Smush Analytics

Is your site ready to be cited by AI answer engines?

Enter any public URL. We grade how well it can be retrieved, parsed and trusted by AI answer engines — then show you exactly what to fix first.

Run a readiness audit

Any public page. No login or Google connection needed.

Ruleset 2026.08.2 (2026-08-19). We report readiness and measurement — never a promise of citations.

Pillar A — the why

Readiness audit

Crawl-based and auth-free. 6 dimensions across three gates — can an engine fetch it, parse a clean answer from it, and trust the source. Tier-1 blockers cap the grade.

Pillar B — the whether

Visibility tracking

Real prompts run across Perplexity, ChatGPT, Gemini, Claude, Google AI Overviews — mention rate, citation rate, position, sentiment and share of voice, per engine, tracked over time.

The loop

Measurement plus diagnosis

Visibility tells you that you are absent for a prompt. The audit tells you the likely reason — a blocked crawler, a client-rendered page, no extractable answer. Fix, then re-measure.

What the Tier-1 blocker checks catch

Blockers gate the grade because nothing downstream matters if an engine cannot fetch the page.
  • robots.txt access for GPTBot, OAI-SearchBot, PerplexityBot, Google-Extended, ClaudeBot and seven more AI user agents — reported per agent, with what each one actually governs.
  • Edge, CDN and WAF blocks — probed with real AI user-agent strings, because a 403 from Cloudflare beats an allowing robots.txt every time.
  • Indexability and HTTPS — noindex directives, X-Robots-Tag headers and canonicals pointing away.
  • JavaScript dependency — whether the primary content exists in the raw HTML at all, or only after hydration.

What we will not claim

The product promises readiness and measurement, never a guaranteed outcome.
  • No rankingAI answers are not a ranked list. There is no “Search Console for ChatGPT”. We report rates and share of voice.
  • DirectionalPublished GEO uplift figures come from a thin research base (Aggarwal et al., “GEO: Generative Engine Optimization”, arXiv:2311.09735, 2024). Where we cite one, it is labelled directional — never a guarantee.
  • Non-deterministicAnswers vary by run, locale and model version. Every prompt is sampled repeatedly and reported as a rate with a confidence indicator.

Already tracking a site?

Projects hold a site, its competitor set, its prompt set and the history of both pillars.