ChatGPT Rebuilt Its Search Tool—Here's What the New Language Reveals

Search Engine Journal discovered that OpenAI replaced ChatGPT's JSON-based search calls with a compact, pipe-delimited query language between August 16 and August 20. The new format reveals five operational patterns: search calls now execute one per line with readable fields; a freshness window (in days) adjusts by query type—2 days for stock prices, 30 for commercial product comparisons, 365 for community forums; ChatGPT segments searches into verticals (fast for web, product for catalogues, business for local, image for images, and genui_run for widgets); domain-targeting now uses a dedicated slot instead of site: operators; and some answers bypass citations entirely by triggering widget functions. For SEO and content strategists, the implications are direct: pricing and buying-intent pages need 30-day refresh cycles to stay within ChatGPT's default freshness window, local businesses must ensure their entity data is accurate in places databases rather than relying solely on website content, and physical product presence in ChatGPT's catalogue layer operates independently from blog content ranking.