How AISutra works
A transparent, repeatable way to measure how ready your website is for AI assistants and AI-driven search.
The audit, step by step
- Crawl. AISutraBot reads up to 15 public pages of your site, plus
robots.txtand your sitemap. It identifies itself, respectsrobots.txt, and waits between requests. About AISutraBot. - Extract. Code pulls out the facts AI systems rely on: business name, address, phone, email, hours, prices, licensing statements, structured data, headings and titles.
- Check. 31 checks and the customer questions below are scored with fixed rules, each backed by evidence from your pages.
- Report. You get your score, the evidence behind it, a prioritized plan and ready-to-paste fixes.
Six categories, 100 points
Weights add up to 100. Each category answers a plain question about your site.
| Category | Points | What it asks | Checks include |
|---|---|---|---|
| Business Clarity | 20 | Can an AI tell what you do, who you serve and why someone should choose you? |
|
| Local & Contact Signals | 15 | Can an AI find your address, phone, hours and the area you serve, and are they consistent? |
|
| Structured Data | 20 | Does your site state its facts in the machine-readable format (schema.org) that search and AI systems prefer? |
|
| Technical Foundations | 10 | Are the technical basics (security, titles, descriptions, headings, sitemap) in place? |
|
| AI Crawler Access | 10 | Are AI crawlers allowed in, and is your content readable without running scripts? |
|
| Customer Question Coverage | 25 | When a parent asks an AI assistant a typical question, does your website contain the answer? | Questions customers ask AI about your business, graded from your pages. See below. |
Customer question coverage
Twenty-five of the 100 points come from how well your site answers the questions customers ask AI assistants about businesses like yours. For child care, that is 22 questions: ages served, hours, tuition, licensing, ratios, curriculum, safety, meals, enrollment, and more, weighted by importance.
A question counts as answered only when a sentence on your site states it directly, partial when the information is vague or says details are "to be confirmed", and missing otherwise. Every answered question shows the sentence and the page it came from.
Evidence and repeatability
- Evidence first. Nothing is marked found, passed, or answered without text from your own pages to show for it. If we cannot show evidence, it is marked "Not found in text".
- Rules, not opinions. Scoring is mechanical: fixed rules applied to your pages. Run the audit twice on the same pages and you get the same score, so before-and-after audits are comparable.
- AI assistance is optional and advisory. An AI model may help write a short summary of your business. It never changes your score.
How recommendations are prioritized
Critical items block AI or customers outright, such as an address or phone number missing from your text, AI crawlers blocked, or conflicting contact details. High, Medium and Low items follow by how many points they cost. Each recommendation shows the verified finding behind it, why it matters, the fix, and the effort.
What AISutra does not measure
- It does not query ChatGPT, Gemini or other assistants, and it cannot predict what they will say. Their answers vary by person, place and day. Your report includes a manual spot-check sheet to try yourself.
- It does not check third-party listings, reviews or social profiles. The report includes an off-site checklist instead.
- It reads text delivered in your page's HTML. Facts shown only in images, PDFs or built entirely by JavaScript are reported as "Not found in text", and flagged when your site looks JavaScript-heavy.
- It reviews up to 15 pages and public pages only.
Industries
Question sets and detection rules are built per industry. Child care and preschool is available now and fully calibrated on real websites. Other small-business industries will open one at a time as each is tested.