OpenAI respects robots.txt directives for its crawlers, specifically GPTBot (used for training AI models) and SearchGPT's crawler (used for their search product). You can block either or both by adding "User-agent: GPTBot" or "User-agent: OAI-SearchBot" followed by "Disallow: /" in your robots.txt file. GPTBot appeared in mid-2023, so any content scraped before you implemented a block remains in OpenAI's training corpus. There's no retroactive opt-out for that data. If you block GPTBot today, you're preventing future scraping cycles, not erasing what's already been ingested. This matters for sites with proprietary content, paywalled articles, or client work samples you don't want feeding into ChatGPT responses. SearchGPT's crawler (OAI-SearchBot) is separate and powers their search features, not model training. Blocking it keeps your pages out of SearchGPT results but doesn't affect GPT model knowledge. Most sites allow OAI-SearchBot because search visibility has direct traffic value, while blocking GPTBot to protect content from becoming free AI training material. At Ottawa SEO, we typically leave both crawlers allowed for informational content and blog posts, since AI-generated answers that cite our domain can drive referral traffic. For client case studies or proprietary methodologies, we block GPTBot to avoid giving away competitive intelligence. We've seen zero negative SEO impact from blocking GPTBot because it doesn't influence Google rankings. One caveat: third-party scrapers and smaller AI companies don't always respect robots.txt. Blocking OpenAI's official crawlers is enforceable because they've publicly committed to honoring it, but bad actors ignore directives entirely. If content protection is critical, consider authentication walls or obfuscation beyond robots.txt. For most sites, controlling GPTBot and OAI-SearchBot gives you reasonable leverage over how OpenAI specifically uses your content without sacrificing discoverability.