google-ai-search

name	google-ai-search
description	Use when a website brief promises Google AI Overviews, AI Mode, AEO/GEO, Search Console AI measurement, local or ecommerce AI readiness, or agentic experience preparation; use seo for implementation after guidance.
metadata	{"portable":true,"compatible_with":["claude-code","codex"]}

Google AI Search

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, [phone].

Translate current official Google Search guidance into qualified website requirements without promising AI feature inclusion.

Use When

• A brief mentions AI Overviews, AI Mode, AEO, GEO, agentic search, or Search Console AI reporting
• claims need official-source verification.

Do Not Use When

• Use seo for routine implementation or unsupported tactics that claim guaranteed AI citations.

Required Inputs

Artefact	Source or provider	Required?	Purpose	If absent
Specific AI-search claim, target market, website type, existing search plan, and current official Google sources	Brief, client, and verified official documentation	yes	Separate supported requirements from myths	Stop when current official guidance cannot be verified

Workflow

1. Frame the claim and verify current official Google sources
2. Distinguish standard search requirements, feature-specific availability, measurement limits, and unsupported AEO/GEO claims
3. Produce qualified requirements for content, entities, commerce/local data, and agentic readiness
4. Stop on unverifiable or guaranteed-outcome language, recover by narrowing the claim, then hand implementation to seo.

Outputs

Artefact	Consumer	Acceptance condition
Google AI Search guidance brief	SEO, content, commerce, local, and measurement owners	Every recommendation cites current official guidance and labels availability or measurement limits

Evidence Produced

Evidence	Consumer	Acceptance condition
Official-source register and claim disposition table	Client and release owner	Each claim is supported, rejected, or marked not assessed with source date

Capability Contract

Read, search, and network verification are required for current guidance. Planning is read-only by default. Website edits, Search Console changes, publication, spending, and account access require explicit authority.

Degraded Mode

Without network or current official sources, return the narrowest qualified list of questions and stable principles, mark current feature claims not assessed, and do not provide tactical certainty.

Decision Rules

Choice	Action	Failure or risk avoided
Claim is supported by current official guidance	Translate it into a testable requirement	Speculative AEO tactic
Brief promises inclusion or ranking	Remove the guarantee and state controllable work	Misleading commercial claim

Quality Standards

• Use official Google sources for current claims, cite at point of recommendation, distinguish availability by market, and never guarantee inclusion or ranking.

Anti-Patterns

• Presenting AEO as a separate technical standard. Fix: map work to documented Search fundamentals and specific eligible features.
• Guaranteeing an AI Overview citation. Fix: state controllable quality and eligibility work only.
• Using a third-party blog as authority for Google behaviour. Fix: verify against official documentation.
• Claiming Search Console exposes unavailable AI detail. Fix: document actual measurement limits.
• Reusing stale feature availability. Fix: verify market and date before advising.

Worked Example

If a proposal guarantees AI Overview placement, replace it with a scoped programme for crawlability, helpful content, entity clarity, merchant or local data, and measurement, citing current official Google guidance.

References

• Website Skills authoring standard

Preserved Domain Use Guidance

• A website must be prepared for Google's generative AI Search features, including AI Overviews and AI Mode.
• The work mentions Google AI search, AI Overviews, AI Mode, AEO, GEO, query fan-out, RAG, Search Console AI traffic, or Google agentic experiences.
• A client asks whether llms.txt, chunking, special schema, AI-specific pages, off-site mention campaigns, Reddit/UGC placement, or "AI citation hacks" are needed for Search or AI assistants.
• The work needs AI-search measurement beyond Google, including Bing Webmaster Tools AI Performance, Citation Share, Intents, Topics, or period comparison.
• The work involves brand-safety risk from user-generated content, poisoned snippets, weak sources, fabricated citations, or volatile ranking updates.
• Existing SEO/GEO recommendations need to be reconciled with official Google Search guidance.

Preserved Domain Exclusions

• The target is primarily ChatGPT, Perplexity, Claude, Bing, or a private RAG system; use seo and its broader AI-search references first.
• The task is only metadata, sitemap, hreflang, or schema implementation with no Google generative-AI decision involved; use seo.
• The user expects guaranteed inclusion in AI Overviews or AI Mode. Google does not guarantee crawling, indexing, serving, or AI-feature inclusion.

Preserved Domain Inputs

• Target pages, page roles, business model, and primary user questions.
• Current SEO baseline: crawlability, indexability, snippets eligibility, canonical/hreflang state, internal links, JavaScript rendering risks, and duplicate-content risks.
• Search Console and analytics access when performance or diagnostics are in scope.
• Local business, ecommerce, Merchant Center, Google Business Profile, product feed, or Business Agent status when relevant.

Preserved Domain Workflow

1. Load references/google-generative-ai-search-playbook.md before making recommendations.
2. Start from the Google rule: generative AI features in Search are grounded in core Search ranking and quality systems, so the work is still SEO.
3. Audit content for non-commodity value: original experience, expert point of view, useful organization, high-quality media, and satisfaction for real visitors.
4. Reject manipulative AI-search tactics: query fan-out page factories, chunking for AI, special Google AI markup, inauthentic mentions, UGC spam, poisoned snippets, and structured-data overuse.
5. Audit technical readiness: crawlability, indexability, snippets eligibility, internal links, textual availability of important content, JavaScript SEO, page experience, and duplicate-content reduction.
6. For local and ecommerce sites, verify Google Business Profile, Merchant Center, product feeds, visible product/service details, and structured data consistency with visible content.
7. Decide content controls deliberately: use Googlebot controls for Search crawling, nosnippet/data-nosnippet/max-snippet/noindex for preview control, and Google-Extended only for non-Search AI training/grounding controls.
8. Measure Google Search through Search Console's Web performance reporting plus analytics outcomes such as conversions, engaged time, subscriptions, leads, or sales.
9. When Bing/Copilot or multi-assistant visibility matters, add Bing Webmaster Tools AI Performance reporting: Citation Share, Intents, Topics, and Compare. Treat citation share as an observational KPI, not traffic share or a ranking score.
10. For editorial, FAQ, review, forum, or community-heavy work, run a source-integrity pass before publishing: primary sources first, named experts where possible, no anonymous UGC as sole proof for money or safety claims, and no copied claims from unverified AI summaries.
11. When browser agents or commerce agents matter, load references/agentic-readiness.md and produce an agent-readiness backlog.

Preserved Domain Quality Guidance

• Never present AEO/GEO hacks as Google requirements.
• Treat llms.txt as an optional broader engine/LLM handover artifact, not a Google AI Search requirement, ranking lever, or default client priority.
• Structured data must match visible page content and exist for rich-result eligibility, not as special AI Overview markup.
• Content recommendations must improve human usefulness first; AI visibility is a consequence of useful, crawlable, indexable content.
• AI-assisted content must meet Search Essentials, spam policies, and the same accuracy, quality, relevance, and disclosure bar as human-written work.
• AI-citation strategy must be source-quality strategy: cite authoritative, durable, original sources; monitor UGC surfaces for poisoning and brand-safety risk; do not attempt to manipulate assistants with planted comments.

Preserved Domain Anti-Patterns

• Creating many near-duplicate pages to capture every fan-out query variant.
• Breaking useful pages into unnatural fragments only for AI parsing.
• Rewriting copy into robotic answer blocks when the human page gets worse.
• Buying or manufacturing off-site mentions for perceived AI authority.
• Hiding weak content behind schema, Markdown files, or crawler files.
• Blocking Googlebot and expecting AI Overviews or AI Mode visibility.

Preserved Domain Outputs

• Google AI Search readiness audit.
• Content and technical implementation backlog.
• Mythbusting note for stakeholders.
• Search Console, Bing AI Performance, and analytics measurement plan.
• Local/ecommerce readiness checklist.
• Source-integrity and UGC-risk notes when the content may be cited by AI tools.
• Agentic-readiness backlog when relevant.

Preserved Domain References

• references/google-generative-ai-search-playbook.md - official Google Search guidance distilled into engine rules.
• references/technical-readiness-checklist.md - implementation checklist for crawlability, snippets eligibility, rendering, page experience, and duplicate control.
• references/content-quality-gate.md - non-commodity, people-first content review gate.
• references/agentic-readiness.md - optional browser-agent and commerce-agent preparation notes.
• ../seo/SKILL.md - downstream implementation for metadata, schema, sitemaps, robots, hreflang, and Search Console setup.
• ../premium-commercial-writing/references/seo-ai-search-writing.md - use when copy must be rewritten for usefulness, proof, entity clarity, and buyer intent.

Notes

• This skill is source-aware: re-check Google's Search Central documentation before major client policy changes because the guidance can change.
• Preserve existing project behavior unless the current task explicitly requires a change.