Generative engine optimization (GEO) for getting cited by AI search platforms — not just ranked in traditional results.
AI SEO, generative engine optimization, GEO, AI overviews, Google SGE, ChatGPT citations, Perplexity SEO, Claude citations, AI search optimization, semantic search, entity optimization, LLM visibility, AI-generated answers, structured data, schema markup, content extractability, AI citability, GPTBot, PerplexityBot, ClaudeBot, answer engine optimization
Before optimizing, settle the inputs below. When one is missing or vague and would change the result, ask about it; when it wouldn't, use a sensible default and say which one you used.
Ask about the two or three inputs that matter most for this request, not the whole list. If the user wants a result straight away, produce it and state your assumptions at the top.
Traditional SEO gets your page ranked. AI SEO gets your content cited. These are different optimization targets.
| Dimension | Traditional SEO | AI SEO |
|---|---|---|
| Goal | Rank on page 1 | Get cited in AI-generated answers |
| Success metric | Click-through rate | Citation frequency |
| Content priority | Keyword density | Answer extractability |
| Authority signal | Backlinks + domain authority | Backlinks + answer quality + attribution |
| User interaction | User clicks your link | AI extracts your answer; user may never visit |
| Content format | Long-form comprehensive | Self-contained extractable blocks |
| Optimization unit | The page | The paragraph or section |
AI systems pull content in chunks. They find the paragraph, list, or definition that directly answers a query and extract it. Your content must be structured so answers are self-contained.
Extractability requirements:
Anti-patterns that kill extractability:
AI systems do not just extract the most relevant answer — they extract the most credible one.
Authority signals in the AI era:
AI systems must be able to find and index your content.
Technical requirements:
Step 1: Bot Access Verification
Check robots.txt for AI crawler permissions:
# Search / retrieval crawlers — blocking these removes you from that platform's AI answers:
Googlebot # Google Search, incl. AI Overviews and AI Mode
OAI-SearchBot # ChatGPT search
Claude-SearchBot # Claude search
PerplexityBot # Perplexity search
# Training crawlers — blocking these does NOT remove you from AI search answers:
GPTBot # OpenAI model training
ClaudeBot # Anthropic model training
Google-Extended # Gemini training + grounding (not AI Overviews / AI Mode)
Applebot-Extended # Apple Intelligence training
If a search/retrieval crawler is blocked, that is the single highest priority fix — zero visibility on that platform until resolved. A blocked training crawler is a policy choice, not a visibility bug. See Bot Access Configuration for the full matrix.
Step 2: Citation Testing
Test top 10 target queries on each platform:
| Platform | How to Test | What to Record |
|---|---|---|
| Perplexity | Search at perplexity.ai, check Sources panel | Cited? Which competitors cited? Content format winning? |
| ChatGPT | Web browsing enabled, check citations | Same |
| Google AI Overviews | Google query, check AI Overview panel | Same |
| Microsoft Copilot | Search at copilot.microsoft.com, check source cards | Same |
| Claude | Web search enabled queries | Same |
Step 3: Content Extractability Scoring
Score each key page (0-7):
Interpretation: 0-3 = needs major restructuring. 4-5 = good baseline. 6-7 = strong.
Step 4: Competitive Citation Analysis
For each target query, document:
Step 1: Lead with the Answer
The first paragraph must contain the core answer to the target query. No preamble, no context-setting, no "In today's landscape..." openers.
Step 2: Structure Self-Contained Sections
Every H2 section must be answerable as a standalone excerpt:
Step 3: Add Extractable Content Blocks
Insert 2-3 of these per key page:
Step 4: Replace Vague with Specific
Find and replace every vague claim:
Step 5: Add Schema Markup
Implement JSON-LD in the page head:
| Content Type | Schema | Impact (as of September 2026) |
|---|---|---|
| Articles and posts | Article | Medium — Article rich result + clear author/date signals |
| Product pages | Product (+ Review snippet) | Medium — Product rich results; product comparison queries |
| Company pages | Organization | Medium — entity authority, logo/knowledge panel signals |
| All pages | BreadcrumbList | Low-Medium — Breadcrumb rich result, site hierarchy |
| Author pages | Person / ProfilePage | Low-Medium — author credibility signal |
| FAQ sections | FAQPage | Low — FAQ rich results removed from Google Search (May 2026); valid vocabulary, no Google search feature |
| Step-by-step guides | HowTo | Low — HowTo rich results removed from Google Search (2023); no Google search feature |
The visible on-page structure (Q&A headings, numbered steps) is what AI systems extract. Google states no special schema is required for AI features — do not sell FAQPage/HowTo markup as an AI-citation lever.
Step 1: Define Your Entity
Ensure your brand exists as a recognized entity across the web:
Step 2: Build Entity Associations
Connect your entity to relevant topics:
Step 3: Strengthen the Citation Chain
Create a network of credible references:
**[Term]** is [concise definition in 1-2 sentences]. [One sentence of context
explaining why it matters or how it differs from related concepts].
Place within the first 200 words. No hedging, no preamble.
Requirements for AI extraction:
Two-column or multi-column tables with clean headers:
| Dimension | Option A | Option B |
|-----------|----------|----------|
| Price | $X/mo | $Y/mo |
| Key Feature | Description | Description |
| Best For | Use case | Use case |
Explicit Q&A pairs. Questions should match natural language queries:
### What is [topic]?
[Direct answer in 1-2 sentences.]
### How does [topic] work?
[Step-by-step explanation.]
The visible Q&A structure is what matters for extraction. FAQPage markup is optional — Google no longer shows FAQ rich results (removed May 2026).
According to [Source Name] ([Year]), X% of [population] [finding].
Complete attribution is critical. Unattributed statistics get deprioritized because AI cannot verify the source.
"[Quote]" — [Name], [Role] at [Organization]
Named experts with credentials produce citable units AI systems pick up.
Google states that no special schema.org markup is needed to appear in AI Overviews or AI Mode (AI features and your website). Prioritize schema that still produces Google rich results or clarifies entities (Article, Product, Organization, BreadcrumbList); treat FAQPage and HowTo as optional vocabulary.
FAQPage Schema (optional — FAQ rich results removed from Google Search in May 2026):
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "What is [topic]?",
"acceptedAnswer": {
"@type": "Answer",
"text": "[Direct answer]"
}
}
]
}
HowTo Schema (optional — HowTo rich results removed from Google Search in 2023):
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "How to [do thing]",
"step": [
{
"@type": "HowToStep",
"name": "Step name",
"text": "Step description"
}
]
}
Article Schema (still eligible for rich results; establishes author and date signals):
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Title",
"author": {
"@type": "Person",
"name": "Author Name",
"url": "https://author-page"
},
"datePublished": "2026-01-15",
"dateModified": "2026-03-01"
}
Validate all schema at schema.org/validator before deployment.
The major AI vendors now split crawling by purpose, so search visibility and training can be controlled separately.
| Vendor | User agent | Purpose | Blocking it means | Honors robots.txt |
|---|---|---|---|---|
| OpenAI | GPTBot | Model training | Content excluded from future training | Yes |
| OpenAI | OAI-SearchBot | ChatGPT search index | Not surfaced/cited in ChatGPT search answers | Yes |
| OpenAI | ChatGPT-User | Fetches a page when a user's request needs it | User-initiated; OpenAI says robots.txt rules "may not apply" | Not guaranteed |
| Anthropic | ClaudeBot | Model training | Future content excluded from training | Yes |
| Anthropic | Claude-SearchBot | Claude search index | Reduced visibility in Claude search answers | Yes |
| Anthropic | Claude-User | Fetches a page when a Claude user asks | Claude cannot retrieve your page for users | Yes |
| Perplexity | PerplexityBot | Perplexity search index | Not surfaced/linked in Perplexity answers | Yes |
| Perplexity | Perplexity-User | Fetches a page for a user's question | Perplexity says this fetcher generally ignores robots.txt | No |
Googlebot | Google Search, including AI Overviews and AI Mode | Removes you from Google Search entirely | Yes | |
Google-Extended (robots.txt token, not a separate crawler) | Gemini training and grounding in some Google systems; also limits training of the models behind Search generative AI features | Does not remove you from Google Search, ranking, AI Overviews or AI Mode | Yes |
Sources: OpenAI crawlers, Anthropic crawlers, Perplexity crawlers, Google common crawlers.
Google AI Overviews / AI Mode opt-out: Google-Extended does not control appearance in them — they are served from Googlebot's index. Your options (as of September 2026):
nosnippet, data-nosnippet, max-snippet limit what can be shown, but also limit how the page appears in regular Search results.noindex — removes the page from Google Search entirely.# Allow AI search / retrieval crawlers (visibility)
User-agent: OAI-SearchBot
Allow: /
User-agent: Claude-SearchBot
Allow: /
User-agent: Claude-User
Allow: /
User-agent: PerplexityBot
Allow: /
# Training crawlers — Allow or Disallow per your content policy;
# this choice does not affect AI search citation
User-agent: GPTBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: Applebot-Extended
Allow: /
Allowing AI citation while blocking training is enforceable for OpenAI, Anthropic, and Perplexity: allow the search crawler (OAI-SearchBot, Claude-SearchBot, PerplexityBot) and disallow the training crawler (GPTBot, ClaudeBot). Blocking one does not block the other.
Google works differently: Google-Extended limits training and Gemini grounding but not appearance in AI Overviews or AI Mode. To keep regular Search visibility but leave AI Overviews / AI Mode, use the Search Console Search generative AI control (as of September 2026); snippet controls and noindex also work but cost regular Search presentation.
Recommendation: Always allow the search/retrieval crawlers if you want AI citation visibility; decide on training crawlers as a separate content-licensing policy.
Test top 10 target queries on Perplexity and ChatGPT:
There is no "AI Overviews" search-type filter in the Performance report. Clicks and impressions from AI Overviews and AI Mode are counted inside the standard Web search type, blended with regular results (AI features and your website).
As of September 2026, Search Console also has a dedicated Generative AI performance report (announced June 2026, rolled out to all sites by August 31, 2026; sites need enough impressions to see data) (announcement, help):
For clicks, use the Web search type in the Performance report plus landing-page analytics; AI-feature clicks cannot be isolated there.
| Signal | What to Check | Tool |
|---|---|---|
| Perplexity citations | Top 10 queries | Manual testing |
| ChatGPT citations | Top 10 queries | Manual testing |
| Google AI Overviews / AI Mode | Impressions (Generative AI report); clicks only blended into Web search type | Google Search Console |
| Copilot citations | Top 5 queries | Manual testing |
| AI bot crawl activity | Crawl frequency and pages | Server logs / Cloudflare |
| Competitor citations | Who is getting cited for your queries | Manual testing |
| Content freshness | Date signals on key pages | Content audit |
Diagnostic checklist when you lose a citation:
Mistake: The page answers the question in paragraph six, after background and brand story.
Why it happens: It was written the way long-form SEO content used to be written, for a reader who scrolls.
Instead: AI systems extract sections, not pages. Put the answer in the first 200 words and make each H2 block citable on its own.
Mistake: OAI-SearchBot or PerplexityBot is disallowed in robots.txt.
Why it happens: A CMS update or security plugin rewrote robots.txt, and nobody checks it after updates.
Instead: Audit robots.txt after every CMS or plugin change, and keep key answers in the initial HTML, because AI crawlers may not run JavaScript.
Mistake: FAQ or HowTo markup goes on every page in the hope of AI citations.
Why it happens: Rich-result tricks once worked for classic search.
Instead: Mark up what the page visibly contains (Article, Product, Organization, BreadcrumbList). Google says AI features need no special schema, and FAQ and HowTo rich results no longer display.
| Problem | Likely Cause | Fix |
|---|---|---|
| Content not cited despite high DA | Poor extractability — answers buried in prose | Restructure with definition blocks, numbered steps, and FAQ pairs in first 200 words |
| Cited on Perplexity but not ChatGPT | Different crawling and indexing pipelines per platform | Verify bot access for all AI crawlers; test rendering without JavaScript |
| AI Overview shows competitor instead | Competitor has more extractable, better-attributed content | Audit competitor's cited content format and match or exceed specificity |
| Citation dropped after site update | Page restructure broke the extraction pattern AI was using | Compare old vs new page structure; restore extractable blocks |
| OAI-SearchBot / PerplexityBot blocked in robots.txt unknowingly | CMS update or security plugin overwrote robots.txt | Audit robots.txt after every CMS or plugin update; set up monitoring |
| Schema markup present but no rich results | Missing required fields or content-markup mismatch | Validate with Google Rich Results Test; ensure schema matches visible page content |
| AI cites your data but not your brand | Missing entity signals — no Organization schema or sameAs links | Implement Organization schema with sameAs to Wikidata, LinkedIn, and social profiles |
In scope:
Out of scope:
Known limitations:
ChatGPT-User, Perplexity-User) may not honor robots.txt# Analyze content for AI citability signals
python scripts/content_scorer.py page.html --json
# Simulate how content might appear in AI search results
python scripts/serp_simulator.py --query "what is cloud cost optimization" --content page.md
# Analyze keyword opportunities for AI search visibility
python scripts/keyword_analyzer.py --keywords keywords.csv --json