How the six AI engines differ
Rubric scores six AI engines, ChatGPT, Perplexity, Google AI Overviews, Gemini, Microsoft Copilot and Claude, and they differ in three ways that change how you optimise: the pool of sources each draws on, whether it runs JavaScript before reading a page, and how it shows citations. A seventh engine, Grok, is advisory only and is not scored.
- →The six scored engines are ChatGPT, Perplexity, Google AI Overviews, Gemini, Microsoft Copilot and Claude.
- →They differ on three axes: source pool, whether they run JavaScript, and citation style.
- →ChatGPT, Perplexity and Claude do not run JavaScript; Google AI Overviews does; Copilot does so unreliably. Grok is advisory only.
The six engines at a glance
Each engine draws on a different source pool, treats JavaScript differently, and shows citations in its own way. The table sets the six side by side. Treat the source pool and citation style as broad descriptions rather than fixed guarantees, because engines change how they retrieve and display sources over time.
| Engine | Source pool (broadly) | Runs JavaScript? | Citation style |
|---|---|---|---|
| ChatGPT | Its own search and crawl (GPTBot, OAI-SearchBot) plus web results | No | Inline links, with a list of cited sources |
| Perplexity | Live web search and its own crawl (PerplexityBot) | No | Numbered inline citations, shown prominently |
| Google AI Overviews | Google's search index | Yes | Links to supporting pages beside the overview |
| Gemini | Google Search and Google's index (Google-Extended) | Yes | Sources surfaced via Google Search grounding |
| Microsoft Copilot | Bing's index | Unreliably | Numbered footnote-style citations |
| Claude | Live web search and its own crawl (ClaudeBot) | No | Inline cited sources |
Why the source pool matters
The source pool is the set of pages an engine can draw on to answer, and it differs by engine. Copilot leans on Bing's index, Google AI Overviews and Gemini on Google's index, and ChatGPT, Perplexity and Claude on their own crawls combined with web search. Because the pools differ, a page can be cited strongly by one engine and missed by another. The practical takeaway is to make sure every engine's crawler can reach your pages: allow GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended and Bingbot, and keep your pages in an XML sitemap. See getting-cited-by-chatgpt for the crawler-access detail.
Why JavaScript rendering matters
Whether an engine runs JavaScript decides what it can see on your page. ChatGPT, Perplexity and Claude do not run JavaScript, so they read the raw HTML your server returns and never execute page scripts. Google AI Overviews renders JavaScript, and Microsoft Copilot renders it unreliably. If your content or schema is injected by JavaScript, half the scored engines cannot see it. This is why "schema readable without JavaScript" is a scored check and why render parity is a recurring theme. Read why-ai-crawlers-dont-run-javascript for the fix.
Why citation style matters
Citation style is how an engine shows the sources behind its answer, and it shapes whether a citation sends you a click. Perplexity and Copilot show numbered citations prominently, so a named source is visible and clickable. AI Overviews and Gemini surface supporting links beside a synthesised answer. ChatGPT and Claude cite inline. In every case the answer is often complete on its own, so the value of a citation is as much about being named and trusted as it is about the click.
Why Grok is advisory only
Grok is surfaced as an advisory note rather than one of the six scored engines. Its most distinctive signal is heavy weighting of recent posts on X, which is off-page and cannot be measured by an on-page audit. Rubric only scores signals it can actually see on your pages, so folding Grok into the score would mean guessing at data Rubric does not have. It is flagged for awareness so you know it exists, without inventing a number.
How Rubric weights the six engines
Rubric weights the six engines because each values the same signal differently. Every one of the 44 checks carries a per-engine weight, so a page's readiness can be read engine by engine. Schema, for example, is weighted more heavily for some engines than others. The overall page score is the weighted result across all six, not a simple average. To see how the weighting is set and why it changes the priority order of your fixes, read how-engine-weightings-are-set.