Session summarization
Session summarization is generating a natural-language account of what happened in a recorded session: the user's apparent goal, the path taken, the friction met, and the outcome. It is the compression step that makes large session archives readable, replacing watch-time with reading-time at scale.
How it works
The input problem dominates the design: raw event streams are too long for model contexts at session scale, so summarization pipelines reduce first, extracting a structured digest of the session (pages, actions on named elements, errors, timing gaps, frustration signals) and prompting the model over the digest rather than the raw stream. Quality follows the digest's fidelity, since a summary can only mention what the reduction preserved, which is why semantically identified elements (the checkout button, not node 4127) matter to the pipeline design. Summaries then become an index layer: they are what humans skim, what search embeds (see semantic session search), and what aggregation runs over when the question is "what were the common struggles this week" rather than "what happened in this session."
Where you encounter it
AI features of replay platforms, rrweb AI's session views, and support workflows where agents read the summary before the replay.