Abhord Quickstart Guide (2026 Refresh)
This practical guide gets new Abhord users from zero to insights—fast. It reflects platform improvements and market shifts through early 2026, including broader LLM coverage, higher-fidelity sentiment, and more reliable share‑of‑voice (SOV) scoring across multimodal answer engines.
What’s new since the last edition
- Broader engine coverage: run surveys across leading LLMs and answer engines (e.g., ChatGPT, Perplexity, Bing/Copilot, Google SGE) with per‑engine weighting.
- Model snapshots: lock runs to a model/version to keep time‑series results comparable.
- Improved entity resolution: fewer false positives (brand synonyms and look‑alikes handled automatically, with manual overrides).
- Sentiment 2.0: aspect‑level polarity and confidence scoring for cleaner, decision‑ready insights.
- Weighted SOV: visibility- and engine-weighted share of voice to better reflect real user impact.
- Alerting: Slack/Teams and email alerts for rank losses, competitor surges, and new claims.
1) Initial setup and configuration
- Create a workspace
- Add your company name and default market (locale and timezone).
- Invite collaborators with roles: Viewer, Analyst, Admin.
- Define entities
- Add your brand, product lines, and known variants (e.g., “Acme Pro,” “AcmePro,” old brand names).
- Add competitors with canonical names and common misspellings.
- Configure disambiguation rules (e.g., “Monday” ≠ “Monday.com”).
- Seed intents and prompts
- Start with templates: “Best-of” discovery, comparisons, pricing, troubleshooting, and alternatives.
- Example seeds:
- “best [category] for [persona]”
- “[brand] vs [competitor]”
- “cheapest/budget [category]”
- “[brand] pricing” and “[brand] alternatives”
- Connect destinations and governance
- Set Slack/Teams channels for alerts, choose weekly digest email.
- Enable SSO and data retention windows (e.g., 180 days).
- Configure privacy filters to redact sensitive strings.
- Budget and run controls
- Assign a monthly cap and per‑survey limits.
- Choose run cadence defaults (ad‑hoc, weekly, or monthly).
- Optional: connect your own API keys for specific models if you want stricter parity with your end‑user surfaces.
Pro tip: Lock “Model Snapshot” for baseline studies so week‑over‑week changes reflect the market, not a silent model update.
2) Running your first survey across LLMs
- Define the objective
- Example: “Measure our visibility and positioning for SMB project management queries in the US.”
- Build the survey
- Pick a template (e.g., Best‑of Discovery).
- Add 10–25 prompts covering variants, modifiers, and personas.
- Select engines/models (aim for 3–5 to start). Keep one as your “north star.”
- Configure run parameters
- Locale/market: e.g., en‑US.
- Persona: neutral (unless you target a niche like “IT admin”).
- Regenerations: 3–5 per prompt/engine to reduce randomness.
- Randomize brand order in comparison prompts to minimize prompt bias.
- Execute and monitor
- Run immediately and watch the queue.
- Use “Preview Mentions” to validate entity capture before scaling.
Quality tips
- Freeze citations capture: store excerpt + URL when engines provide it.
- Enable “Strict Deduping” to avoid double‑counting repeated mentions in a single answer.
- If budgets are tight, downsample regenerations but increase prompt diversity.