Downloads content for AI model training without direct attribution.
Recommended action: Review robots.txt policy and decide if training access is acceptable.
Category
AI Data Scraper
Primary use case
AI model training
Trust level
Review recommended
Trust Levels
- Trusted
- Generally safe
- Review recommended
- Caution advised
Trust levels are an indication based on category, operator, and robots.txt compliance. Always review bot activity for your specific situation.
Learn how we assess trustrobots.txt
Respected
GPTBot Traffic (Last 90 Days)
What is GPTBot?
GPTBot is OpenAI's web crawler that collects data from publicly accessible web pages to improve AI models like ChatGPT. Site owners can control access via robots.txt.
What GPTBot means for your site
GPTBot downloads your content to include in datasets used to train AI models, operated by OpenAI. Your text becomes part of the AI's general knowledge, but without direct attribution or links. This is a key distinction: training crawlers take your content, AI assistants cite it. You can control training access via robots.txt without affecting citations.
What should you do?
- Decide whether you want OpenAI to train on your content
- Block via robots.txt if unwanted: User-agent: gptbot / Disallow: /
- Monitor crawl patterns for unexpected spikes
- Review BotSights data to see which pages are targeted
See GPTBot on your own site
BotSights tracks every GPTBot visit in real time, including which pages it crawls, how often, and from where.
How to identify GPTBot
GPTBot uses the user-agent "gptbot" and respects robots.txt. It crawls broadly and systematically, often downloading full page content.
gptbotGPTBotHow to block GPTBot
Three robots.txt options below. Pick the one that matches your goal. Each snippet lists every known GPTBot user-agent pattern so the rules apply regardless of which one the bot announces.
Edit robots.txt with care
A single misplaced line can de-index your entire site. Common mistake: pasting User-agent: * followed by Disallow: / blocks every bot, not just GPTBot, including Googlebot. Always paste the snippet between existing rules (not over them), keep the User-agent line scoped to GPTBot's patterns, and verify with Google's robots.txt tester before deploying. If you are not sure, ask a developer first.
Option 1: Block all access
Tells GPTBot not to crawl any URL on your site. Use this when you want the bot completely off your content.
User-agent: gptbot
User-agent: GPTBot
Disallow: /Option 2: Block specific paths only
Keep public content crawlable but exclude sensitive or non-public sections. Add one Disallow: line per path. Replace the example paths with your own.
User-agent: gptbot
User-agent: GPTBot
Disallow: /admin/
Disallow: /private/
Disallow: /checkout/Option 3: Slow down with a crawl delay
Crawl-delay is a voluntary directive that asks the bot to wait the given number of seconds between requests. Useful when GPTBot is hammering your origin and slowing the site down for real visitors, but you do not want to block it outright. The value is in seconds, so 10 means at most one request every ten seconds. Not all bots honour this directive (Googlebot ignores it; Bingbot, Yandex, and many AI crawlers do respect it).
User-agent: gptbot
User-agent: GPTBot
Crawl-delay: 10Frequently Asked Questions
What is the User-Agent for GPTBot?
GPTBot identifies itself with the User-Agent string "gptbot" (alternate forms: GPTBot). Use this exact string in robots.txt rules to control access.
Can I stop GPTBot from using my content for AI training?
Yes. Add this to your robots.txt: User-agent: gptbot / Disallow: /. OpenAI commits to respecting robots.txt for training data.
Will blocking GPTBot affect my AI citations?
No. GPTBot is a training crawler, separate from real-time AI assistants. For example, blocking GPTBot does not block OpenAI's user-prompt assistants from citing your content live.
What's the difference between GPTBot and an AI assistant bot?
GPTBot crawls broadly to build training datasets — your content becomes part of the model's general knowledge but without direct attribution or links. AI assistant bots (like ChatGPT-User, Claude-User) fetch specific pages in response to user prompts and cite sources back. They use separate User-Agents and can be controlled independently.
How do I verify that a request is really from GPTBot?
User-Agent alone is not enough — anyone can claim to be GPTBot. OpenAI publishes IP ranges at openai.com/gptbot.json so you can verify the source IP. BotSights flags spoofed traffic automatically.
Is my content being used without permission?
Training crawlers collect publicly accessible content. The legal landscape around this is rapidly evolving (lawsuits in the US, EU AI Act, etc.). Robots.txt remains the most practical opt-out mechanism today, plus emerging standards like ai.txt.
How often does GPTBot crawl?
Training crawlers usually visit periodically — weekly or monthly waves rather than daily. If you see sudden spikes, monitor whether the bot is honoring Crawl-delay directives in your robots.txt.
See which pages AI training crawlers target
Monitor training-oriented bots, identify the content they access most, and decide what to allow or block.
- Track training crawler activity per page
- See exactly which content is being scraped
- Make smarter allow or block decisions
Free plan available. No credit card required. Setup in 2 minutes.