What KPIs should a brand be tracking for AI-era reputation health?

Five KPIs cover a brand’s AI-era reputation health: AI sentiment, AI source quality, AI peer comparison, AI accuracy and AI narrative drift. Each one has to be tracked per engine rather than averaged. The same query can return materially different answers across ChatGPT, Gemini, Copilot, Perplexity, Claude, Grok, Google AI Overviews and Google AI Mode, and an average across models hides where the problem actually sits.

The five KPIs, and what each one captures

AI sentiment (per model)

The tone of each engine’s responses about the entity: positive, neutral or negative. The models diverge, so one engine can be warm on a topic while another is hostile about it.

AI source quality

Which sources each model draws on. AI engines assemble answers from underlying source content and weight that content by credibility, so a narrative resting on weak or hostile sources is fragile no matter how positive it reads today.

AI peer comparison

How the entity is positioned against its competitors inside the engines’ answers. Reputation is relative, so the comparison carries as much weight as the standalone read.

AI accuracy

Whether the engines state correct facts about the entity. Fluent, confident misinformation is its own risk, separate from tone.

AI narrative drift

How the framing changes over time. Model updates and shifting source pools can move the narrative when nothing about the entity itself has changed.

Why per-engine, not an average

Averaging the KPIs across engines hides divergence. A single fix does not propagate uniformly across the models, so a problem confined to one engine disappears inside a blended score. Tracking each KPI per model, the approach Five Blocks uses in AIQ™, is what shows where a narrative is breaking down.