The questions people actually ask

One crawled URL. A scan that fetches 400 pages spends 400 credits from the month's allowance. Your plan's allowance is spent first, then any add-on credit you have bought, then anything support has granted — the billing screen in the product shows all three balances and what is left.

Keep them. Those are research tools: they tell you what to go after. They will not tell you that last night's deploy dropped the canonical off your category pages. PixyScan watches the site you already have and reports what changed since it was last fine. Different job, works alongside.

An audit is checking for smoke once a year. This is the smoke detector. A single audit gives you a list you will mostly already know; the value here starts with the second scan, because that is the first time a finding means something changed.

No. PixyScan reads the HTML you serve — structure, links, markup and machine readability — not how fast a browser paints it. It does not run Lighthouse and does not report LCP, INP or CLS. Use PageSpeed Insights or your RUM provider for those.

Not yet, and we would rather say so. Scan events raise in-app alerts — scan failed, score dropped, score below your threshold — which appear on the bell and on the Alerts screen. Email delivery for those is the next thing on the list, not a shipped feature. Email today is used for signup codes, password resets, invitations and billing notices.

Partly, and here is the exact line. Each site can issue a client secret; `npx @pixyscan/scan-sdk run` scans the branch's environment URL and exits non-zero when the score falls below your threshold, which is enough for your pipeline to fail the build. What does not exist is a GitHub app that comments on the pull request or blocks the merge itself. Your pipeline makes that call.

PixyScan reads the server-rendered HTML, which is what a crawler sees first. That is deliberate — several checks exist precisely to catch metadata that only appears once JavaScript has run. If your title is injected client-side, PixyScan will tell you.

Yes, that is what branches and environments are for. A scan records the branch it ran against and is only ever compared with the previous scan on the same branch. Map a branch pattern to a URL under Settings → SDK & CI/CD, and a preview run is filed as `preview` rather than becoming production's baseline.

Five weighted groups, all of them things the crawler can actually see: answer signals (does the page answer a question in a liftable form), AI engine access (llms.txt and llms-full.txt, and the author and freshness signals engines weigh), machine-readable meaning (JSON-LD valid for the type it claims), document structure, and extractable prose. It does not claim to predict whether a model will cite you.

Cancelling means one thing: the subscription stops auto-renewing. Every entitlement is honoured to the end of the period you have already paid for, there is no mid-cycle termination, and your data stays where it is. What there is not is a refund — a purchased period runs to its end and the payment for it is final. Downgrades are deferred to the end of the cycle for the same reason.