPT Work can open sites in a cloud browser and pause while you sign in.

OpenAI says robots.txt "may not apply" to ChatGPT-User, since you asked for the page. That didn't help in our test, because X's server refused the request anyway.

OpenAI also says what the cloud browser can finish "depends on the website". We haven't tried it on x.com, and it would read X as you, through your own account.

We ran this test on 21 October 2026, on a public post from @XDevelopers:

GET https://x.com/XDevelopers/status/2019881223666233717

Claude Code's WebFetch tool    → 402 Payment Required, no post text
Request named ChatGPT-User     → 402
Request named Claude-User      → 402
Request named Perplexity-User  → 402

We sent the last three ourselves, labeled with each fetcher's published name, so they show how x.com treats that name. The post is public, and you can open it in a browser. The fetchers inside AI apps get the error instead.

Search crawlers hit a different wall. Here are the key lines of X's robots.txt that day:

User-agent: Googlebot
User-agent: Bingbot
Allow: /*?s=
...
User-agent: facebookexternalhit
...
User-agent: *
Disallow: /

Only Google, Bing and one Facebook crawler get rules that let them in. The catch-all rule at the end tells every other bot to stay out, AI search crawlers included.

Can Claude read tweets?

No, not from a link you paste either. Anthropic's help page says that with web search on, Claude can open URLs you give it. It calls this web fetch.

Anthropic's developer docs list two limits that rule out x.com:

  • No JavaScript pages. Web fetch "does not support websites dynamically rendered with JavaScript", and x.com is one.
  • robots.txt counts. A URL that robots.txt blocks comes back as a url_not_allowed error.

Anthropic also says all its bots honor robots.txt, including Claude-User, the one that opens pages for you. X's file shuts out every bot it doesn't name.

Can Perplexity search X?

Not directly. Perplexity's search crawler, PerplexityBot, falls under X's "Disallow: /" rule. Its link fetcher, Perplexity-User, "generally ignores robots.txt rules", Perplexity says, but x.com still answered that name with a 402.

Perplexity does take remote MCP servers: its help center tells you to open Plugins and choose Add custom. We haven't tested Weir there yet.

Here is what you get from the other three, as of October 2026:

Assistant Crawler allowed by X Pasted x.com link Custom MCP connector
ChatGPT No 402 in our test Business, Enterprise, Pro
Claude No 402 in our test Every plan, one on Free
Perplexity No 402 in our test Yes, from Plugins

All three need you to hand them the post.

How to make ChatGPT or Claude read tweets

Pick the route by how often you need X:

  • Paste it when you need one post you're already looking at.
  • Connect it when you want the assistant to find posts, read replies or follow an account.

Pasting works on any plan. OpenAI says image inputs work on Free and paid ChatGPT plans, so a screenshot is fine. Anthropic's upload guide lists PNG and JPEG, pasted straight from your clipboard.

A connector uses MCP, the Model Context Protocol, which lets your assistant call outside tools. Weir's server gives it tools like get_x_post, search_x and get_x_replies. You sign in with OAuth, so you copy no key and need no X account or developer app.

  1. Open the connector settingsChatGPT: chatgpt.com/plugins, then +. Claude: Customize, then Connectors.
  2. Add the serverChoose a custom MCP server or connector and paste https://api.weirapp.io/mcp.
  3. Sign in to WeirApprove the consent screen. Your X account stays out of it.
  4. Ask about a postPaste an x.com link with your question. In ChatGPT, @mention the connector.

The full setup, plan by plan, is in our guides for ChatGPT and Claude.

We pasted the @XDevelopers link that got the 402 and asked how the post did. With Weir connected, Claude made one call on 21 October 2026:

→ get_x_post {"id": "https://x.com/XDevelopers/status/2019881223666233717"}
← 1 post · 2 credits

What came back, trimmed:

{
  "text": "Officially launching X API Pay-Per-Use ...",
  "createdAt": "2026-02-06T21:09:45.000Z",
  "author": { "username": "XDevelopers", "followersCount": 697891 },
  "likeCount": 5596,
  "retweetCount": 896,
  "replyCount": 792,
  "quoteCount": 596,
  "viewCount": 2557918
}

You get the full post with its counts, ready to summarize. It cost two credits: one for the call and one for the post.

Other prompts you can try:

  • "Summarize the replies to this post." The assistant calls get_x_replies.
  • "What did @OpenAI post this week?" It reads the timeline with get_x_user_posts.
  • "Read this X Article in full." get_x_article returns the whole body, as our guide to reading X Articles via API explains.

If you write code, you can call the same data from Weir's REST API instead.

When not to use Weir

  • You only read a post now and then. Pasting the text costs nothing on any plan.
  • Your assistant needs to post or reply. Weir is read-only: no posting, liking or following.
  • Price per post matters most. TwitterAPI.io lists $0.15 per 1,000 tweets and has its own MCP server.
  • Your company needs a deal with X itself. Weir isn't licensed by X, so access can change.

FAQ

Can ChatGPT access Twitter?

Not on its own. X's robots.txt keeps ChatGPT's search crawler out, and x.com refused its link fetcher in our test. You can paste a post's text or screenshot, or add an MCP connector that reads X.

Can ChatGPT search X?

Only with a connector. Its own web search can't, because X's robots.txt keeps OpenAI's crawler out. With an X connector, you can have ChatGPT search posts by keyword, account and date. On Weir, that costs one credit per call plus one per post returned.

Does ChatGPT Plus support an X connector?

OpenAI's help center lists custom MCP servers for Business and Enterprise/Edu on the web, plus Pro with read and fetch tools. It doesn't list Plus or Free, as of October 2026. Claude allows one custom connector on its Free plan.

Is it allowed to have an AI read tweets?

Platform terms matter. X's terms prohibit "crawling or scraping the Services in any form" without its written consent, and this isn't legal advice. Weir reads public posts only, never posts, and isn't licensed by X, so access can change.

Sources

All pages checked 21 October 2026 unless noted.

Keep reading