News & analysis
OpenAI's Aura Check Can Block ChatGPT From Your Travel Site
Search Engine Journal reports that OpenAI's internal Aura safety check can stop ChatGPT's browsing tools from opening a website entirely, ignoring robots.txt and local settings, with no public appeal process.
By Roomer Editorial · · 3 min read
What is OpenAI's Aura safety check?
Aura is OpenAI's internal system for deciding whether ChatGPT's browsing tools and Codex's Browser Use are allowed to open a given link, according to Search Engine Journal's reporting. It sits separately from OpenAI's named crawlers like GPTBot and OAI-SearchBot, which publish IP addresses and follow standard web rules.
If Aura flags a site, browsing tools refuse to load it at all. Local settings and robots.txt files don't matter. The decision is checked against an API endpoint and cached for 24 hours, with no public way to find out why a site was flagged or ask for a review.
OpenAI's help pages say agents may be blocked for security, safety, and compliance reasons using a global blocklist, and that sites linked to illegal activity are blocked. The report notes OpenAI hasn't explained the technical workings behind the system publicly.
How did a publisher end up blocked, and what did OpenAI say?
The report describes an international media publisher, distributing news and magazine content with fully open robots.txt rules, that found ChatGPT repeatedly failing to load its main site starting in May 2026. Server logs showed zero incoming traffic from ChatGPT, meaning the block happened before any request reached the publisher's servers.
Tests with Codex CLI and desktop apps returned an explicit rejection message citing site-safety policy as the reason, with no user permission prompt offered. The same chat sessions had no trouble opening competing news sites.
Traffic data showed OpenAI's training crawlers made roughly 5,000 successful connections to the site over 30 days, compared to 477,000 from Google and 564,000 from Anthropic in the same window. OpenAI support staff eventually confirmed in writing that a safety block was active and that specialist teams were reviewing whether it was intentional, without ever naming Aura directly.
How can a site owner tell if they're blocked by Aura?
Compare behavior across tools: if ChatGPT and Browser Use refuse to open your site while server logs show no incoming agent traffic at all, the block is happening inside OpenAI's system, not on your end. The report recommends ruling out your own firewall and blocklist settings first, then building a documented pack of failed requests and clean logs before contacting support, since there's no self-service tool to request a review.
Roomer Editorial's take
A travel creator's independence rests on being reachable — by readers, by search, and increasingly by AI agents pulling live information to answer a question like "best hotel in Lisbon's Alfama." If Aura mislabels your domain, none of that matters. The agent doesn't see an error about your content; it just moves to the next source, and you never find out.
This is a separate problem from AEO or structured content work. You can have clean entity markup, open robots.txt, and solid citations, and still be invisible to ChatGPT's browsing tools because of a backend classifier you can't query or appeal. The source makes clear this isn't theoretical — a publisher with legitimate traffic and healthy crawler activity got flagged anyway.
Roomer Editorial's view: creators running their own site built from their videos should treat this as a reason to diversify where their bookable pages get discovered, not a reason to abandon direct booking. Your commissions on your own site, and the booking rails through Vio, Stay22, GetYourGuide, Viator and Klook, don't depend on any single AI tool finding you. If ChatGPT can't open your page today, a traveler searching Google — where roughly a million Roomer pages are already indexed — still can.
The exception: if a meaningful share of your traffic already comes from ChatGPT referrals, a silent Aura block is worth escalating immediately, using the comparison-testing method the source lays out, before you spend time or budget chasing a visibility fix that can't work because the access problem sits upstream of content entirely.
One action: run the same query about your site in ChatGPT's browsing mode and in a plain Google search side by side this week, and if ChatGPT refuses to open your domain while competitors load fine, start building a log-based case rather than optimizing content you already know is solid.
Source: Search Engine Journal — OpenAI's Aura Safety Check Can Block Your Site From ChatGPT Browsing
