The AI-readiness score methodology (0-100): five categories and weights
Updated 2026-08-03
What the AI-readiness score measures
The AI-readiness score is a single number from 0 to 100 that shows how easily an AI assistant — ChatGPT, Perplexity, Google AI Mode — can find your store, understand your products and recommend them to a shopper. The free Botmetria audit builds that number from five categories, each with its own weight. The weights are not arbitrary: they mirror what actually decides whether your product lands in an AI answer.
Across our data — 63 stores in Moldova and Romania — the market median is 38 out of 100. Half of all shops do not even reach the middle of the scale. Here is what each category checks and why the weights sit where they do.
The five categories and their weights
| Category | Weight | What it checks |
|---|---|---|
| Discovery | 20% | robots.txt, sitemap.xml, llms.txt |
| Product data | 30% | JSON-LD Product: price, currency, availability |
| Logistics | 15% | shipping and returns readable to AI |
| Hygiene | 15% | response codes, images, speed |
| Agentic Markdown | 20% | whether the site serves clean text/markdown |
Discovery — 20%
Checks the basics: can bots find your site at all. We look at robots.txt (is AI-bot access blocked by mistake), sitemap.xml (is there a map for crawlers) and llms.txt (a curated map for language models). If these files are missing, or robots.txt blocks useful bots by accident, the model never reaches your catalog — and the other categories stop mattering. Building an llms.txt is covered in a separate guide.
Product data — 30%
The heaviest category. We check the JSON-LD Product markup from schema.org: does offers carry price, priceCurrency and availability (InStock/OutOfStock). These three fields are exactly what an AI agent reads to quote a price and stock to the shopper. No structured price, and the assistant either stays silent about your product or quotes a competitor who did the markup.
Logistics — 15%
Whether shipping and returns are readable to AI. Assistants often answer not only "where to buy" but also "how fast does it ship" and "can I return it". If those terms are baked into an image or a PDF, the model cannot see them.
Hygiene — 15%
The technical health of the site through a bot's eyes: response codes, speed, image availability. This is where the market's single most common problem lives: product images return 403 or 406 to bots. A human sees the photo, an AI agent gets refused — and the product shows up in the answer with no image.
Agentic Markdown — 20%
Whether the site can serve clean text/markdown on request instead of heavy HTML. Agents read Markdown faster and cheaper, and more tools now ask for it specifically. This is less about SEO and more about agentic commerce — when the AI does not just recommend but places the order itself.
Why product carries the most weight
The weights answer one question: what decides whether your product lands in an AI answer, at the right price.
- Product data — 30%, because price and availability are the reason an assistant mentions a product at all. Without a structured price you effectively do not exist.
- Discovery and Markdown — 20% each: if a bot cannot find or cannot read the site, it never reaches the product.
- Logistics and Hygiene — 15% each: important, but refinements on top of the essentials. Returns and speed shape the answer less than the price itself.
The five add up to 100. A heavier category moves the total more — so fix product first, rather than chasing a couple of points in logistics.
How to read the score bands
The final number falls into one of three bands:
- ≤ 40 — red. The store is nearly invisible to AI. The market median (38) sits right here, so this is the norm, not a verdict.
- 41–70 — yellow. The basics are in place but with gaps — usually
productorhygieneis dragging. - > 70 — green. The store reads well to AI assistants and is ready for agentic scenarios.
The same score powers the MD/RO store rating, the AI traffic index and the PDF audit — one methodology, so the numbers are directly comparable.
How to find your score
Run the free URL audit: in about a minute it returns a 0-100 number, the breakdown across five categories and a list of concrete issues — from a blocked robots.txt to images returning 403. Start with product (weight 30%) and hygiene (those 403/406 photos) — usually the fastest way to pull a score out of the red band.
FAQ
Why does "Product data" weigh 30% instead of an equal share?
Because price and availability decide whether an AI assistant names your product. Without a structured price in JSON-LD the model stays silent or quotes a competitor. Discovery and readability matter, but they only lead to the product — the product itself decides.
What does a market median of 38 out of 100 mean?
Of 63 stores in Moldova and Romania, half score 38 or lower — inside the red band. A low start is the market norm, which is exactly why even small fixes make a store stand out fast.
Is the score calculated the same way in the audit, the rating and the PDF?
Yes. The methodology and weights are identical for the free audit, the MD/RO rating and the PDF report, so the numbers are directly comparable.
Related guides
Check your store for free
A 60-second audit shows how ready your store is for AI agents — and exactly what to fix first.