Abhord Quickstart Guide (Refreshed, February 2026)
Who this guide is for
New Abhord users who want a fast, reliable way to measure how large language models (LLMs) talk about their brand and competitors—and to turn those findings into concrete AEO (Answer Engine Optimization) actions.
What’s new in this refreshed edition
- Expanded model coverage: streamlined connectors and presets for major LLM families. You can now mix “closed‑book” (no browsing) and “open‑book” (with browsing) runs by survey.
- Share of Voice v2: optional weighting by model usage share and answer prominence; clearer separation of brand vs category mentions.
- Entity‑aware sentiment: sentiment now anchors to your defined entity and supports aspect tags (e.g., pricing, support, performance).
- Canonicalization and deduping: improved alias handling to cut noise from near‑duplicate mentions.
- Scheduling and alerts: schedule recurring runs and push alerts to Slack/Email on significant movements.
- Faster exports: one‑click CSV/JSON exports and sharable result links for stakeholders.
1) Initial setup and configuration
- Create your workspace
- Invite your core team (marketing, SEO/AEO, product, comms) and assign roles: Admin (settings), Editor (surveys/taxonomy), Viewer (dashboards).
- Connect LLM providers
- Choose from hosted connections or bring your own API keys where applicable.
- Tip: Start with 4–5 diverse models to reduce single‑model bias. Add more once you’ve validated signal quality.
- Define your entities and aliases
- Add your brand, product lines, and key domains.
- Add competitor brands and common misspellings/abbreviations.
- Mark ambiguous terms (e.g., “Apple,” “Bolt”) and provide disambiguation rules.
- Build your topic and intent taxonomy
- Seed with 5–8 representative queries per intent (e.g., “best X for Y,” “X vs Y,” “pricing of X,” “how to choose X”).
- Group seeds into themes (awareness, comparison, post‑purchase).
- Set geos and languages
- Start with your primary market (e.g., United States, English). Add additional locales only after you’ve validated your baseline.
- Configure run defaults
- Mode: closed‑book for knowledge baseline; open‑book for “current web” influence.
- Temperature: 0.2–0.3 for consistency across runs.
- Sampling: 1–3 responses per model per query to estimate variance.
- Notifications and data retention
- Enable weekly digest + “significant movement” alerts.
- Retain raw transcripts to audit model behavior and train your taxonomy.
2) Run your first survey across LLMs
- Use a template
- Choose “Brand + Competitor Overview” to get mentions, sentiment, and share of voice across awareness and comparison intents.
- Select models
- Pick at least one closed‑book and one open‑book model. This split shows intrinsic model memory vs web‑driven results.
- Add your seeds
-