How engine weightings are set
Rubric scores six AI engines, ChatGPT, Perplexity, Google AI Overviews, Gemini, Microsoft Copilot and Claude, and each of the 44 checks carries its own weight for each engine. The engines weight the same signal differently because they retrieve and render pages differently. The clearest split is JavaScript: three of the six do not run it at all.
- →Each of the 44 checks carries a per-engine weight, so the same signal can matter more to one engine than another.
- →Render parity is the biggest divider: ChatGPT, Perplexity and Claude do not run JavaScript, Google AI Overviews does, and Microsoft Copilot does so unreliably.
- →Grok is not scored. Its main signal is off-page, so it appears only as an advisory note.
Why do the six engines weight the same signal differently?
The six engines weight the same signal differently because each check carries its own weight for each engine, and the engines value signals differently. They draw on different source pools, render pages differently and cite in different styles, so a signal one engine leans on heavily is one another barely uses. Schema, for example, is weighted more heavily for Gemini than for some other engines. Rubric holds a separate weight per engine for every check, which is why the report can show how a single page reads to each engine rather than giving you one blended number with no explanation.
How does JavaScript rendering split the engines?
Render parity is the match between what a browser shows after running JavaScript and what an engine sees when it does not run JavaScript, and it is the single largest source of divergence between the six engines. If your schema or key content is injected by a script, it is present for an engine that renders JavaScript and absent for one that does not. That one difference reweights a whole cluster of checks at once: for the engines that do not render, "schema readable without JavaScript" and the content it describes effectively fail, no matter how good the markup looks in your browser.
Which engines run JavaScript?
The table below shows how the six scored engines handle JavaScript and what each means for how you build a page.
| Engine | JavaScript rendering | Practical rule |
|---|---|---|
| ChatGPT | Does not run it | Server-render everything you need seen |
| Perplexity | Does not run it | Server-render everything you need seen |
| Claude | Does not run it | Server-render everything you need seen |
| Google AI Overviews | Renders it | JS-injected content can still be read |
| Microsoft Copilot | Renders it unreliably | Do not depend on JS rendering |
| Gemini | Renders it (a Google engine) | JS content can be read, but server-render to be safe |
The pattern is clear enough to act on without knowing every weight: at least three of the six do not run JavaScript, so anything you need an engine to read should be present in the served HTML.
Why isn't Grok scored?
Grok is not one of the six scored engines because its distinctive signal is off-page. Grok weights recent posts on X heavily, and an on-page audit like Rubric's cannot see or measure activity on X. Scoring Grok would mean inventing a number for something Rubric cannot observe, so instead Grok is surfaced as an advisory note. This is a deliberate honesty rule: Rubric only scores what it can actually read on the page, and it flags the rest as context rather than folding a guess into your citability score.
How does per-engine weighting show up in your report?
Your report can show how a page reads to each of the six engines, so you can see which engine a page is weakest for and why. In practice, though, the fixes that help the strictest engines help all six. An engine that does not run JavaScript, wants sourced facts and needs a clear entity is the hardest to satisfy, so building for that reader tends to lift the page across the board. Treat the per-engine view as a diagnosis, not a set of six separate to-do lists, and only chase a single engine when your measured citations show it dominates your traffic.