From a domain to an audit in minutes.
Here's exactly what happens between you typing your URL into the form and getting a finished report. No black box.
You give us your domain, brand and category
Three fields. The brand name has to be exactly how you brand it (e.g. "JB Hi-Fi" not "jbhifi") because that's what AI engines write — we match on the literal string. Category sharpens the buyer questions we'll test.
- Domain becomes the audit subject (also the site we screenshot).
- Brand name + aliases are how we detect "you appear in this answer."
- Category drives the auto-generated query set (e.g. "best family hotels in Bali" or "alternatives to Airbnb").
We generate a 40-question buyer-query set
Five intent buckets: brand (people who already know you), category (discovery), problem (the pain that drives buying), comparison (head-to-head decisions), and vertical (industry-specific intent). The free preview uses 8 of these; full audit uses all 40.
We query 5 AI engines in parallel
Each engine sees the same questions and gets to use its native web search. We capture the full answer text plus the source citations.
- Google AI Overviews via the Apify google-search-scraper actor
- ChatGPT (gpt-5-mini with built-in web search)
- Claude (Haiku 4.5 with built-in web search)
- Perplexity (Sonar — search-native model)
- Gemini (2.5 Flash with grounding)
In parallel: we run a GEO audit on your site
While the AI engines are answering, we directly inspect your site with 15 technical checks — GEO (Generative Engine Optimisation). Crawlability (robots.txt allows GPTBot, sitemap.xml, llms.txt), structured data (Organization, FAQPage, Product schema), metadata (description, OG, canonical), content shape (single H1, server-rendered HTML), and performance (HTTPS, mobile viewport, response time).
Each check returns a status (pass/warn/fail), what we'd expect, why it matters, and how to fix it. This catches the silent killers — like a robots.txt that accidentally blocks ChatGPT, or a JS-only site invisible to AI crawlers. Learn more about GEO →
We score every AI answer two ways
First: deterministic. Regex on the brand name. URL parse on every cited domain. Substring match for competitors. Fast, repeatable, no LLM needed for these signals.
Second: LLM scoring (paid tiers only). A second-pass call to Claude Haiku checks each answer against your business's ground-truth facts and tags sentiment, accuracy, and any specific false claims.
We assemble the report
Headline metrics, engine heatmap, top-cited domains, competitor share-of-voice, hallucination flags, per-query drill-down. The hosted report is a single self-contained HTML page with inline CSS — opens in any browser, prints to PDF cleanly.
- CSV export for your team
- PDF export delivered with the report email
- Raw responses archived so we can re-score offline if you need to pivot the analysis
You get the report by email
One email. Hosted report link. PDF attachment. CSV attachment. Subject line: "Your [Brand] AEO audit is ready." Delivered within minutes of payment — usually under 10.
What we don't do
- We don't scrape the AI engines without permission. We use OpenRouter's official tool-use API for the LLMs and Apify's compliant SERP scraper for Google.
- We don't store your audit data forever. Raw responses live on our servers for 90 days then are deleted. The hosted report stays accessible for 12 months.
- We don't do an SEO audit. If you want backlink analysis, on-page recommendations or keyword research, use Ahrefs or Semrush. We measure something different.
- We don't claim guaranteed AI rankings. No one can. AI engines change weekly. What we give you is measurement and direction.