Method
What we check, how we ask, and what the evidence says
Two separate things: a readiness check (can AI search read your site?) and a measurement (what do the engines actually answer?). Only the second tells you what buyers see.
1. The readiness score
We fetch your homepage, robots.txt and sitemap once, the way a crawler would, with the user agent described on our bot page. Each check passes (full weight), warns (half) or fails (none). The score is the share of weight earned. Checks with weight 0 are shown as notes and never move the score.
| Check | Weight |
|---|---|
| AI search crawlers can read your site (robots.txt) | 15 |
| Firewall does not turn AI crawlers away | 10 |
| Search engines may index the page | 10 |
| Your words are in the HTML without JavaScript | 10 |
| Clear title and description | 6 |
| Sitemap | 5 |
| Contact details are easy to find | 5 |
| Structured data says who you are | 4 |
| Answers to buyer questions on the page | 4 |
| Secure (HTTPS) | 4 |
| Canonical URL | 3 |
| Share preview (Open Graph) | 3 |
| One main heading | 3 |
| Shopify: product pages have Product markup | 8 |
| AI training crawlers | 0 |
| llms.txt | 0 |
The heavy weights are the clear-cut ones: if a crawler is blocked by robots.txt or a firewall, or your text only appears after JavaScript runs, an engine can’t read the page. The engines’ own documentation describes their search crawlers and how robots.txt controls them (for example OpenAI and Google). Blocking training crawlers is your choice and is not scored.
2. What public studies say
- llms.txt. Ahrefs looked at about 137,000 domains with an llms.txt file and found that 97% of those files received no traffic at all in May 2026 (Ahrefs, June 2026). We list it as optional and don’t score it.
- Schema markup. Ahrefs compared 1,885 pages that added schema with 4,000 that didn’t: ChatGPT citations changed by about +2.2% and Google AI Mode by +2.4%, which the authors call statistically indistinguishable from zero. The study covered pages that were already being cited, so it doesn’t prove schema never helps (Ahrefs). We give it a low weight, because it still states your facts unambiguously.
So we present every fix as best practice, not as a way to get recommended. Nobody outside the engines can promise that.
3. How the paid audit asks the engines
- Each question is sent as written, through the engine’s official API, with web search on where the API offers it. One fresh request per question: no account history, no location, no personalisation.
- We record the answer excerpt, whether your business is named, who is named instead, and the sources the engine cites, with the date and the exact prompt.
- API answers can differ from what a signed-in person sees in the consumer app, and answers change from day to day. Treat the result as a dated sample, not a ranking. Re-run within 30 days to compare.
- The questions come from your own list, topped up by drafts written from your site (by an AI model when one is available, otherwise from templates).
4. What we don’t do
No ranking or recommendation promises, no fake reviews, no link schemes, no content written to manipulate engines. We are not affiliated with OpenAI, Perplexity, Anthropic, Google or Shopify.