You want AI systems to cite your pages, not your competitors. This guide gives you a clear definition, a practitioner checklist and reproducible assets you can ship today. It serves B2B SaaS, e-commerce and multi-location teams that need vendor-neutral steps, copyable schema and measurement tied to revenue. The Floyi GEO Blueprint anchors the method and artifacts.
Use this page to:
- Learn what GEO is and how it differs from SEO
- Ship quick wins with schema and provenance that AI can verify
- Measure AI-citation lift and business impact with a reproducible method
What is Generative Engine Optimization (GEO)?
Generative Engine Optimization (GEO) optimizes content, structured data and provenance so AI systems return your brand as a trusted answer.
- When GEO matters: when ChatGPT, Perplexity, Gemini, and Google AI Overviews build concise answers and choose sources for AI-citation
- How GEO differs from SEO: SEO tunes ranking signals for SERPs. GEO packages verifiable claims, schema and provenance so generative systems select and cite your page
- One action to start today: publish a provenance-backed JSON-LD Claim on one high-value page with author, date and an evidence URL visible near the claim
GEO is the practice of packaging verifiable claims with schema and visible provenance so generative engines can select, cite and summarize your page. It emphasizes answer-first copy, JSON-LD that maps claims to evidence and stable IDs that let systems verify authorship, date and sources.
Some teams refer to GEO and adjacent concepts like “answer engine optimization (AEO)” or “large language model optimization (LLMO),” but this guide focuses on schema, provenance and selection for AI-citation.
GEO works best on sites with strong topical authority and a clear topical map, because engines can verify claims across a coherent cluster and cite the most canonical node.
Why does GEO matter now?
AI answers are concise and cite a few sources. Engines favor reproducible, provenance-backed claims that are easy to extract. Package your claims so they can be selected, cited and summarized.
Design for current engine behavior:
- Extract short answers from clearly labeled sections
- Prefer reproducible claims with visible provenance
- Consolidate signals so engines can choose a small set of sources
Engines prefer clusters with complete coverage. Publish the lead claim on the hub page, then reinforce it with supporting subtopics. Use internal links that mirror your topical map.
How is GEO different from SEO?
- SEO focuses on ranking signals and click-through from traditional SERP listings
- GEO focuses on selection and AI-citation in ChatGPT, Gemini, Perplexity, and Google AI Overviews and AI Mode.
- GEO uses answer-first packaging, JSON-LD schema and machine-readable provenance
SEO vs. GEO at a glance
| Dimension | SEO | GEO |
| Primary goal | Rank in SERPs | Be selected and cited in AI answers |
| Packaging | Long-form page with headings | 40–80 word lead claim with provenance |
| Markup | Basic schema | Claim, FAQ and HowTo schema with explicit claim-to-evidence mapping |
| KPI | CTR and organic sessions | AI-citation share, answer presence, assisted conversions |
How does GEO affect conversion?
Being the cited source reduces leakage in zero-click experiences. Citation with brand, author, date and evidence URL raises trust and clicks.
B2B SaaS benefits from clearer problem and solution alignment. E-commerce benefits from clearer specs and compatibility. Multi-location benefits from stable hours and services.
What should you implement this week for GEO?
Ship GEO on five high-value pages using three moves: audit targets, write a 40–80 word lead answer, and add schema with provenance for AI-citation.
1. Audit pages
Pick five pages that match high-value intents:
- FAQs for product, policy or troubleshooting
- How-to guides for setup or usage
- Product or service specs with compatibility details
- vs. and comparison pages
- Location pages with hours, services and NAP
Selection rules:
- Start with pages that already convert
- Add queries from Search Console with who, what, how, vs.
- Include one page per intent type to compare impact
- Prioritize hubs with incomplete subtopic coverage in your topical map
- Choose one leaf per cluster to compare impact across depths
- Add a descriptive internal link from each leaf to the hub and between siblings
2. Write the 40–80 word lead answer
Place a short, direct answer near the top. Add a one-line provenance anchor under it.
- Lead answer template: “State the canonical claim in one short paragraph that matches user intent and cites on-page evidence.”
- Visible provenance line: “Source: [Author or Team], [YYYY-MM-DD]. Evidence: [URL or #section-id].”
Learn more about Search Intent vs User Intent Differences and the new Generative Intent.
3. Add schema and provenance that engines can verify
Paste one JSON-LD block. Replace bracketed fields. Keep IDs stable so claims map to evidence.
FAQPage (copyable)
{
"@context": "https://schema.org",
"@type": "FAQPage",
"@id": "[https://www.example.com/page#faq]",
"mainEntity": [
{
"@type": "Question",
"@id": "[https://www.example.com/page#q1]",
"name": "[Question in user language]",
"acceptedAnswer": {
"@type": "Answer",
"text": "[40–80 word answer that matches the on-page lead]",
"citation": { "@type": "CreativeWork", "url": "[#evidence-anchor-or-url]" },
"author": { "@type": "Organization", "name": "[Brand]" },
"datePublished": "[YYYY-MM-DD]"
}
}
],
"mainEntityOfPage": { "@id": "[https://www.example.com/page]" }
}
HowTo (copyable)
{
"@context": "https://schema.org",
"@type": "HowTo",
"@id": "[https://www.example.com/page#howto]",
"name": "[Task users search for]",
"description": "[Short task summary aligned to the lead answer]",
"totalTime": "PT[minutes]M",
"step": [
{ "@type": "HowToStep", "name": "Step 1", "text": "[Do X]", "url": "[#step-1]" },
{ "@type": "HowToStep", "name": "Step 2", "text": "[Do Y]", "url": "[#step-2]" }
],
"author": { "@type": "Organization", "name": "[Brand]" },
"mainEntityOfPage": { "@id": "[https://www.example.com/page]" },
"citation": { "@type": "CreativeWork", "url": "[#evidence-anchor-or-url]" },
"datePublished": "[YYYY-MM-DD]"
}
Claim (copyable)
{
"@context": "https://schema.org",
"@type": "Claim",
"@id": "[https://www.example.com/page#claim-001]",
"claimText": "[Your canonical claim in one sentence]",
"author": { "@type": "Organization", "name": "[Brand]", "url": "https://www.example.com" },
"datePublished": "[YYYY-MM-DD]",
"identifier": { "@type": "PropertyValue", "name": "ClaimID", "value": "claim-001" },
"citation": { "@type": "CreativeWork", "url": "[#evidence-anchor-or-url]" },
"isBasedOn": { "@type": "CreativeWork", "url": "[#supporting-dataset-or-method]" },
"mainEntityOfPage": { "@id": "[https://www.example.com/page]" },
"about": [
{ "@type": "Thing", "name": "Generative Engine Optimization" },
{ "@type": "Thing", "name": "GEO" },
{ "@type": "DefinedTerm", "name": "Topical authority", "inDefinedTermSet": "https://floyi.com/topical-map#authority" }
],
"isPartOf": { "@type": "CreativeWorkSeries", "name": "GEO Cluster", "url": "https://floyi.com/topical-map#geo-cluster" }
]
}
Implementation notes:
- Place JSON-LD once per page in the head or just before </body>
- Make the evidence anchor real. Match the on-page claim, author, date and supporting material
- Keep Claim IDs permanent
4. Publish and request indexing
- Add one internal link from a thematically close page with a descriptive anchor
- Update the XML sitemap, lastmod and canonical
- Request indexing after you publish the lead answer and schema
5. Track AI-citation and outcomes
- Capture a pre-change baseline: AI answer presence, citations and share of voice
- Check weekly for 4 weeks. Record AI-citation count, answer presence, CTR and assisted conversions
- Escalate pages that move toward consistent AI-citation. Iterate the claim, schema and evidence on laggards
You should see faster selection and AI-citation on pages with a clear claim, schema and provenance, cleaner answers in AI search engines, and better conversion efficiency on product, how-to and comparison pages.
How do you run a reproducible GEO experiment?
Run a matched-pair, 12-week test on ten pages. Apply GEO artifacts to five Treatment pages (page in your experiment that receives the GEO package you are testing), keep five as Controls and track AI-citation share and outcomes with an open, pre-registered protocol.
How should you design the study?
Use a simple, controlled design that isolates the GEO package.
- Define scope: one site section, language and country
- Select pages: ten that already rank or convert
- Match pairs: pair by intent and traffic
- Randomize: assign one per pair to Treatment, one to Control
- Intervene: add lead answer, schema and provenance on Treatment only
- Freeze edits: no non-intervention changes during the window
- Observe: collect weekly telemetry for 12 weeks
- Analyze: compare change in AI-citation share and conversions
Which pages qualify for inclusion?
Choose one page per intent type with similar crawlability and internal depth.
- FAQ
- How-to
- Product or service spec
- vs. or comparison
- Location
- Each page belongs to a single cluster in your topical map
- Hubs and leaves have consistent anchors and reciprocal internal links
What goes into the Treatment?
Ship the same GEO package on every Treatment page.
- 40–80 word lead answer near the top
- Visible provenance line with author, date and evidence anchor
- JSON-LD Claim plus FAQPage or HowTo that maps claim to evidence
- One descriptive internal link from a related page
How do you measure outcomes?
Track machine-readable metrics weekly and compare against Control.
| Metric | Definition | Source | Target signal |
| AI-citation share | Share of answers that cite your domain for the tracked query set | Weekly checks in ChatGPT, Perplexity, Gemini, and Google AI Overviews | +10% vs. Control for 3 consecutive weeks |
| Answer presence | Binary: your domain appears in answer citations | Same as above | Presence in ≥50% of weeks |
| Share of voice | Your citations divided by all cited domains for the same answers | Same as above | +5 percentage points vs. baseline by week 8 |
| CTR | Click-through from organic and AI surfaces where available | Analytics, GSC, platform logs | Lift vs. baseline |
| Assisted conversions | Conversions with assisted sessions touching test pages | Analytics with model-friendly attribution | Lift vs. Control |
What telemetry format should you publish?
Use tidy, stable IDs and share both CSV and JSON.
CSV header
week_start,engine,query,page_url,variant,answer_present,our_domain_cited,cited_domains,ai_citation_share,sov,notesRow example
2025-09-01,Perplexity,"what is generative engine optimization geo",https://example.com/geo,Treatment,1,1,"example.com;floyi.com;vendor.com",0.50,0.33,"Lead answer revised on 2025-09-08"JSON record
{
"week_start": "2025-09-01",
"engine": "Perplexity",
"query": "what is generative engine optimization geo",
"page_url": "https://example.com/geo",
"variant": "Treatment",
"answer_present": true,
"our_domain_cited": true,
"cited_domains": ["example.com", "floyi.com", "vendor.com"],
"ai_citation_share": 0.5,
"sov": 0.33,
"notes": "Lead answer revised on 2025-09-08"
}
How should you analyze results?
Compute weekly metrics, then apply difference-in-differences across weeks 1–12.
lift = (Treatment_week12 − Treatment_week1) − (Control_week12 − Control_week1)Flag success when Treatment outperforms Control by 10% or more AI-citation share for three consecutive weeks, then confirm movement in CTR and assisted conversions.
How do you publish the open dataset and methods?
Host everything under one canonical URL with a clear license.
/dataset/telemetry.csv
/dataset/pages.csv
/dataset/queries.csv
/notebooks/analysis.ipynb
/methods/methods.md
/methods/preregistration.md
/LICENSE
/README.mdPreregistration (copyable)
Hypothesis: Treatment pages with lead answers, schema and provenance increase AI-citation share by ≥10% vs. Control over 12 weeks.
Units: Page-week.
Primary metric: AI-citation share.
Secondary metrics: Answer presence, share of voice, CTR, assisted conversions.
Inclusion: Ten pages matched by intent and traffic; five Treatment, five Control.
Intervention: Lead answer, visible provenance, JSON-LD Claim plus FAQPage or HowTo.
Analysis: Difference-in-differences; success if lift ≥10% for three consecutive weeks.
Protocol deviations: Logged with timestamps in methods.md.How do you ensure validity and compliance?
- Note model updates, seasonality and major site changes
- Follow each engine’s terms for data collection
- Record all deviations with timestamps
What is the replication checklist?
- Pages and pairs logged in pages.csv
- Queries fixed and published in queries.csv
- Telemetry captured weekly with stable IDs
- Notebook runs end-to-end on dataset files
- Results table and one-paragraph narrative added to this article
What developer artifacts can you copy now?
Here are ready-to-use schema blocks, a provenance file pattern and tested RAG prompts so you can ship GEO quickly and earn AI-citation.
Which JSON-LD snippets should you use?
Use one content pattern per page. Keep IDs stable.
WebPage + Claim + Dataset (provenance-aware)
{
"@context": "https://schema.org",
"@type": "WebPage",
"@id": "https://www.example.com/geo#webpage",
"url": "https://www.example.com/geo",
"name": "Generative Engine Optimization (GEO): Definition and Playbook",
"datePublished": "2025-08-31",
"author": { "@type": "Organization", "name": "Your Brand" },
"mainEntity": {
"@type": "Claim",
"@id": "https://www.example.com/geo#claim-001",
"name": "GEO optimizes content, structured data and provenance so AI systems return your brand as a trusted answer.",
"description": "Short, canonical claim used in the lead answer.",
"author": { "@type": "Organization", "name": "Your Brand" },
"datePublished": "2025-08-31",
"citation": { "@type": "CreativeWork", "url": "https://www.example.com/geo#evidence" },
"isBasedOn": { "@id": "https://www.example.com/geo#dataset" }
},
"citation": { "@id": "https://www.example.com/geo#dataset" }
}Dataset (telemetry you publish with the experiment)
{
"@context": "https://schema.org",
"@type": "Dataset",
"@id": "https://www.example.com/geo#dataset",
"name": "GEO AI-citation Telemetry",
"description": "Weekly AI-citation, answer presence and share of voice for matched-page experiment.",
"creator": { "@type": "Organization", "name": "Your Brand" },
"datePublished": "2025-08-31",
"license": "https://creativecommons.org/licenses/by/4.0/",
"distribution": [
{
"@type": "DataDownload",
"encodingFormat": "text/csv",
"contentUrl": "https://www.example.com/dataset/telemetry.csv"
}
]
}FAQPage (drop-in)
{
"@context": "https://schema.org",
"@type": "FAQPage",
"@id": "https://www.example.com/geo#faq",
"mainEntity": [
{
"@type": "Question",
"@id": "https://www.example.com/geo#q1",
"name": "What is Generative Engine Optimization (GEO)?",
"acceptedAnswer": {
"@type": "Answer",
"text": "GEO optimizes content, schema and provenance so AI systems select and cite your page.",
"author": { "@type": "Organization", "name": "Your Brand" },
"datePublished": "2025-08-31",
"citation": { "@type": "CreativeWork", "url": "https://www.example.com/geo#evidence" }
}
}
],
"mainEntityOfPage": { "@id": "https://www.example.com/geo" }
}HowTo (drop-in)
{
"@context": "https://schema.org",
"@type": "HowTo",
"@id": "https://www.example.com/geo#howto",
"name": "Implement GEO on five pages in one week",
"description": "Audit targets, write a 40–80 word lead answer, add schema with provenance.",
"totalTime": "PT180M",
"step": [
{ "@type": "HowToStep", "name": "Audit pages", "text": "Pick five high-value pages.", "url": "https://www.example.com/geo#audit" },
{ "@type": "HowToStep", "name": "Write the lead answer", "text": "Place a 40–80 word claim near the top.", "url": "https://www.example.com/geo#lead" },
{ "@type": "HowToStep", "name": "Add schema and provenance", "text": "Paste JSON-LD and anchor evidence.", "url": "https://www.example.com/geo#schema" }
],
"author": { "@type": "Organization", "name": "Your Brand" },
"mainEntityOfPage": { "@id": "https://www.example.com/geo" },
"citation": { "@type": "CreativeWork", "url": "https://www.example.com/geo#evidence" },
"datePublished": "2025-08-31"
}How should you publish a provenance file?
Expose a small, machine-readable index that ties claims to on-page anchors and datasets. Link it from the head with a rel=alternate tag.
/provenance.json (pattern)
{
"publisher": "Your Brand",
"updated": "2025-08-31",
"claims": [
{
"claim_id": "claim-001",
"page_url": "https://www.example.com/geo",
"anchor": "#evidence",
"claim": "GEO optimizes content, structured data and provenance so AI systems return your brand as a trusted answer.",
"author": "Your Brand",
"date": "2025-08-31",
"dataset_url": "https://www.example.com/dataset/telemetry.csv",
"checksum_sha256": "abc123..."
}
]
}Head tag
<link rel="alternate" type="application/json" href="/provenance.json">How do you set up RAG so answers cite your page?
Keep retrieval narrow. Prefer claims and evidence anchors.
Retrieval steps
- Normalize the query to intent
- Search your vector index with a site filter to your domain
- Boost passages that contain a claim_id or evidence anchor
- Return top k passages with their URLs and fragment IDs
System prompt
You answer with facts from the provided context only. Cite sources with [site-url#anchor]. Keep answers to 80 words unless asked for more. If the context is insufficient, say what is missing.User prompt template
Question: {{user_query}}
Context:
{{top_k_passages_with_urls_and_anchors}}
Task: Give a short, direct answer. Include 1–3 citations that point to the most relevant anchors.Expected answer shape (for testing)
{
"answer_text": "Generative Engine Optimization (GEO) packages content, schema and provenance so AI systems can select and cite your page.",
"citations": [
{ "url": "https://www.example.com/geo#evidence", "claim_id": "claim-001" }
]
}
Where do you place schema and provenance in common CMSs?
- WordPress: add a script tag with type application/ld+json in the theme head or via a code injection plugin. Keep one JSON-LD block per type
- Shopify: insert JSON-LD in theme.liquid head or in template sections for product and article types
- Webflow: paste JSON-LD in the Page Settings custom code area before </body>
- Wix: use the SEO panel’s Structured Data or Tracking & Analytics to inject head code
- Headless (Next.js, Gatsby): render JSON-LD in <Head> per route and serve /provenance.json at the site root
How do you validate before you ship?
- Validate JSON-LD with a structured data tester
- Click every #anchor you reference
- Confirm IDs stay stable across deploys
- Check robots, sitemaps and canonical URLs
- Request indexing after the visible lead answer and schema go live
How should you measure and monitor GEO?
Track AI-citation lift and business impact with a small metric set, a simple dashboard and clear alert rules. Use weekly telemetry and a monthly outcome review with defined stop conditions.
What metrics should you track?
Track AI-citation share, answer presence, share of voice, CTR and assisted conversions. Keep field names consistent with your telemetry.
| Metric | Definition | Calculation | Source | Threshold | Cadence |
| AI-citation share | Share of answers that cite your domain for the tracked query set | Mean of ai_citation_share per page or pair | Weekly engine checks | Treatment ≥ Control + 10% for 3 consecutive weeks | Weekly |
| Answer presence | Binary: your domain appears in the answer citations | Mean of answer_present (0 or 1) | Weekly engine checks | Presence in ≥50% of weeks | Weekly |
| Share of voice | Your citations divided by all cited domains for the same answers | Mean of sov | Weekly engine checks | +5 percentage points vs. baseline by week 8 | Weekly |
| CTR | Click-through from organic and AI surfaces | Sessions with clicks ÷ impressions | Analytics, GSC | +5% vs. baseline by week 8 | Weekly |
| Assisted conversions | Conversions with assisted sessions touching test pages | Model-friendly attribution | Analytics | +10% vs. Control by week 12 | Monthly |
| Topic coverage | Percent of required subtopics published in the cluster | published_subtopics ÷ planned_subtopics | Site map or CMS | ≥80% by week 8 | Monthly |
| Hub link ratio | Ratio of links to higher-level topic (e.g., main topic) | Internal links from leaves to hub ÷ total internal links on leaves | CMS export | ≥0.30 | Monthly |
How do you wire a Google Sheets starter?
Use three tabs and keep IDs stable.
- telemetry_raw: paste weekly rows with headers week_start,engine,query,page_url,variant,answer_present,our_domain_cited,cited_domains,ai_citation_share,sov,notes
- metrics_weekly: group by week_start,variant, compute means
AI-citation share
=AVERAGEIFS(telemetry_raw!I:I, telemetry_raw!A:A, A2, telemetry_raw!E:E, B2)Answer presence
=AVERAGEIFS(telemetry_raw!F:F, telemetry_raw!A:A, A2, telemetry_raw!E:E, B2)Share of voice
=AVERAGEIFS(telemetry_raw!J:J, telemetry_raw!A:A, A2, telemetry_raw!E:E, B2)Diff-in-diff in summary
=(AVERAGEIFS(metrics_weekly!C:C, metrics_weekly!A:A, MAX(metrics_weekly!A:A), metrics_weekly!B:B, "Treatment")
-AVERAGEIFS(metrics_weekly!C:C, metrics_weekly!A:A, MIN(metrics_weekly!A:A), metrics_weekly!B:B, "Treatment"))
-(AVERAGEIFS(metrics_weekly!C:C, metrics_weekly!A:A, MAX(metrics_weekly!A:A), metrics_weekly!B:B, "Control")
-AVERAGEIFS(metrics_weekly!C:C, metrics_weekly!A:A, MIN(metrics_weekly!A:A), metrics_weekly!B:B, "Control"))How do you build a Looker Studio starter?
Connect the Sheets file, then add calculated fields and filters.
- Calculated fields: AI Citation Share = AVERAGE(ai_citation_share), Answer Presence = AVERAGE(answer_present), Share of Voice = AVERAGE(sov)
- Charts: time series by week split by variant, table by engine and query, bar chart by engine for share of voice, scorecards for CTR and assisted conversions
- Filters: Variant (Treatment, Control), Engine (ChatGPT, Perplexity, Gemini, and Google AI Overviews), Intent type (FAQ, How-to, Spec, vs., Location)
What alert rules should you set?
Use simple, testable triggers.
- Success: Treatment AI-citation share ≥ Control + 10% for 3 consecutive weeks
- Regression: week-over-week drop ≥20% in Treatment AI-citation share
- Quality: Answer presence < 25% for 2 consecutive weeks
- Outcome: CTR or assisted conversions down ≥10% vs. baseline after 4 weeks
What cadence should you follow?
Keep a steady loop that favors small, fast iterations.
- Weekly: scrape AI answers, update telemetry, review dashboard, log notes
- Biweekly: adjust underperforming pages, refine lead answers and anchors
- Monthly: review CTR and assisted conversions, decide to end or extend the test
- Stop rules: stop when the success trigger holds for 3 consecutive weeks or after 12 weeks without lift
How do you keep provenance and IDs stable?
Stability helps engines trust your claims.
- Keep claim_id, page_url and anchor unchanged across deploys
- Log every change in methods.md with timestamps
- Version the dataset with a date suffix and update the Dataset JSON-LD distribution URL
What should you do next with GEO?
Publish one flagship resource with a clear definition and copyable schema. Pair it with a reproducible dataset, then run the 12-week test and measure AI-citation lift.
- GEO in one line: package content, schema and provenance so AI systems select and cite your pages. SEO ranks. GEO gets selected.
- Quick wins: audit five high-value pages, write 40–80 word lead answers, add FAQPage, HowTo and Claim schema, anchor visible provenance
- Experiment: match pairs, randomize Treatment vs. Control, freeze edits, track AI-citation share, answer presence, share of voice, CTR and assisted conversions
- Developer artifacts: use WebPage, Claim and Dataset JSON-LD, publish /provenance.json, keep IDs stable, test RAG prompts that return citations to your anchors
- Measurement: wire Sheets and Looker Studio, set alert rules, stop when Treatment beats Control by 10% for three weeks or after 12 weeks without lift
Make this URL your canonical hub for GEO, then fill the surrounding leaves to raise topical authority and improve selection. Host the dataset, methods and notebooks here. Link to it from related pages with descriptive anchors. Pitch the study to tradepress and a neutral co-author. Your goal is simple: be the source AI search engines cite.
Generative Engine Optimization (GEO) FAQs
Is GEO just schema?
No. Schema helps, but GEO requires answer-first copy, visible provenance and stable IDs that map claims to evidence.
Does GEO replace SEO?
No. SEO drives rankings and clicks in SERPs. GEO helps your page get selected and cited inside generative answers.
How long until I see AI-citation lift?
Many tests see movement within 3–6 weeks. Use the 12-week protocol to confirm.
How do topical maps improve GEO selection?
Topical maps turn isolated pages into a coherent cluster. Engines can trace the hub claim, check supporting leaves and assign more trust. Build the hub first, add 3–5 leaves per cluster, and link hub ↔ leaves with clear anchor text.
