The table-stakes technical checklist

The checklist

None of the AI-era tactics matter if the classics fail — AI Overviews cite ranking pages, and agents act on working ones. Run this list, with Claude, before and after any GEO push.

1. Core Web Vitals. Targets: LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1 — and CLS now counts double: for users, and for agents clicking moving targets.
2. Mobile-first and HTTPS. Indexing is mobile-first; a valid certificate is assumed.
3. Fast TTFB and clean crawl. Server response under ~0.8s; no crawl traps, redirect chains, or orphan pages.
4. Sitemaps — XML plus image/video. Honest lastmod values, declared in robots.txt.
5. hreflang if international. Reciprocal pairs, an x-default, no mixed signals.
6. Structured data that validates. Rich Results Test plus the schema.org validator.
7. Local: Google Business Profile + NAP. For local intent, a claimed profile and consistent name-address-phone everywhere — Google's own AI guidance calls local out explicitly.
8. Accessibility. Labeled controls and a sane accessibility tree — compliance for people, parseability for agents.

Validation is half the checklist

Lighthouse (including the Agentic Browsing category), Search Console's coverage and enhancement reports, the Rich Results Test, and the schema.org validator between them catch most of what silently breaks. Re-run after every meaningful deploy, not once a year.

I'm pasting my Lighthouse/PageSpeed output, my sitemap, and my robots.txt. Produce a ranked fix list: every failing table-stakes item (Core Web Vitals, mobile, HTTPS, TTFB, sitemap hygiene, hreflang, schema validation, accessibility) with severity, the specific fix, and which files or pages it applies to. Order by impact on classic ranking first, agent usability second.