v0.0.19: Answer Archive, GSC Prompt Import & AI Mode Finalized

Turn what you already know into what you track

This release is about surfaces over the data the product already stores. The scan-data programme: a browsable evidence trail for every answer, and your own real Google queries turned into tracked prompts.

The Answer Archive

Every verbatim AI answer is stored per scan. That has always been true, but it was a database row, not something you could read. The archive makes it a surface: from the rankings drill-down, any past scan’s answer to a question can be reopened, with the wording changes against the scan before it shown as additions and removals, alongside the claims the answer newly makes or dropped. It is the answer text diffing that makes a hallucinated-claims problem visible: what an engine changed its answer to, and when.

GSC Queries → Tracked Prompts

The Search Console top-queries table now lets you select the queries that are already earning you clicks and import them as tracked prompts in one click: real demand, straight from your own property, instead of a synthetic prompt set. Imports land inactive (activation stays behind your plan’s active-prompt cap), dedupe case-insensitively against your existing prompts, and carry an estimated monthly volume wherever a keyword provider is configured. The tracking Google search performance tutorial covers the Search Console side; the choosing tracked prompts guide covers what happens once they are in your library.

Google AI Mode tracking, finalized

One Less Thing to Choose

Google AI Mode joined the weekly SEO-data pass in v0.0.18. This release closes the loop: it no longer appears in the AI engine picker in Settings. Like Google AI Overviews, it is not a model you select: on paid plans it runs automatically every week alongside AI Overviews, answered by a SERP lookup rather than an LLM call, and billed on the SEO-data meter instead of your credit allowance.

About Search Console’s Generative AI Report

Search Console’s generative AI performance report (announced June 2026) is worth knowing about but not yet worth building on: it is impressions-only (no clicks, CTR or position), it reports AI Overviews and AI Mode together rather than split apart, it is rolling out to properties incrementally, and it is not exposed through Google’s API at all. The knowledge base now says exactly that, so nobody reads more into the report than Google publishes. AI Mode visibility in LLM Metrix comes from SERP lookups on the weekly SEO-data pass instead.

Microsoft Copilot Tracking

Microsoft Copilot joins the weekly SEO-data pass as its own SERP surface: the answer panel from Bing’s copilotsearch page, read through a SERP provider like the Google AI Overviews shape. Copilot has no public chat-completions API, so this is a lookup, never a model call, billed on the SEO-data meter and dropped from scans when no provider serves it.

Deep Research Monitoring

A Deep Research run asks a research model to answer your tracked prompts with a full report, then grades each report for brand mention, rank, stance and claims. On paid plans it is billed from your credit allowance on its own meter.

Closed-Loop Execution

A recommendation on the board can now be turned into a deliverable (an llms.txt file, schema JSON-LD, or a content brief), copied or downloaded right from the item, with every generation audit-trailed.

The Public API Platform

The developer API grows from two read-only endpoints to a read+write REST surface: trigger scans, retry jobs, manage prompts and the webhook URL, plus an MCP server for AI agents, all on the same personal access tokens, with the OpenAPI contract generated from the request schemas.