OpenAI / ChatGPT Has A Web Cache

OpenAI’s ChatGPT web search feature does offer a web cache, an offline, locally stored version of web pages that have been previously crawled. This should not come as a surprise, as all search engines store cached versions of web pages (even if they don’t show it to us).

Amy Rigby posted about this on LinkedIn and wrote, “For months, SEO/AEO pros have claimed that OpenAI is building a “hidden cached index” of webpages (H/T James Berry and Jérôme Salomon). Did OpenAI just confirm it in this help article.”

Here is that help document and you can see it says, “Offline web search is a web search configuration for eligible ChatGPT workspaces. It allows ChatGPT to use OpenAI’s indexed and cached web content instead of using live web search at the time of each request.”

Here are the lines that mention the cache:

• It allows ChatGPT to use OpenAI’s indexed and cached web content instead of using live web search at the time of each request.
• When offline web search is enabled, ChatGPT uses OpenAI’s indexed and cached web content for web search.
• This means ChatGPT can return information from pages that are already available in the OpenAI index or cache.
• If a page or URL is not available in the index or cache, ChatGPT will not not be able to retrieve it through offline web search
• Use a regulated workspace configuration where cached or indexed web access is enabled by default.
• A particular page may be refreshed on a different schedule depending on the site, crawl access, caching, popularity, and other technical factors.
• If a user asks ChatGPT to use a specific URL, offline web search can use that URL only if the page is available in the index or cache.
• Offline web search results may not include the exact time when a page was indexed or cached.
• It limits web search to OpenAI’s indexed and cached web content instead of using live external web search at request time.
• Offline web search does not eliminate all risk. Indexed or cached web content can still contain inaccurate information, incomplete information, outdated information, or malicious instructions.
• If offline web search is enabled without a broader security mode, most ChatGPT features should not change solely because search uses OpenAI’s indexed and cached web content.
• When offline web search is enabled, users may notice that ChatGPT can still search the web, but results come from OpenAI’s indexed and cached web content instead of a live external search provider at request time.
• Offline web search uses indexed and cached content.
• Can ChatGPT show when a page was cached? Not always. If your workflow requires a reliable timestamp, use uploaded source material, official documents, or another approved source-of-record process.

Use this table to decide whether offline web search is appropriate for a workflow.

Forum discussion at LinkedIn.