If ChatGPT or Codex suddenly can’t open your pages, a hidden safety check may be why. OpenAI Aura is the name for a central check that decides whether OpenAI’s browsing tools may load a given site, according to a Search Engine Journal report published October 29, 2026.
It’s worth knowing about, but the evidence is thin. It comes from one case study, and OpenAI hasn’t publicly named the system.
Here are the quick facts:
- Aura is a safety check used by ChatGPT live search and Codex’s Browser Use, per the report.
- A flagged site is refused even if robots.txt allows it and the user’s browser setting says “Always allow.”
- Blocks are cached locally for 24 hours, and there’s no clear appeal path.
- The only documented case is one unnamed international publisher.
Treat this as a heads-up, not a confirmed wave of blocked sites.
What OpenAI Aura actually is
The author, Dan Taylor, found the name in network traffic. ChatGPT calls a status endpoint, chatgpt.com/backend-api/aura/site_status, before it opens a page. OpenAI’s public help pages don’t mention Aura by name.
What the help pages do say is that browsing is blocked on sites linked to illegal activity, using a global blocklist for security, safety and compliance. So the idea of a blocklist is official. The details of how Aura works are not.
One detail matters for site owners. If the status check fails because of a network glitch, the report says it defaults to allowing access. An explicit block, though, is strict.
The case that started the story
The case study begins in May 2026. ChatGPT kept failing to load one publisher’s pages, and weeks later all five tested URLs were still blocked.
The publisher’s robots.txt was healthy and crawler traffic looked normal. Server logs showed no ChatGPT requests at all, which means the refusal happened before any request left OpenAI’s side. Competitor sites loaded fine in the same sessions.
Support later confirmed in writing that a safety block was active and under internal review. Staff didn’t say why.
The author also notes that the publisher ran an ad campaign with OpenAI and that he advised the publisher. That’s useful context when you weigh the report.
What’s confirmed and what isn’t
Most of the claims here come from a single case. This table separates what was observed from what is still a guess.
| Claim | Status |
|---|---|
| Aura exists as an internal safety check for ChatGPT and Codex browsing | Reported from network traffic, not named by OpenAI |
| A flagged site can be refused even with robots.txt and browser settings open | Observed in one publisher case |
| Blocks are cached for 24 hours | Reported by the author |
| Why that publisher was flagged | Unknown |
| The prompt-history theory | Unverified |
| How many sites are affected | Unknown |
The prompt-history theory is the most speculative. Some researchers think status checks, which send chat details along with links, could cause site-wide blocks tied to what users asked. Nobody has verified that.
How to check if ChatGPT can reach your site
You can test this in about ten minutes. Start with the pages an assistant would most likely cite, such as your homepage, docs, pricing and contact page.
- Ask ChatGPT live search and Codex to open each URL, and note the exact error and the time.
- Check your server, CDN and firewall logs. If no request arrived, your own rules probably aren’t the cause.
- If requests did arrive, look for 403, 429 or 5xx responses, redirects, geo-blocks and bot challenges.
- Test a few competitor pages in the same session for comparison.
That last step shows whether the problem is your site or the tool. If only your domain fails and your logs are clean, the block is likely on OpenAI’s side.
What to do if your site is blocked
Save your evidence first: error messages, timestamps with time zones, and log screenshots. Then open a support case with OpenAI and ask whether your domain has a site-safety classification and how to request a review.
Don’t loosen your security or whitelist all AI traffic in a panic. Narrow fixes tied to evidence are safer. The report also warns against trying to bypass the block, since OpenAI’s usage policies treat that as a violation.
It helps to retest key URLs monthly. A site being “up” no longer guarantees an AI agent can read it. If you care how OpenAI’s agents are watched for other problems, our piece on OpenAI’s AI monitors covers a different side of that safety work, and the OpenAI Model Spec rewrite explains the company’s wider rules.
Frequently Asked Questions
What is OpenAI Aura?
It’s the name OpenAI’s systems use for a safety check that decides whether ChatGPT and Codex browsing tools can open a website. OpenAI hasn’t named it in public help pages, so the name comes from the Search Engine Journal report.
Does robots.txt stop Aura from blocking my site?
No. In the reported case, the block applied even though robots.txt was healthy and the user’s browser was set to “Always allow.”
How do I know if ChatGPT is blocking my site?
Try opening your URLs in ChatGPT, then check your server logs. If the tool refuses the page and your logs show no ChatGPT request, the block likely happens before the request is sent.
Is my site probably affected?
Probably not. The report documents one case and offers no estimate of how many sites are blocked.
Can I appeal an Aura block?
There’s no clear public appeal process. The report’s advice is to open a support case with evidence and ask for a review.
Our take
Don’t panic, but do a quick test. One anonymized case, from an author with ties to the publisher, is not proof of a widespread problem. Still, the fix costs you ten minutes, and finding a silent block early beats discovering it after your pages quietly vanish from AI answers. Run the check once a month and keep your evidence ready.
Leave a Reply