What is Conversation thread retention?
Conversation thread retention is the per-engine memory window inside a single AI search session — the number of prior turns the engine carries forward into the retrieval, rerank, and synthesis stages of the next turn. Mid-2026 per-engine anchors: ChatGPT Search carries 8–12 turns of state on commercial sessions; Perplexity carries 5–8 turns; Google AI Mode carries 4–7 turns; Microsoft Copilot carries 6–10 turns; Amazon Rufus carries 3–5 turns with asymmetric weighting toward product-discovery state; Claude carries 10–15 turns with the deepest entity-graph retention. The retention window is not symmetric — the most recent 1–2 turns carry 60–80% of the state weight, with older turns carrying decayed influence. Retention windows matter because the conversion-driving decisions cluster on turns 2–4 of a commercial conversation, well inside the retention window of every major engine — which is why multi-turn engineering is structurally available, not theoretical.
How it relates to AI UGC
Visual thread retention runs on the same per-engine windows — the multimodal substrate carries the head-turn carousel state across the same retention horizon as the text substrate carries the cited-URL state. Persona-locked visual sets compound across the full retention window; rotating imagery resets the visual entity signal turn-by-turn.
Key statistics
- Per-engine mid-2026 retention anchors: ChatGPT Search 8–12 turns, Perplexity 5–8, Google AI Mode 4–7, Copilot 6–10, Rufus 3–5, Claude 10–15 (engine-retention audits, 2026).
- The most recent 1–2 turns carry 60–80% of the state weight across every major engine — older turns decay but do not reset (state-weight-decay audits, 2026).
- Commercial-conversion-driving decisions cluster on turns 2–4 of a typical commercial AI search conversation, well inside every major engine's retention window (commercial conversation-shape audits, 2026).