# Link Building for GEO (getting cited by AI answer engines)

> Only the non-obvious moves. Standard hygiene (E-E-A-T, freshness, sitemaps) is assumed. Priority marks: 🔴 critical · 🟠 high · 🟡 medium.
> Online version and Excel sheet: https://huskyhamster.com/knowledge-base/checklists/link-building-for-geo/

## A. Measurement (do this before touching anything)
- [ ] 🔴 Longitudinal presence measurement: min. 4 runs/month on 100+ prompts; report % presence and rotation (entries/exits), never a single position
- [ ] 🟠 Scrape public UIs (ChatGPT/Gemini/Perplexity HTML), skip APIs — API answers correlate worst with what humans see
- [ ] 🟡 Simulate 2–3 buyer personas with history/context; compare against anonymous answers
- [ ] 🟠 Mention × citation quadrant per brand: many mentions / few citations → defend mentions; few / many → push your own domain's citability
- [ ] 🟠 Track 6 intents separately: definition, validation, exploration, comparison, ranking, recommendation
- [ ] 🟡 Training-data test: query with web search OFF; recommended from memory = you're in training (mark as secured)

## B. Getting into training data
- [ ] 🟠 Full, valuable English version of key pages (≈95% of Common Crawl is English — no blind auto-translate)
- [ ] 🟡 Don't block CCBot in robots.txt; link/ping key URLs; goal = presence in the NEXT training run

## C. Architecture
- [ ] 🟠 Title + meta as a full summary the model can answer from without entering the site
- [ ] 🟠 Explicit entities in JSON-LD (Organization/Person/Product) + sameAs; A/B on subpages
- [ ] 🟡 Key answers as standalone prose Q→A paragraphs (2–4 sentences); lists only when natural

## D. Entity and gap work
- [ ] 🔴 Gap matrix: competitor × intent × category × funnel stage; publish exactly where "competitor is, we're not"
- [ ] 🟠 Own the "definition" intent for your category — the most stable layer
- [ ] 🟠 Dominate comparisons: "X vs Y" content, own + UGC, be the default mention
- [ ] 🟡 Export domains citing competitors in a given intent; buy placements exactly there
- [ ] 🟠 Repeat across different intents/contexts (one consistent story), not 100× the same claim

## E. Platforms
- [ ] 🟠 Match source type to platform: Gemini → gov/edu/medical; Perplexity → e-commerce/comparison; ChatGPT → Reddit/YouTube/forums
- [ ] 🟠 Authentic UGC layer: Reddit threads, YouTube with transcripts, industry forums
- [ ] 🟠 Full transcripts on-site for every video/podcast
- [ ] 🟠 Parallel presence on 3+ platforms (brands on 3 platforms hold ~86% of mentions)
- [ ] 🟡 Complete map profiles (Apple/Bing/HERE/Waze standard = Google-level completeness)
- [ ] 🟡 Bing Places yes; but don't bet on IndexNow as the ChatGPT lever — ChatGPT cites Bing-invisible pages too
- [ ] 🟠 Narrow targeting: win TOP 1–3 of one narrow intent/category, not "visibility anywhere"

## What NOT to do
- [ ] Don't monitor sentiment as a KPI (~0.16% of AI mentions are negative — AI recommends, it doesn't critique)
- [ ] Don't ship one-off position reports (50% of #1s are unstable within 30 minutes)
- [ ] Don't fire 100 identical brand mentions into one intent
- [ ] Don't fake reviews — one penalised account wipes months of work
- [ ] R&D only, never client work: new-entity creation, cloaking for chat crawlers

---
**Husky Hamster**: link building for SaaS, AI products and software houses
Questions about this checklist? Andreea Paluszak, Head of Outreach: hello@huskyhamster.com · Book 30 minutes: https://huskyhamster.com/book-a-call/
20-22 Wenlock Road, London N1 7GU, United Kingdom · huskyhamster.com