v0.0.7 — Tracked Prompts, Site Audit & Grounded Answers
Track your own prompt list rather than discovered questions, audit your whole site for how well it answers AI queries, query engines with live web search, and add Google AI Overviews as a seventh surface.
A larger definition of a scan
Five expansions to what a scan covers and how much of it you control.
Track Your Own Prompts
Rather than relying only on automatically discovered questions, you can maintain your own list of prompts per project, and scans draw from that list. Each scan runs up to your plan’s per-scan prompt limit, so order the list by what matters most. Suggested prompts come with topic clustering, and with estimated search volume when a keyword-data provider is configured.
Site-Wide Content Audit
Point the audit at your domain and it discovers and crawls your pages, scores each one for how well it answers the questions AI engines are asked, and aggregates the result into the Content Gaps view.
Search-Grounded Answers
Engines can be queried with their provider’s live web search enabled, so an answer comes back with the URLs it actually cited rather than domains inferred from the prose. This is a deployment-level capability, not an account setting — there is no switch for it in the dashboard, it is off by default, and where it is enabled a grounded prompt carries a credit multiplier, because a grounded query costs materially more to run than a plain one. Perplexity and Google AI Overviews are the exceptions in both directions: they search on every call regardless, so their result is grounded already and carries no surcharge.
The visible consequence, when it is off, is that a citation is a domain rather than a page. Every URL-level derivation in the product reports that URL data is unavailable rather than running on empty — without that guard, zero citation URLs would make every page you rank for read as “not cited”, which is a confident and entirely fictitious finding.
Google AI Overviews as a Seventh Surface
Scans can include Google AI Overviews, the answer block above Google’s results. It works differently from the other six: rather than querying a chat model it reads a search-data provider, so it is included only when one is configured, and skipped from a mixed scan when none is. That makes seven surfaces in total.
Two More Free Tools
The free tool set gains an llms.txt audit and a schema markup generator.
Want to shape what we build next?
See what we're building next, and tell us what you need.