There are 3 main user agents used by ChatGPT:
• ChatGPT-User
• OAI-SearchBot
• GPTBot
ChatGPT-User is a real-time bot and acts immediately based on prompts.
None of the 3 OpenAI bots can render JavaScript. Your main content has to be in the html.
Foundational steps to rank in ChatGPT Search:
1. Be indexed in Bing
2. Be accessible for OpenAI's bots
Searches on ChatGPT are mainly long-tail searches.
Bing supports IndexNow. If you are on Shopify, Wordpress, or Wix, check out the available IndexNox plugins.
ChatGPT sometimes uses multiple documents from the same domain.
Sometimes ChatGPT hallucinates URLs. If ChatGPT keeps visiting or hallucinating the same URL, consider creating or redirecting it.
ChatGPT is not crawling llms.txt files at this time. This makes sense since llms.txt is more relevant for agentic use cases.
If you mention your xml-sitemaps in your robots.txt, ChatGPT will crawl them.
With Inspect > Network > Fetch/XHR you can access a JSON file of all the information that went into generating the result. You must be logged in to access the file. There are different types of URLs in the JSON:
• search_results: all URLs
• sources_footnotes: the main citations
• supporting_websites
• safe_urls
• blocked_urls: URLs ChatGPT cannot access or does not want to show to the user for safety treasons
You can use tools like Peec AI to track a few hundred prompts. Then look at the list of the most important sources across all prompts. I do this as well!
Many of the most important sources will be sites that your traditional PR efforts would have missed.
You can ask ChatGPT which attributes it considers when making a recommendation. Use this to create truly holistic content!
Many people assume structured data helps with ranking in ChatGPT. But no one has ever shown data to proof it.
This post was originally shared by on Linkedin.