Real-time knowledge

Real-time knowledge is an LLM's access to information from the past minutes/hours/days via live data feeds — Grok's X firehose, Perplexity's web search, ChatGPT's browse — separate from the model's static training cutoff.

Frontier LLMs have a knowledge cutoff (the date their training data ends) that's typically 6-12 months before release. Real-time knowledge bridges this gap: Grok pulls live X (Twitter) posts, Perplexity and ChatGPT search the web, Gemini queries Google. By 2026 real-time access is table-stakes — almost every consumer chat product ships it. The implementation matters: native feeds (Grok + X) are faster but narrower; web search (Perplexity, ChatGPT, Gemini) is broader but slower and can hit paywalls or stale caches. Production agents combine both via [[tool-use]] — the model decides when to call the search tool vs answer from training data.

When to use real-time knowledge

• Current events, news, market data.
• Anything past the model's training cutoff.

Common mistakes

• Trusting the model when it doesn't search — knowledge-cutoff facts can be stale by months.
• Skipping source citations — users can't verify real-time claims.

FAQ

Related terms

• Knowledge cutoff — The knowledge cutoff is the date after which a language model has no training data — anything that happened after is unknown to it unless supplied at inference time.
• AI search engine — An AI search engine answers a user's query by retrieving relevant web sources and synthesising a cited answer with a language model — the category that includes Perplexity, ChatGPT Search, Claude with web, and Gemini AI Overviews.
• Grounding — Grounding is any technique that ties a language model's output to verifiable sources — retrieved documents, tool results, structured data — instead of pure memory.

Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/real-time-knowledge.md.