Abhord Quickstart Guide (Refreshed Edition)
This practical guide helps new Abhord users stand up a generative engine optimization (GEO/AEO) program in under a week. It covers setup, running your first cross-LLM survey, interpreting results, competitor tracking, and turning insights into action.
What’s new in this refreshed edition
- Normalized share-of-voice (SOV) across models and locales: apples-to-apples comparison is now default.
- Source Trace 2.0: improved citation capture and clustering to reveal which pages most influence answers.
- Multi-locale presets: faster rollout in new markets with language, model, and query-tuning bundles.
- Mentions+ taxonomy: entity disambiguation for brand variants and product lines (e.g., “Acme” vs “Acme Pro”).
- Action Playbooks: one-click recommendations mapped to web, content, PR, and product tasks.
1) Initial setup and configuration
Goal: create a clean data foundation so results are consistent and actionable.
- Create a workspace
- Organization name, primary domain(s), and brand handles.
- Choose home locale and time zone for reporting.
- Define entities
- Add your brand, products, and key people as entities.
- Enter canonical names, common aliases, and disallowed collisions (e.g., “Atlas” the product vs. “Atlas” the gym).
- Optional: upload a factsheet (CSV) with product specs, pricing tiers, and must-have claims for validation checks.
- Connect sources (optional but recommended)
- Verify site ownership (TXT or file) for deeper crawl and structured data checks.
- Add owned channels you want LLMs to cite (docs, blog, help center).
- Select model roster
- Start with the default balanced roster spanning leading US/EU and APAC models.
- Keep “Auto-refresh models” on so your studies track newly-deployed versions without manual edits.
- Tip: Retain at least one consistent “control model” for longitudinal comparability.
- Configure locales and languages
- Use multi-locale presets (e.g., EN-US, EN-GB, DE-DE, JA-JP).
- Turn on “regional compliance” if your category has location-specific claims.
- Set governance and alerts
- Choose PII redaction level.
- Set alert thresholds for sentiment drops (e.g., -10 points week-over-week) and competitor surges (+5 pts SOV).
- Save as Baseline v1
- This locks current configuration so you can measure improvement against a stable starting point.
2) Running your first survey across LLMs
Goal: capture how models talk about your brand today, by intent and locale.
- Pick a template
- Start with Brand Perception Baseline or Alternatives & Recommendations.
- Templates include a proven “question bank” for discovery, comparison, pricing, and troubleshooting intents.
- Define intents and prompts
- Include at least 12–18 prompts spanning:
- Discovery: “Best X for Y?”, “Top tools for …”
- Comparison: “Brand A vs Brand B”, “Is Brand A worth it?”
- Task-based: “How do I … with X?”, “Fix … in X”
- Commercial: “Pricing for X”, “Discounts/coupons for X”
- Add 2–3 “negative” prompts to probe risk: “Why avoid X?”, “Common complaints about X.”
- Sampling and run settings
- Sample size: n=50 per model per locale (good first pass).
- Temperature sweep: 0.2 and 0.7 to capture deterministic and creative modes.
- Turn on Source Trace 2.0 and Mentions+.
-