Kuroma
Page Performance Indicators
What this check reads on your page
A page that returns a 4xx or 5xx status scores zero here, and a server response slower than three seconds forfeits a quarter of the factor. Otherwise the check counts what a crawler has to wait for before there is a page: script tags with a src and no async, defer or module attribute, plus stylesheet links other than print ones, graded on a four-step ladder. It then divides the number of HTML elements by the number of words of visible content, so a page that spends thousands of elements on a few dozen words is graded as markup a crawler has to parse per answer. No rendering, Lighthouse run or Core Web Vitals measurement is involved.
Our verdict on the evidence
Deliberately down-weighted Speed correlations with AI visibility are weak and driven by extreme outliers; no platform names speed as a factor, so this stays a low weight and a coarse ladder. A fetch that fails scores zero. Otherwise the page is graded on what a crawler has to do to read it: how many scripts and stylesheets block the page appearing, and how much markup it spends per word of content.
Cross-engine citation-driver study, deep vs shallow content and freshness gatekeeping (arXiv 2605.25517) · Vercel and MERJ AI crawler study (569M GPTBot requests: zero JavaScript execution, no image fetches)
This number is not a speed measurement and will not track your Lighthouse score: a page can be slow for a visitor and score well here, or fast and score badly because it wraps each word in three divs. Read it as a cost per answer. The check asks what a fetcher that runs no JavaScript has to download and parse before it reaches the text, because that is the only sense in which speed reaches an AI crawler at all.
What it is worth
10% of Technical Optimization, which is 2 points of the overall 100.
How audited pages score on it
Based on 30,738 audited pages.
The median page scores 65, and the middle half of pages score between 51 and 78.
- Good (75 to 100)8,508 (28%)
- Moderate (50 to 74)15,958 (52%)
- Needs work (25 to 49)5,531 (18%)
- Critical (0 to 24)741 (2%)
20% of audited pages score below 50 on it.
Where it fails most and least
Industries that fail it most
- Marketplace: 26% of 211 audited pages fail it
- Enterprise Software / ERP: 26% of 114 audited pages fail it
- Restaurant Tech: 25% of 481 audited pages fail it
- Medical Devices: 24% of 190 audited pages fail it
- Retail Tech / POS: 24% of 100 audited pages fail it
Industries that fail it least
- AdTech / Advertising: 7% of 138 audited pages fail it
- Banking / Neo-banking: 9% of 363 audited pages fail it
- AI / Machine Learning: 12% of 477 audited pages fail it
- FinTech / Payments: 14% of 201 audited pages fail it
- Supply Chain / Logistics: 14% of 359 audited pages fail it
How to fix it
- Add defer or async to every external script that is not needed to draw the page, or load it as type=module.
- Merge stylesheets so that scripts and stylesheets together leave no more than two blocking requests, and mark print styles with media=print.
- Cut wrapper elements: aim for no more than one HTML element per word of content.
- Make sure the page answers in under three seconds and never returns an error status to a crawler.
What has changed in this factor
- Rubric version 2026-07-07: weight in its category moved from 25% to 10%