Eleven checks. Every weight published.
75 points for the five AI checks, 25 for the six search basics. Every grade is a fact about your store on the day of the scan. Same captured inputs, same method, same result, so two runs can be set side by side.
How a score is worked out
Compare your score only with your own earlier scores. It is not a ranking against other shops, and it does not predict whether an assistant will mention you.
AI checks
5 checks · 75 pointsWhether the automated visitors that read your site for AI tools are let in, or turned away by your host, your security plugin or anything in between without telling you. Nothing else matters if this one fails.
Whether price, stock, sizes and product codes are written on your product pages in a form software can read, and whether they match what a shopper sees.
Whether your shop has a small llms.txt file that says in plain words what you sell, how it is grouped, and what your policies are. It saves an assistant from guessing its way around your pages.
Whether your product and category pages state materials, sizing, delivery and returns in words rather than inside a picture.
Whether your page arrives quickly, and shows its detail without waiting on code that these automated visitors do not run.
Search basics
6 checks · 25 pointsWhether each page says what it is, once, in the short title and summary that search engines and assistants read first.
Whether your shop has a sitemap.xml, the file that lists all your pages, whether robots.txt points to it, and whether it matches the products you sell today.
Whether your security certificate is valid, every page uses the secure https address, and the older addresses jump straight there in one step.
Whether the different versions of a page, the ones with sizes, filters or tracking tags in the address, all point back to one main address. Otherwise one product can be read as five.
Whether each page has one main heading that names what is on it, with smaller headings in order underneath.
Whether every image that means something carries a short written description. It helps shoppers using a screen reader first, and software second.
The points are fixed and published here. They follow what stops an assistant first, not what is easiest to sell. We check that the search basics are in place. We do not track keywords, positions or rankings, and we never report on them.
A worked example
The same sums the scanner does. You can check them with a calculator.
| Check | Weight | Grade | Points |
|---|---|---|---|
| AI assistant access | 26 | Needs work | 13.0 |
| Product details | 19 | Broken | 0.0 |
| Shop summary | 13 | Broken | 0.0 |
| Answers on the page | 9 | Needs work | 4.5 |
| Page speed | 8 | Good | 8.0 |
| Titles and descriptions | 5 | Needs work | 2.5 |
| One main web address | 4 | Good | 4.0 |
| Headings | 3 | Broken | 0.0 |
| Image descriptions | 3 | Needs work | 1.5 |
| Page list (sitemap.xml) | 5 | Good | 5.0 |
| Secure address | 5 | Good | 5.0 |
Step 1. Add the AI-check points: 13.0 + 0.0 + 0.0 + 4.5 + 8.0 = 25.5 of 75
Step 2. Add the search-basics points: 2.5 + 4.0 + 0.0 + 1.5 + 5.0 + 5.0 = 18.0 of 25
Step 3. All eleven checks gave a result, so the points counted add up to 100: 25.5 + 18.0 = 43.5, which rounds to 44, hard to read
What a score can and cannot say
Every report sorts its evidence into the same three layers, so it is always clear which part is measured and which part is not claimed at all.
Technical facts
What the checks measured on your public pages on the day of the scan. These are facts about your store, they are the only thing we will ever promise to do, and you can measure them again yourself.
Observed visibility
What assistants actually said when we asked them, on the dates given. That is a dated observation with a written method behind it, not a forecast, and it is not promised to repeat.
Business outcome
Not promised. Nothing here predicts rankings, mentions, visits or sales, and no number in a report should be read as a promise of any of those.
What we will never claim
What we do not measure
Keyword rankings, traffic, or whether any assistant mentions you. Nobody can measure that honestly, so we do not report on it.
See these checks run on your store
Every finding in your report shows the points it carries and the grade it earned, so you can add them up yourself.
Method changes
2026.09 (6 September 2026)
Monitoring comparisons restart at a method change.
- Assistant access is graded on sampled public URLs.
- Training crawlers (GPTBot, ClaudeBot, Google-Extended) are reported on their own unscored line; the access check scores the six search and user-fetch crawlers. Google-Extended also controls Gemini grounding; it does not affect Google Search inclusion or ranking.
- Product prices need a currency. A Product record without a name now fails.
- The sitemap check passes when any of /sitemap.xml, /sitemap_index.xml, /wp-sitemap.xml or a robots.txt Sitemap directive parses.
- Duplicate-record and non-standard-wrapper warnings count toward the product details credit.
- FAQ page data needs at least one question with an answer.
- Product discovery works across store platforms; product pages nested up to three levels deep and Shopify ProductGroup records are recognised.
- One linked help or FAQ page is sampled.
2026.08 (published weights, first version)