PDP Score
PDP Score grades every D2C and retailer-hosted product-detail page on its AI-readability. It looks at 25 criteria across five buckets — GEO, Content, Semantic, Visual, Technical — and produces a 0-100 score plus a ranked list of recommendations the merchant can act on.
What it measures
PDP Score answers one question per SKU: how AI-readable is this page? The number is calibrated against citation outcomes in Prompt Runs. Pages that get cited tend to score above 70. Pages that stay invisible tend to score below 50. The score is not a vibe — it is reverse-engineered from "what predicts citation."
The five buckets
Twenty-five criteria. Five buckets. Five criteria each. The buckets map to the five signal types AI engines actually weight. See AI visibility for the underlying mechanism.
| Bucket | Criterion | What it checks |
|---|---|---|
| GEO | Crawler allow rules | robots.txt permits GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, Applebot-Extended. |
| llms.txt presence | Valid llms.txt at the root with product-relevant sections. | |
| AI Agent Access | Each bot returns 200 with content (not 403 / not blocked). | |
| Robots meta tag | No noindex or noai directives that conflict with crawl rules. | |
| Sitemap inclusion | The PDP URL appears in a referenced sitemap. | |
| Content | Title clarity | Title contains brand, product name, and at least one differentiator. |
| Description completeness | Description covers what it is, who it is for, and at least one feature in plain prose. | |
| Bullet specificity | Bullets quote concrete specs and use cases, not marketing adjectives. | |
| FAQ presence | At least four shopper-language Q&A pairs visible in text. | |
| Review aggregation | Review count, star rating, and at least three quoted reviews visible in plain text. | |
| Semantic | Product JSON-LD validity | Valid Product schema with name, image, description, brand. |
| Offer schema | Offer with price, priceCurrency, availability, and itemCondition. | |
| AggregateRating | AggregateRating with reviewCount and ratingValue. | |
| BreadcrumbList | BreadcrumbList connecting the PDP to category and home. | |
| Entity consistency | Brand name spelled identically across title, schema, alt text, OG tags. | |
| Visual | Image alt text | Each product image has descriptive alt text including product name. |
| Image filename | Image filenames contain the SKU or product name (not IMG_4302.jpg). | |
| Open Graph image | og:image present at recommended dimensions. | |
| Twitter Card image | twitter:image present. | |
| Image count | At least three distinct product images, including one feature shot. | |
| Technical | Response time | HTML response < 1500 ms p50 from a US data centre. |
| Mobile rendering | Content visible without JS execution; viewport meta tag present. | |
| Canonical handling | Canonical URL self-referential; no canonical chains. | |
| JS rendering | Core content is in initial HTML (not JS-dependent for retrieval). | |
| Page weight | Total page weight < 2 MB on mobile. |
Running an audit
From the Products table
Click any row. The detail view opens with the score recomputed if it is older than seven days, served from cache otherwise.
From a Shopify webhook
If the Shopify integration is connected, eCommerce Insights re-audits the SKU when product:update fires. The audit runs asynchronously and the score is updated on the table.
Bulk
Settings → Catalog → Re-audit all. Useful after a site-wide template change. Runs in batches of 25 with backoff.
On a schedule
Configure in Scheduler. Daily / weekly / monthly cadences supported.
Reading the result
The detail view stacks the five buckets vertically. Each bucket shows its composite (0-20) and the five criteria with pass / partial / fail markers. Click any criterion to expand the underlying observation (e.g. for "Image alt text" you see the actual alt-text strings the audit found).
The header strip shows three numbers: total score, delta since last run, and the timestamp of the last successful audit. Hover the delta to see a sparkline of the last twelve runs.
Recommendations
Below the bucket detail is the recommendation list. Each recommendation has three properties.
- Priority
- High / medium / low. High means the criterion has high predicted citation impact and is currently failing. Low means it is a nice-to-have or the criterion has small impact.
- Effort
- 5 min / 15 min / 1 hr / multi-day. Effort is computed from the criterion type — JSON-LD updates are 15 min, robots.txt edits are 5 min, image regeneration is multi-day.
- Action
- The actual change to make. For content recommendations, the action links to the Listing rewrite capsule with the proposed diff. For technical recommendations, the action gives the exact line of code.
Re-running and tracking drift
The Score column on the Products table shows the latest score with a small trend indicator. Open the detail to see the twelve-run sparkline. Drift downward is usually a signal that the underlying PDP template changed or a metafield got cleared during an inventory sync.
Common questions
What is a passing PDP Score?
How long does a PDP audit take?
Does PDP Score apply to Amazon SKUs?
Can I customise the bucket weights?
What if a PDP is blocked by Cloudflare?
Ask AI about this page
Open this topic directly in your favourite AI assistant — the prompt is prefilled.
Try this yourself
See eCommerce Insights on your own catalog. The free trial covers your full SKU set across five AI engines.
Start free trial