How the score works
Our score is a weighted average of 30 checks in four areas. Here is every check and every weight, so you can see exactly how we got your number.
What is the score, in short?
The score is a number from 0 to 100. It measures how ready your website is to be read and quoted by AI answer tools.
It does not measure whether an AI tool names you today. That also depends on your reputation and on what other websites say about you. We say this plainly because no tool can promise you a mention.
These rules are version 2026.2. When we change a rule, we change the version number, so old and new scores are never mixed up.
What are the four areas?
Each area answers one question. These are the weights for a general website.
Does it give clear answers?
People ask AI questions. Pages that answer them plainly, with real facts like prices and opening hours, get used.
Can AI read it?
AI reads the text inside your page's code. Words hidden in pictures, or added late by small programs, may be invisible to it.
Would AI trust you?
AI prefers businesses that show who they are. A name, contact details, dates and customer reviews all help.
Can AI get in?
A website can tell AI programs to keep out. Sometimes that is set by mistake. We check the door is open.
How is the score worked out?
- Each check gets a result. A pass is worth 1. A half pass (we call it a warning) is worth 0.5. A fail is worth 0. If a check does not apply to your site, we leave it out.
- Each area gets a score from 0 to 100. It is the average of its checks, where serious checks count more (see the table below).
- Your overall score adds up the four area scores, each multiplied by its weight. If an area has no checks that apply, we share its weight among the others.
A worked example. Say a site scores 74 for answers, 88 for readability, 76 for trust and 100 for crawler access. The overall score is 74 × 0.35 + 88 × 0.25 + 76 × 0.25 + 100 × 0.15 = 81.9, which rounds to 82. That is a B.
How much does each check count?
| How serious | Weight |
|---|---|
| Fix now (critical) | 4 |
| Fix soon (high) | 3 |
| Worth fixing (medium) | 2 |
| Small fix (low) | 1 |
| Grade | Score |
|---|---|
| A | 85 to 100 |
| B | 70 to 84 |
| C | 55 to 69 |
| D | 40 to 54 |
| F | 0 to 39 |
Do weights change by type of site?
Yes. A law firm is judged more on trust, and an online shop more on readable product facts. You can pick your type of site before a check, or we use General.
| Type of site | Answer-readiness | Machine readability | Authority signals | Crawler access |
|---|---|---|---|---|
| General (default) | 35% | 25% | 25% | 15% |
| Online shop | 30% | 30% | 25% | 15% |
| Law firm | 30% | 20% | 35% | 15% |
| News or magazine | 25% | 25% | 25% | 25% |
| Local business | 25% | 30% | 30% | 15% |
| Software | 35% | 25% | 25% | 15% |
Which checks do we run?
All 30 checks run on every plan. Each one has an ID that appears in your report. Paid plans add 3 more checks where an AI model reads your main pages and judges the writing; those are clearly marked in the report.
Answer-readiness: Does it give clear answers?
| ID | What we check | How serious |
|---|---|---|
| AR-01 | Pages open with a direct answerWhen it fails we say: Your pages don't answer the question right away | Fix soon |
| AR-02 | FAQ or Q&A section existsWhen it fails we say: You don't have a questions-and-answers section | Fix soon |
| AR-03 | Pages contain specific factsWhen it fails we say: Your pages are short on hard facts | Worth fixing |
| AR-04 | Comparison or spec tables on product/service pagesWhen it fails we say: Prices and details aren't laid out in a table | Worth fixing |
| AR-05 | Key pages have adequate content lengthWhen it fails we say: Your main pages are too short | Worth fixing |
| AR-06 | Headings are phrased as questions or clear topicsWhen it fails we say: Your headings don't sound like real questions | Small fix |
Machine readability: Can AI read it?
| ID | What we check | How serious |
|---|---|---|
| MR-01 | Main content is present in the raw HTMLWhen it fails we say: AI tools see an almost empty page | Fix soon |
| MR-02 | JSON-LD structured data is on the homepageWhen it fails we say: Your homepage has no hidden "business card" for AI | Fix soon |
| MR-03 | JSON-LD parses and declares @context and @typeWhen it fails we say: Your hidden "business card" for AI has mistakes | Fix soon |
| MR-04 | Appropriate schema types are present on key pagesWhen it fails we say: Your pages don't say what kind of page they are | Fix soon |
| MR-05 | Title tag is present and 30-65 charactersWhen it fails we say: A page title is missing or the wrong length | Worth fixing |
| MR-06 | Meta description is present and 70-160 charactersWhen it fails we say: A page summary is missing or the wrong length | Worth fixing |
| MR-07 | Exactly one H1 per pageWhen it fails we say: A page doesn't have exactly one main headline | Worth fixing |
| MR-08 | Canonical tag is present and self-consistentWhen it fails we say: Pages don't say which web address is the official one | Worth fixing |
| MR-09 | Semantic HTML structure (nav, main, article/section)When it fails we say: Your page layout isn't labelled | Small fix |
| MR-10 | Images have alt textWhen it fails we say: Some images have no description | Small fix |
Authority signals: Would AI trust you?
| ID | What we check | How serious |
|---|---|---|
| AU-01 | A named author or organization is identifiedWhen it fails we say: It's not clear who is behind the site | Fix soon |
| AU-02 | An About page exists and is linkedWhen it fails we say: There's no "About us" page we could find | Worth fixing |
| AU-03 | Contact information is visibleWhen it fails we say: Your contact details are hard to find | Worth fixing |
| AU-04 | Content pages show a published or last-updated dateWhen it fails we say: Your pages don't show a date | Worth fixing |
| AU-05 | Content links out to credible external sourcesWhen it fails we say: You don't link to outside sources | Small fix |
| AU-06 | Privacy policy and terms pages are linkedWhen it fails we say: No privacy or terms pages are linked | Small fix |
| AU-07 | Reviews or testimonials are presentWhen it fails we say: There are no customer reviews on your site | Worth fixing |
Crawler access: Can AI get in?
| ID | What we check | How serious |
|---|---|---|
| CA-01 | robots.txt is present and readableWhen it fails we say: Your site has no clear "who may enter" file | Fix soon |
| CA-02 | AI crawlers are allowed on the homepageWhen it fails we say: Your site tells AI helpers to stay out | Fix now |
| CA-03 | AI crawlers are allowed on key pagesWhen it fails we say: Some of your pages are closed to AI helpers | Fix soon |
| CA-04 | Key pages are not noindex or nosnippetWhen it fails we say: Some pages say "don't list me" | Fix now |
| CA-05 | sitemap.xml exists and is referenced in robots.txtWhen it fails we say: You have no page map for AI helpers | Worth fixing |
| CA-06 | Key pages return HTTP 200 without redirect loopsWhen it fails we say: Some of your pages are broken or loop around | Fix soon |
| CA-07 | llms.txt file is present (informational)When it fails we say: You don't have an llms.txt file | Small fix |
What does the score not tell you?
- It does not ask ChatGPT or any other AI tool about you. It looks only at your website.
- It does not measure your reputation, your reviews on other sites, or press about you. Those matter a lot to AI tools.
- It reads a sample of pages (10 on the free check), not your whole site.
- Some checks use simple rules, such as looking for a date near the words "last updated". Rules can miss things or be fooled. Read each finding, not just the number.
The file that tells crawlers about your site, llms.txt, is new and its effect is not proven. We list it, but it can never make a check fail. You can read the proposal at llmstxt.org. Structured data follows the shared vocabulary at schema.org.