Product Guides2 min read • Jan 24, 2026By Jordan Reyes

How to interpret AI sentiment scores for your brand (Jan 2026 Update 2)

Abhord Quickstart Guide (Refreshed 2026 Edition)

Abhord Quickstart Guide (Refreshed 2026 Edition)

Who this is for: New Abhord users who want a fast, reliable way to learn what large language models (LLMs) say about their brand, competitors, and category—and turn that into action.

What’s new in this edition

  • Stratified model mixes: guidance for balancing frontier and lightweight/open-source models.
  • Updated defaults: lower temperatures for recall tasks; improved cost caps and throttling.
  • Mentions v2: better aliasing, deduplication, and confidence scoring.
  • Sentiment v2: domain-aware, neutral-friendly scoring with adjustable thresholds.
  • Share of Voice (SOV) refresh: rolling 28‑day window, direct vs. inferred mentions, and per-model weighting.
  • Competitor Tracker upgrades: alias manager, false‑positive suppression, and entrant detection.
  • Action integrations: cleaner exports to Slack, Sheets, Jira/Linear, HubSpot.

1) Initial setup and configuration

  • Create a workspace and roles

- Add teammates; assign roles (Admin, Analyst, Viewer).

- Enable SSO if available; create per-project API tokens for least-privilege access.

  • Connect integrations

- Slack for alerts, Google Sheets/BigQuery for exports, Jira/Linear for issue creation, HubSpot/Salesforce for handoffs.

  • Define entities and aliases

- Add your brand, products, competitors, and common aliases (spelling variants, abbreviations).

- Include negative keywords to suppress false positives (e.g., “apple the fruit” for Apple).

  • Choose model roster and controls

- Start with a stratified set: 2–3 frontier models, 2 cost‑efficient APIs, 1–2 popular open-source models via your preferred gateway.

- Defaults: temperature 0.2–0.4 for recall/listing tasks; 0.6–0.7 for ideation.

- Set monthly cost caps and per‑run throttles; choose regions to satisfy data policies.

- Enable PII redaction and domain-specific profanity/guardrail handling.

  • Project templates

- Use “Brand + Category Pulse” to get mentions/SOV; “Competitor Delta” for movements; “Messaging Tests” for copy evaluations.

2) Run your first survey across LLMs

  • Frame the objective

- Example: “Identify the top 10 CRM tools for SMBs, reasons they’re chosen, and how AcmeCRM compares.”

  • Build the prompt pack

- System message: define evaluator role and criteria.

- User prompts: 3–5 variations covering direct, comparative, and scenario prompts.

- Add disambiguation instructions (“If unsure, say ‘unknown.’ Do

Jordan Reyes

Principal SEO Scientist

Jordan Reyes is a 15-year SEO and AI search veteran focused on search experimentation, SERP quality, and LLM recommendation signals.

Ready to optimize your AI visibility?

Start monitoring how LLMs perceive and recommend your brand with Abhord's GEO platform.