AI Summary
This checklist is the same one we use to baseline new client engagements at OrganikPI: 50 specific signals across six sections, scored 1 or 0, producing a number that tells you exactly where to start fixing.
AI-referred website sessions grew 527% between January and May 2025 across 19 tracked properties, according to Previsible’s AI Traffic Report. The businesses showing up in those sessions did not get there by publishing more content. They got there by passing the structural checks below. Score each item 1 if implemented, 0 if not. The interpretation guide at the end tells you what your score means and what to fix first.
Section 1: Technical foundations (10 points)
Why it matters: AI crawlers follow robots.txt and serve clean HTML for citation extraction. A single misconfigured directive can make your entire site invisible to citation bots, regardless of content quality. How to check: Fetch your robots.txt in a browser and search for GPTBot, Google-Extended, PerplexityBot. Run a log file analysis over 30 days to confirm actual bot hits, not just permitted access.
- robots.txt allows GPTBot, Google-Extended, PerplexityBot
- Sitemap.xml is clean, only 200-status URLs, submitted to GSC
- Site uses semantic HTML5 (article, section, nav, header, footer)
- Largest Contentful Paint under 2.5 seconds
- Mobile-responsive design verified across viewports
- HTTPS enforced site-wide
- No “click to expand” patterns hiding content from crawlers
- Clean URL structure (no session IDs, query strings)
- Structured data via JSON-LD on every meaningful page
- llms.txt file present at root
Section 2: Entity and authority signals (10 points)
Why it matters: AI engines do not cite sources they cannot identify. Entity signals tell the model who you are, what you do, and whether you are a credible source. A 2024 study by Sequeda et al. (arXiv 2405.11706) showed LLM accuracy on fact extraction improved from 16% to 54% when content included structured data. How to check: Validate your Organization schema at schema.org/Organization and confirm sameAs links resolve to live profile pages.
- Organization schema implemented site-wide
- Author schema on every editorial page
- sameAs properties linking to verified profiles (LinkedIn, GitHub, Twitter)
- About Us page with substantive company detail
- Contact page with physical address
- Author bio pages for every named author
- External citations to credible primary sources
- Trust signals (case studies, customer logos, certifications)
- Leadership team profiles published
- Wikidata entity with bidirectional sameAs
Section 3: Content quality signals (10 points)
Why it matters: AI engines extract sentences, not pages. Our May 2026 study of 153,425 citations found cited sentences average 9.27 words, with the 6-10 word range accounting for 45.2% of all citations. Sentences over 18 words were never cited. Content must be structured for extraction, not just for human reading. How to check: Take your top five pages and measure average sentence length in key claim sections. Count atomic facts vs. discursive prose. Target at least 60% atomic sentences in any section making factual claims.
- Inverted-pyramid writing, answer first, context after
- Original statistics, data, or research embedded
- Expert quotes or sources cited
- Clear definitions of technical terms
- FAQ schema on pages with common questions
- Tables or lists used for structured data
- Descriptive image alt text on all images
- Content updated within last 6 months
- Citations to primary sources, not aggregator pages
- No obvious AI-generated fluff (human-verified)
Section 4: Semantic structure (10 points)
Why it matters: AI parsers use your HTML heading hierarchy as a navigation map. Skipped heading levels, layout tables, and JavaScript-rendered content all introduce extraction errors. 74.9% of cited sentences in our May 2026 study came from the first half of the document: front-loaded structure compounds with good heading hierarchy to dramatically increase citation probability. How to check: Run your page through a DOM inspector and verify H1-H6 hierarchy with no skipped levels. Confirm lists use ul/ol tags and tables use thead/tbody. Check that critical content exists in the initial HTML response, not injected via JavaScript.
- H1-H6 hierarchy is logical and consistent
- Headings contain natural keyword variants
- Lists use proper ul/ol tags, not styled paragraphs
- Tables use table tags for actual data, not layout
- Blockquotes used for actual quotes
- Strong/em used semantically, not decoratively
- No intrusive ads breaking content flow
- Breadcrumb schema implemented on every page
- Related content linking present
- Internal linking minimum 3 contextual links per page
Section 5: Citation-worthy assets (5 points)
Why it matters: 76.95% of URLs cited by AI engines in our May 2026 study were not in the organic top-10 search results. Traditional ranking does not predict citation. What predicts citation is content that AI engines cannot source anywhere else: original research, proprietary data, and expert synthesis. The arXiv GEO paper (2311.09735) confirmed that cite-sources and statistics methods produce 30-40% more AI visibility. How to check: List all content assets published in the last 12 months and tag each as original research, comparison, how-to, or case study. Any asset lacking a unique data point or proprietary methodology is a weak citation candidate.
- Original research or proprietary data published in last 12 months
- Expert thought leadership with credentialed authors
- Comparison or “vs.” pages for direct competitors
- How-to guides with specific, actionable steps
- Case studies with quantitative results
Section 6: Measurement and optimization (5 points)
Why it matters: GEO is not a one-time project. AI training cycles, retrieval-index refreshes, and competitor changes move citation positions on a monthly cadence. Without measurement, you cannot tell whether your changes are working or whether a competitor has displaced you. How to check: Open GA4 and filter sessions by source containing chat.openai.com, perplexity.ai, and gemini.google.com. If you see zero or “not set,” your AI referral attribution is broken. Set up a monthly prompt set of 20-30 category queries and run them manually or via a tracking tool to measure share of voice.
- GA4 configured to track AI referral traffic (chat.openai.com, perplexity.ai)
- Citation tracking tool or process in place
- Regular content audits scheduled (quarterly minimum)
- Competitor citation benchmarking running monthly
- Feedback loop from citation data to content roadmap
Score interpretation
40-50 points: GEO-ready
Strong foundations. Focus on asset velocity and competitive expansion. You are well-positioned for compounding citation growth. The next lever is publishing more citation-worthy assets at a higher cadence and tracking competitor share of voice monthly to identify gaps before they cost you.
25-39 points: Moderate readiness
Foundations are partial. Prioritize missing technical and entity signals first. They are force-multipliers for everything else. A site with excellent content but broken robots.txt permissions earns zero citations. Fix access first, then schema, then content structure.
10-24 points: Significant gaps
Multiple foundation gaps. Do not publish more content yet. Spend 30-60 days on technical and entity foundations before adding asset velocity. Publishing into a broken infrastructure compounds the problem by diluting crawl budget across pages that cannot be cited.
Under 10 points: Start from zero
Treat this as a greenfield build. The good news: you can implement most foundation items in 4-6 weeks of focused work. Most sites scoring under 10 are missing robots.txt permissions entirely and have no schema markup, which means a single technical sprint closes the gap dramatically.
What to fix first
The sequence matters. An AI engine that cannot access your site cannot cite it, regardless of how good the content is. Fix in this order:
- robots.txt and llms.txt, enable AI engines to access and understand your site at all
- Organization and Author schema, the entity recognition foundation every other signal depends on
- Top 5 highest-traffic pages: structure, FAQ, semantic markup to maximize extraction accuracy
- About and author bio pages with credentials and sameAs links
- Citation tracking in GA4, you cannot improve what you do not measure
The highest-leverage technical checks in more depth
robots.txt permissions for AI bots
AI crawlers operate in three tiers. Training bots (GPTBot, Google-Extended, anthropic-ai) consume content to improve model weights. Citation bots (OAI-SearchBot, PerplexityBot) fetch pages to generate real-time answers. User-triggered bots (ChatGPT-User, Claude-User) retrieve pages when a user follows a citation. Most businesses benefit from allowing all three tiers. Blocking GPTBot blocks training access but does not block real-time citation. Blocking PerplexityBot blocks both. For most B2B SaaS companies, the right policy is allow all and invest in content quality instead. Read the full AI crawler control guide to understand the tradeoffs.
Schema markup and extraction accuracy
Structured data is not optional for GEO. A 2024 study (arXiv 2405.11706, Sequeda et al.) found that providing content via a knowledge graph raised LLM accuracy from 16% to 54% on fact-extraction tasks. That 38-percentage-point gain comes from the same principle: structured signals help the model identify what a fact is, who asserts it, and whether it is trustworthy. For GEO purposes, the highest-leverage schema types are Organization (entity disambiguation), Article (content attribution and freshness), FAQPage (direct answer extraction), and Person (author authority). See the schema decision guide for implementation priorities.
Content freshness signals
AI engines weight recency. The article:modified_time Open Graph tag and the dateModified field in Article schema both signal freshness to crawlers. Pages updated within 60 days receive measurably higher citation rates than stale content, because many AI retrieval systems de-prioritize older content for rapidly evolving categories like AI search itself. Schedule quarterly content audits and update statistics, examples, and links on your highest-traffic pages to maintain freshness signals. The content freshness guide covers the mechanics.
Run it yourself or hand it off
This checklist pairs with the DIY visibility audit (three free methods for the prompt-side baseline) and the AI search readiness audit framework for a full framework approach. The open-source GEO/AEO Tracker handles ongoing measurement once foundations are in place. If you want the full-depth version with competitive source mapping and a prioritized roadmap, that is the GEO audit service: $4,500 flat, two weeks, fee credited if you continue into a retainer.
Background reading: what GEO is, the 42,971-citation study this checklist is built on, and the GEO vs SEO breakdown for context on where this fits in your overall strategy. For the agency-vs-DIY question on who executes the fixes, see the GEO agency vs DIY guide.
Once your audit surfaces gaps, the right tracker helps you watch them close. Compare the top GEO tools for 2026.