How to use this checklist
Work top to bottom and mark each item pass or fail — no partial credit, because every item is binary-verifiable with evidence you can screenshot. The four groups are ordered by dependency: technical access gates everything, content makes access worthwhile, sources make content believable, and measurement tells you whether any of it worked. Method details for each group live in the GEO Wiki; the scoring philosophy follows the open measurement standard.
Group 1: Technical access (14)
| # | Check |
|---|---|
| 1 | Google Search Console verified; sitemap submitted; no mass exclusions of canonical pages |
| 2 | Bing Webmaster Tools verified; brand and category pages confirmed indexed |
| 3 | IndexNow connected; page updates push automatically |
| 4 | robots.txt allows OAI-SearchBot and ChatGPT-User |
| 5 | robots.txt allows PerplexityBot |
| 6 | GPTBot, ClaudeBot and other training bots configured deliberately — not by accident in either direction |
| 7 | Google-Extended decision documented (remember: training only, no effect on AI Overviews) |
| 8 | Key content server-rendered; no JavaScript dependency for answer-bearing passages |
| 9 | Organization, Article and FAQPage schema deployed and matching visible content |
| 10 | Product/Offer/Review schema on commerce pages; prices as text, not images |
| 11 | llms.txt published and synchronized with the real page set |
| 12 | Server logs segmented by AI user agents; fetch frequency reviewed monthly |
| 13 | Core pages interactive within 3 seconds on mobile |
| 14 | No staging, faceted or parameter URLs wasting AI crawl budget |
Group 2: Content engineering (14)
| # | Check |
|---|---|
| 15 | Every core page opens with a bolded 40-80 word answer block |
| 16 | H2/H3 headings are real buyer questions |
| 17 | Key categories have comparison tables with an honest “who this suits” column |
| 18 | Every core page carries 2-4 genuine FAQs with matching FAQPage schema |
| 19 | Every statistic states source and date |
| 20 | Prices, specs and certifications exist as text, never only as images or PDFs |
| 21 | Publish and update dates agree across meta, schema and visible page |
| 22 | Target-market content is native-quality, not machine translation |
| 23 | Quarterly refresh ritual exists for top citation pages |
| 24 | Paragraphs are self-contained: one claim, fully stated, quotable out of context |
| 25 | Fan-out sub-questions for the main topics are mapped and covered |
| 26 | No keyword-stuffed legacy pages surviving on authority alone |
| 27 | Author entities real and consistent: names, bios, credentials |
| 28 | Contact, company and policy pages complete — trust surfaces engines can verify |
Group 3: Source graph (12)
| # | Check |
|---|---|
| 29 | 20-30 real buyer questions tested; the domains engines actually cite are listed |
| 30 | Every industry directory on that list is claimed and consistent |
| 31 | Review-platform profiles (G2, Capterra, Trustpilot or category equivalents) claimed and complete |
| 32 | Compliant review invitation flow running: no gating, no incentives |
| 33 | Every review answered, including negative ones |
| 34 | Genuine expert participation on Reddit, Quora or vertical forums under real names with disclosure |
| 35 | Trade media or podcast appearances on record for the current year |
| 36 | Brand facts — founding date, address, description — byte-identical across site, directories and socials |
| 37 | Organization schema carries alternateName and a complete sameAs set |
| 38 | Wikipedia/Wikidata handled honestly: present if genuinely notable, otherwise secondary anchors (Crunchbase etc.) |
| 39 | No purchased spam directories or low-quality link packages in the backlink profile |
| 40 | Local businesses: Google Business Profile complete plus NAP audit across all citations |
Group 4: Measurement discipline (10)
| # | Check |
|---|---|
| 41 | A fixed question set of 50-100 buyer questions is frozen, sourced from real customer language |
| 42 | The set is organized by decision cluster: shortlist, compare, verify, price, risk |
| 43 | The set runs on a fixed date each month, per engine, never averaged across engines |
| 44 | Every answer screenshotted and archived with date, engine and question ID |
| 45 | Four-dimension scores — Visibility, Citation Rate, Semantic Coverage, Brand Perception — recorded monthly, 0-100 |
| 46 | Hallucinations and factual errors logged separately and tracked to source-side fixes |
| 47 | Attribution notes link content and source-graph actions to score changes |
| 48 | Vendor or agency acceptance criteria written only as re-verifiable terms: baseline-to-retest deltas |
| 49 | Every new cycle structured as a 90-day POC: baseline, intervention, re-measurement |
| 50 | Zero-click economics (Ahrefs, February 2026: −58% position-1 CTR with AI Overviews, ~68% zero-click) understood by whoever reads the report |
How do you score and prioritize?
Score each group against its item count: technical out of 14, content out of 14, sources out of 12, measurement out of 10. Any group under 60% is next quarter’s priority. Typical sequences: technical gaps close in two weeks; content and source-graph groups take a quarter each; measurement starts this month because every other group is blind without it. Then re-run the whole checklist quarterly as part of the quarterly GEO audit, and validate the result against your live scores in the GEO checker.
Frequently asked questions
How long does it take to pass all 50 checks?
Technical items close in about two weeks, entity and profile fixes in two to four weeks, and content plus source-graph groups take a quarter or more. Full conformance is a 6-12 month asset-building program, not a sprint.
Can I use this checklist to evaluate an agency?
Yes — that is a design goal. Ask the agency to deliver evidence per item: screenshots, profile links, deployment records. Any item they cannot evidence is a risk point in the engagement.
Which group should I start with?
Measurement, paradoxically: it takes a day to freeze a question set and run a baseline, and every other group’s progress is invisible without it. Then technical access, because crawler blocks gate everything downstream.