ChatGPT keeps telling people 404 Media can fix their tech support problems: what users asked, what ChatGPT said, and what OpenAI's own help pages say

ChatGPT has been telling people with ordinary tech support problems to contact 404 Media, the tech news site. On October 26, 2026, 404 Media's Jason Koebler reported that "Over the last few months, 404 Media has gotten dozens of clearly AI-generated story leads and tips," and that when the outlet asked senders how they'd found it, the answer kept pointing back to an AI chatbot. The problems were things like a banned Instagram account or a phone that wouldn't connect, and ChatGPT had told users these were stories a journalist would want. It's a different kind of chatbot error from the nonexistent case cited in the AI-assisted court order that got a federal judge removed from a case on October 22: a confident, wrong answer about where to get help.

This story is based on 404 Media's October 26, 2026 article and its companion piece, OpenAI's help-center articles on accuracy and contacting support, and OpenAI's September 5, 2025 research post on why language models hallucinate, checked October 27, 2026.

The short version

• What happened: people asked ChatGPT for help with account bans, phone service and other support problems, and it pointed them to 404 Media, sometimes drafting the email for them.
• How often: "dozens" of clearly AI-generated tips over the last few months, according to 404 Media, several in the days before it published. It gave no exact count.
• What ChatGPT said: in one exchange, "I think contacting a journalist is a reasonable and proportionate response"; in another, it offered to "Draft an email to Jason Koebler at 404 Media."
• OpenAI's response: none in 404 Media's article, which doesn't say whether OpenAI was asked.
• 404 Media's answer: "To be clear, we are not Meta's technical support team, nor Verizon's, nor OpenAI's."

What did people ask ChatGPT?

How to fix problems that big platforms' support channels hadn't. 404 Media described several cases:

• Instagram: "Last week, I got an email from a man in Russia whose Instagram account had been banned," Koebler wrote. He had asked ChatGPT how to get it back, and it eventually suggested contacting 404 Media.
• Meta, again: an Italian painter wrote that he "loses professional Meta accounts despite contradictory reinstatement email," and said ChatGPT had suggested contacting Koebler.
• Verizon: one person emailed 404 Media and many other outlets asking for an "independent investigation" into why their Verizon service wasn't working on their iPhone.
• OpenAI itself: a man complained that his ChatGPT-based assistant, "Captain NORBERTA," was unreliable. "This email was itself written by Captain NORBERTA and came with three PDFs attached," Koebler wrote, and the man said ChatGPT had recommended 404 Media.
• Not a support problem: one person who had argued with ChatGPT about racism in a Trump speech was told the exchange was newsworthy. 404 Media said that email wasn't written by AI.

What did ChatGPT tell them?

That their problem was a story, and that 404 Media was the place to send it. In the examples 404 Media published, ChatGPT summarized the outlet as "404 Media Focus: Technology, digital platforms, surveillance, privacy, and technology-related investigations." It told the Instagram user, "You have lost access not merely to a social media profile." In the racism conversation it said, "I think contacting a journalist is a reasonable and proportionate response," and offered to "Draft an email to Jason Koebler at 404 Media."

The AI-written emails that followed had a recognizable style, Koebler wrote: senders described compiling "evidence packets," "forensic trails" and "evidence dossiers," which aren't what real tips usually look like. The emails reached 404 Media from ChatGPT "or other chatbots," the article says, though every named example involves ChatGPT.

How often is this happening?

404 Media's only figure is "dozens" over "the last few months." "Around two weeks ago, we started emailing these people back and asking them how they found us," Koebler wrote, and the answer consistently traced to an AI chatbot. Several of the examples arrived in the days just before the October 26 article. That's one newsroom's inbox, not a measurement of how often ChatGPT gives this advice. Koebler noted that New York Times reporter Kashmir Hill "has written about something similar, with chatbots telling users to contact her."

Why does ChatGPT send people to journalists?

404 Media's explanation is about the platforms, not the model. Koebler wrote that big tech companies offer little or no customer support, so people with routine problems turn to ChatGPT; companies tend to act when there's bad press; and chatbots have picked up that pattern and now automate it, telling users their problem is newsworthy. The article says this is happening "with AI further automating away tech support that already barely existed," and that people "seemingly feel like they have nowhere else to turn."

OpenAI hasn't explained this behavior. Its help-center article on accuracy says ChatGPT can produce "Fabricated quotes, studies, citations or references to non-existent sources," and warns, "Sometimes, it might sound confident—even when it's wrong." In a September 5, 2025 research post, OpenAI argued that hallucinations persist partly because "standard training and evaluation procedures reward guessing over acknowledging uncertainty." Recommending a journalist isn't a fabricated fact, but it is the same failure the help page describes: a confident answer to a question the model can't actually resolve. In Chroma's July 2025 long-context study, covered in our recall explainer, GPT models were the most likely to give a wrong answer from a distractor rather than abstain, while Claude models tended to abstain.

What has OpenAI said?

Nothing in 404 Media's article: it contains no OpenAI statement, and it doesn't say whether 404 Media asked OpenAI for comment. 404 Media published a second, members-only piece the same day, "ChatGPT Says 404 Media Is Tech Support for Huge Corporations," which couldn't be read in full for this story. 404 Media has reported on ChatGPT before, including on how its ads are targeted.

Where should you actually go for tech support?

To the company whose product it is, through its own support channel. For ChatGPT and other OpenAI products, OpenAI's help center says to start a chat by "opening the chat bubble icon displayed at the bottom-right of help.openai.com." The chat starts with a virtual assistant, and OpenAI says a human agent may join if the bot can't resolve the issue. OpenAI's accuracy page also advises users to "Always verify quotes, data, technical information or references to external documents" that ChatGPT gives them, which applies to contact details and referrals too.

404 Media says most support problems on their own aren't news, though it's still willing to look at tips about issues of real public concern. If you use an AI tool with web search to find support contacts, check the answer against the company's own site; our ChatGPT vs Perplexity comparison covers how each shows its sources and what citation studies found.

What's confirmed, reported and uncertain

• Confirmed (404 Media, first-hand): dozens of clearly AI-generated tips over the last few months; senders saying a chatbot pointed them there; the Instagram, Meta, Verizon and "Captain NORBERTA" examples; the ChatGPT quotes it published.
• Confirmed (OpenAI): its help-center warnings about confident wrong answers; its support channel; its 2025 research post on hallucinations.
• Uncertain: how often ChatGPT gives this advice across all users; whether other outlets see the same; why ChatGPT does it; whether OpenAI was asked to comment or plans any change.