# Can ChatGPT read X (Twitter) replies?

No. In our test on November 15, 2026, ChatGPT could not read the replies under a public X post, and both times it described the post itself wrongly. We sent the same public X post twice and asked for the 3 most common complaints with one quoted comment each. It said it would not make up replies, but it did make up what the post was about, and differently each time. The two answers share nothing. If you only skim the first lines, you would not notice.

Tested 2026-11-15. Post: https://x.com/WallStreetApes/status/2042095685575831769

## What ChatGPT said

- Answer 1: it could not read the replies (“the X page returned 403”), then said the post was Wall Street Apes sharing a claim that California's governor ran bot farms.
- Answer 2: same 403, but now the post was about Crumbl Cookies: “one cookie has as much sugar as 11 Krispy Kreme donuts and 69 ingredients.”

## What the post actually has

- @WallStreetApes (1,046 likes): A US Federal Judge dismissed a proposed class action lawsuit accusing the maker of ‘Stanley tumblers’ of hiding the popular water bottles contained lead “The manufacturer has acknowledged a small quantity of lead is used” These are very popular drink tumblers. Be careful

The actual post, posted April 9, 2026. It is about Stanley tumblers and lead, not about bots or cookies.

## Why it fails

ChatGPT said the post page returned an error (403) and that search and mirror sites showed no replies. With only fragments to go on, it filled in the post itself differently each time.

## What Buzzsift does instead

Our connector searches X posts that name a brand and returns each post with likes, date and a link. It does not read the replies under one chosen post: it finds the posts people write about the brand.

find_mentions: 100 points a search, up to 150 posts and comments. A full report: $9.99. Set up: https://sociallisteningai.com/claude-chatgpt-connector/

## Questions

**Can ChatGPT read X (Twitter) replies?** No. In our test on November 15, 2026, ChatGPT could not read the replies under a public X post, and both times it described the post itself wrongly. Both answers are quoted on this page.
**How did you test it?** On 2026-11-15 we opened a new ChatGPT chat twice and sent the same request for a public X post: "Read the comments under this post and list the 3 most common complaints, quoting one comment for each." (in Chinese; quotes here are translated). The post is the one in the Stanley sample report on this site.
**Can I still analyze X comments with ChatGPT?** Yes, if the comments reach it as text. Paste them in, or connect Buzzsift to ChatGPT or Claude so the assistant fetches them itself (100 points a search, $1; a new account gets 100 points).
**Will this change?** It may. The test is dated 2026-11-15; chatbots and platforms change their access often, and we will rerun it.