Bot traffic split by purpose: Bot traffic now classifies each hit as AI Answers, Indexing, Training, or Other instead of grouping every crawler together. Category tabs filter the chart, bot list, page list, and coverage views, and the classification applies to historical data too. The same breakdown reaches get_analytics, get_analytics_overview, and AI-generated analytics recaps.
Broken-page tracking for AI crawlers: Analytics now records the pages AI crawlers request that return a 404, without counting those requests as successful crawls. Repeated requests for a missing URL become a concrete list of pages assistants expect to find. Settings → Analytics → Add to App includes the not-found capture hook and updated instructions for supported server and edge stacks.
AI Visibility Score: The analytics dashboard now calculates a deterministic 0–100 AI Visibility Score from crawl coverage, crawler breadth, referral volume, assistant breadth, and discovery-file health. Open the score badge to see the points contributed by each component. The score uses the existing bot-traffic request, returns 0 when there is no traffic, and is also available through get_analytics_overview.
Automated AI visibility checks: A new AI visibility check automation asks buyer-style questions, then reports whether your brand appears, its prominence and share of voice, cited sources, competitors, and unanswered content gaps. It runs through the Claude Code, Codex, or Cursor subscription already connected to the automation, without requiring a separate metered API key. Enable it from Automations under Analytics & insights or from the Automations button in the analytics header; its quantitative baseline comes from the computed AI Visibility Score.
AI referrals from real visitors: A new AI Assistants card separates human visits sent by assistants from crawler activity. It shows total referrals, period-over-period changes, the assistants that sent people, and their most-visited landing pages, with coverage for ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, Meta AI, DeepSeek, Mistral, You.com, Poe, Phind, Hugging Face, and more. The card loads on demand and the same data is available through get_analytics.
Grok CLI and OpenCode providers: Grok CLI and OpenCode now appear anywhere you choose an assistant, including chat, agents, automations, tasks, and workspace defaults. Aiola discovers Grok models and reasoning options from the installed CLI and reads the exact provider/model pairs exposed by OpenCode. Both providers include setup and update controls in Settings → CLI tools, plus usage reporting on the dashboard.
Navigation and reporting for long-running chats: Long threads now have a vertical turn rail with hover previews and click-to-jump navigation, plus an active marker that follows your scroll. When an assistant starts background work, the session remains active and posts a new reply when the task finishes. Background tasks and spawned agents also show live running and completed status rows with expandable summaries.
One model and reasoning switcher: The chat composer now combines model, reasoning effort, context window, and fast mode in one switcher. Open it to move through the provider and model list, choose a supported context window, adjust effort on a color-coded slider, or toggle fast mode. The same control appears everywhere the chat composer is used.
GUI Lifetime enforcement and offline licenses: GUI Lifetime now enforces the full local workspace while clearly gating hosted analytics and integrations, backed by signed offline licenses that resolve before central authentication. A license whose tier no longer matches your account is replaced rather than trusted, and the cached license and credential files are written with owner-only permissions.