AI Search Optimizer

AI Search Optimizer

Overview

Be the answer AI gives, not just a result it lists. AI search platforms synthesise answers and cite sources — your content must be crawlable, extractable, authoritative, and fresh. This skill covers HOW AI finds and recommends your products. It does NOT cover human conversion psychology (see conversion-psychology) or page structure (see ecommerce-growth).

Boundary: product / page AI visibility → here; a person's AI findability (how ChatGPT/Perplexity/Gemini answer "who is ") → career-online-presence; an AI agent transacting on the buyer's behalf (ACP / AP2 / delegated payment / llms.txt) → agentic-commerce-readiness.

Discovery vs transaction: this skill gets you recommended, so the buyer lands on your site and checks out normally. That is the proven, measurable half. Agents completing purchases without a site visit is a separate and far less mature surface — do not fund it from this skill's evidence.

Top 5 Essentials

1. Allow AI search bots, block AI training bots in robots.txt
2. Complete Product schema on every product page (empty schema = 18% citation penalty)
3. Answer capsule — first 40-60 words of key pages must directly answer the primary query
4. Q&A / FAQ content on product and category pages — for AI extraction, not for a SERP rich result (Deprecated)
5. Google Merchant Center feed — kept current with real stock/pricing

How AI Platforms Find Products

Answer Engine Optimization (AEO)

The Answer Capsule

Every key page needs a direct answer in the first 40-60 words after the H1:

BAD: "Welcome to GamingBuilds! We offer a wide range of gaming PCs..."
GOOD: "The best gaming PC for Fortnite under £1,000 is the Titan X,
featuring an RTX 5060, Ryzen 5 9600X, and 32GB DDR5. It delivers
144+ FPS at 1080p on competitive settings with a 1TB NVMe SSD."

Content Formats AI Prefers

Formatting Rules

• Clear H2/H3 headings with descriptive text
• Most important answer FIRST (inverted pyramid)
• Consistent data formats (always GBP, always "X cores / Y threads")
• Original data: benchmarks, FPS figures, test results
• Content must render in initial HTML (AI crawlers don't execute JS)

Conversational Query Strategy

Content strategy: Build Q&A content around natural questions. Target "best gaming PC for [game] under [budget]" pages, GPU comparison tables, and buyer's guides with FAQ schema.

Schema Checklist

Priority here is ranked for AI CITATION, which is not the same ranking as SERP presentation. For schema status — whether a type is alive, deprecated, or removed — seo-structure-architect is the canonical owner; this table deliberately does not restate it. That split is why FAQPage once carried three different verdicts across three skills.

Deliberately NOT using CRITICAL/HIGH/MEDIUM/LOW. Those words are seo-structure-architect's SERP-impact vocabulary. Reusing them for a different axis produced three false contradictions the moment this table was written — BreadcrumbList read MEDIUM here and HIGH there, which is indistinguishable from disagreement even though both were correct on their own axis. When two skills rank the same object on different axes, they must not share a verdict vocabulary.

AI-1 = ship before anything else · AI-2 = ship next · AI-3 = keep, don't prioritise. These tiers rank for citation only and say nothing about SERP value or schema status.

Implement the full set from seo-structure-architect. This table ranks; it does not enumerate. Two types (MerchantReturnPolicy, WebSite) were missing here entirely until 2026-07-26, so anyone following only this skill shipped an incomplete Offer.

Schema Rules

• Never implement empty schema — 18% citation penalty vs no schema
• Populate ALL relevant attributes before adding
• JSON-LD format (preferred by Google and AI systems)
• Validate with Google Rich Results Test

GTIN/MPN for Custom PCs

Custom-built PCs don't have GTINs. Use mpn (Manufacturer Part Number) with your own SKU system. Set identifier_exists: false in product feeds. Google and AI systems accept MPN as alternative for custom/assembled products.

AI Crawler Management

robots.txt — Allow Search, Block Training

# AI SEARCH bots (ALLOW — these power answers users see now)
User-agent: ChatGPT-User
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Claude-SearchBot
Allow: /

# AI TRAINING bots (BLOCK — no search benefit)
User-agent: GPTBot
Disallow: /
User-agent: ClaudeBot
Disallow: /
User-agent: CCBot
Disallow: /
User-agent: Meta-ExternalAgent
Disallow: /

# Google-Extended — NOT a search bot. Judgement call; default ALLOW. See below.
User-agent: Google-Extended
Allow: /

Key distinction: Search bots power real-time results users see. Training bots collect data for model training — blocking them doesn't affect your search visibility.

Google-Extended is the exception to that rule — read before copying

Google-Extended is not a crawler and has no user-agent string — it is a robots.txt product token only. Per Google's own crawler documentation it does not affect Google Search inclusion, is not a ranking signal, and does not affect AI Overviews eligibility. Filing it under "search bots that power search results" is wrong.

It is the one token where allow search / block training conflicts, because it controls both:

Default recommendation: ALLOW — you keep Gemini-app grounding, and the cost is consenting to Gemini model training. Block it only if you have a deliberate no-training policy, and know you are trading away Gemini grounding to get it. Either way, do not expect any Search or AI Overviews effect in either direction.

Product Feed Priority

1. Google Merchant Center (Priority 1) — WooCommerce feed plugin, maximise attributes, keep stock current
2. Perplexity Merchant Program (Priority 2) — Free, apply at perplexity.ai
3. Bing Merchant Center (Priority 3) — Powers Copilot shopping
4. ChatGPT feed (Priority 4) — ACP plugin for WooCommerce (verify current enrollment process)

Small team minimum viable: Start with GMC only. Add others when capacity allows. Stale feeds actively harm your brand.

Entity Building (Brand Authority)

AI checks multiple independent sources before recommending ("consensus signal"):

Monitoring AI Visibility

Manual testing (free): Ask each platform product questions and check citations:

• "Best gaming PC shop near Rotherham"
• "Best pre-built gaming PC under £1000 UK"

Analytics tracking: Monitor referrals from chatgpt.com, perplexity.ai, copilot.microsoft.com, claude.ai, gemini.google.com

⚠️ Use chatgpt.com — not chat.openai.com, which OpenAI retired in 2024. A segment still filtering on the old domain returns zero and reads as "AI sends us no traffic" rather than as a broken filter.

Judge these segments on CVR and AOV, not session volume. AI referral volume is small by construction (zero-click discovery leaves no referrer at all), so sessions understate the channel. The instrument that actually sees dark/zero-click discovery is a post-purchase survey — "How did you first hear about us?" — plus branded-search lift in Search Console.

Deprecated — Do NOT Implement

This surface changes faster than any other in the SEO family. Check this table before acting on any recommendation in this skill, and add a row the moment something dies — a dead recommendation left at CRITICAL priority costs more than a missing one.

Do not read the Instant Checkout row as "agentic commerce is dead" — the product failed, the protocol standardised. ACP's stable spec is dated 2026-04-17 (after the shutdown) and AP2 went to the FIDO Alliance on 2026-04-28 with 60 organisations. See agentic-commerce-readiness for the standards layer and what a merchant actually implements. What died was one consumer surface, not the category.

Why this table exists. Four defects were found in this skill in one review (Jul 2026), all of the same kind: correct-when-written guidance that the world moved past. Recency of authorship is not the risk factor — rate of change of the subject is. Any skill covering a fast-moving surface should carry a table like this.

Anti-Patterns