Skip to content
Primores AI wiki

AI-Era SEO — Overview

AI-Era SEO

TL;DR: This domain covers what “SEO” means once AI answers sit between your content and your buyer. Three facts reframe everything: most searches now end without a click (~47% CTR collapse under AI Overviews, Pew-measured; vendor estimates run higher), AI surfaces are separate discovery layers (only ~13.7% citation overlap with classic rankings — ranking #1 no longer implies being cited), and the query being answered isn’t the one the user typed (engines fan it out into synthetic sub-queries). The response isn’t a tactic — it’s an operating model: be citable (two gates: retrieved and cited), be crawlable (access control is now a policy decision), and run a production system (seo/articles-engine) instead of publishing ad hoc.

The shift, in three numbers

  1. ~47% — the measured CTR collapse when an AI Overview appears (Pew Research clickstream, primary; the widely-quoted 64.82% zero-click figure is a flagged vendor estimate). Strategy consequence: seo/zero-click-strategy — brand-and-visibility-first, PR strategy is SEO strategy.
  2. 13.7% — citation overlap between AI answers and top-10 rankings across Ahrefs’ 1B+ data-point studies (seo/ahrefs-ai-search-studies-2026). AI search is a separate discovery layer, not a summary of the SERP.
  3. ~2/3 — the share of ChatGPT prompts that get rewritten into multiple synthetic sub-queries before retrieval (glossary/query-fan-out). You optimize for the model’s reformulations, not the user’s phrasing.

The visibility stack (core pages)

Measurement

Access control — the flip side

Visibility assumes crawlability, and crawlability is now a policy decision: seo/ai-crawler-access covers the training / retrieval / user-fetch bot taxonomy, why a WAF enforces where robots.txt only asks, Cloudflare’s default-block + pay-per-crawl, and current UA strings. Blocking training bots while allowing retrieval bots is a coherent position — but only if you know which is which.

The production system

Individual tactics don’t compound; a system does. seo/articles-engine is the six-stage loop (map → produce → publish → interlink → measure → refresh) that turns the pages above into an operating cadence — with the measured freshness bias (updating counts nearly as much as publishing) and Google’s spam line (purpose and user value, not production method) as guardrails. Upstream of it: glossary/topical-authority and the niche-selection logic in seo/new-site-ranking.

The creative reverse-engineering cluster

A self-contained methodology cluster lives in this domain: deconstructing winning ads into reusable templates — seo/ai-creative-reverse-engineering-complete-methodology (pillar), with the craft distinction (seo/surface-vs-structural-mimicry), when it beats briefs (seo/ai-reverse-engineering-vs-creative-briefs), when not to use it, team structure, ethics, and seo/meta-ad-library-api for sourcing. It connects outward to the marketing domain’s production system (marketing/static-ad-template-system).

Getting started

  1. Audit — run tools/ai-visibility-audit against your domain; check how AI currently describes you.
  2. Unblock — verify crawler access (seo/ai-crawler-access); publish llms.txt; fix rendering.
  3. Systematize — stand up the seo/articles-engine loop on a niche you can own (seo/new-site-ranking).
  4. Measure — track citations and share-of-model, not just rankings (seo/geo-aeo-benchmarks-2026).

Key Takeaways

  • AI search is a separate discovery layer — classic rank and AI citation overlap weakly; optimize both, measure both.
  • The two-gate chain (fan-out → retrieval → citation) means titles and coverage target the model’s rewrites, not the typed keyword.
  • E-E-A-T is a binary filter and brand mentions outpull backlinks — PR and SEO have merged.
  • Crawler access control is strategy, not ops trivia: know which bot types you’re blocking and why.
  • Run a production loop, not one-off posts — freshness is measured to matter, and updating existing pages captures much of it.

Sources

This is a navigation hub — sources live on the linked pages, each carrying its own credibility grading.

By Andrej Ruckij