Crawlability and indexability first
Start with whether search engines can reach the page at all: check robots.txt for accidental blocks, confirm important pages aren't set to noindex by mistake, and verify the XML sitemap actually lists the URLs you want found (and doesn't list ones you don't).
Then check for duplicate-content traps — multiple URLs serving the same content without a canonical tag pointing to one version — since that alone can split ranking signals and confuse which page an AI system should even consider citing.
Site speed and Core Web Vitals
Every crawler, human or automated, has to load and render a page before it can evaluate anything else about it. Slow or janky-to-render pages are worse candidates by default, independent of content quality.
Run the page through PageSpeed Insights or a CrUX-based tool and check the three Core Web Vitals — LCP (loading), CLS (visual stability), and INP (responsiveness) — against Google's public thresholds, not just a subjective "feels fast" impression.
Structure, internal linking, and structured data
Check that the site's information architecture actually reflects importance — a flat structure where every important page is a few clicks from the homepage, with internal links using descriptive anchor text rather than generic "click here."
Finish with a structured-data pass: is schema.org markup present and valid (Article, Organization, FAQPage where relevant), and does it match what's actually on the page — mismatched or missing schema is one of the most common, cheapest-to-fix findings in a real audit.