Net Sentiment Score

Every AI response that mentions a brand is classified by an LLM judge as Positive, Neutral or Negative based on the full context of the answer (not a keyword score). The judge handles hedging, qualified praise and conditional framing — a sentence like “X is solid for small teams, but enterprise buyers should look elsewhere” lands as Mixed, not Negative.

Net Sentiment Score is computed as:

NSS = (% Positive − % Negative + 100) / 2

The +100 / 2 normalization shifts the raw difference (which ranges from -100 to +100) into a 0–100% scale that's easier to read and benchmark. A brand with 60% Positive and 10% Negative responses has an NSS of 75%; a brand with 40% Positive and 40% Negative has an NSS of 50% — neutral framing on average.

NSS is computed per brand and per engine, because the same brand can land at very different scores across ChatGPT, Claude, Perplexity, Gemini and AI Overviews on the same prompt set.

Raw Positive / Neutral / Negative percentages tell you the distribution; NSS tells you the headline. Both are useful for different reasons:

Raw sentiment answers “what does the response mix look like?” — useful for spotting shifts in the Negative tail or the rise of Mixed responses, which signal hedging.

Net Sentiment Score answers “is the framing net favorable or unfavorable?” — useful for tracking direction over time, benchmarking against competitors, and reporting to executives in a single number.

Treat NSS as the headline metric and raw distribution as the diagnostic. A brand whose NSS is rising while Negative is also rising (because Neutral is collapsing) is a different story from a brand whose NSS is rising because Positive is genuinely growing.

AI assistants now describe brands to buyers before the buyer ever lands on a vendor site. Whether ChatGPT calls your product “the obvious choice for fast-growing teams” or “a viable but pricier option” shapes consideration directly — and that framing is invisible to traditional brand metrics.

NSS makes that framing measurable. Trending NSS up over a quarter means buyers are increasingly hearing favorable descriptions of your brand inside AI; trending down means the opposite. Watching competitor NSS alongside your own surfaces who is winning the framing battle in your category, engine by engine.

Most usefully, NSS pinned to specific drivers — pricing, support, integrations, performance — turns a single number into an actionable view. A 6-point NSS drop driven entirely by negative pricing comparisons is a different problem from one driven by support complaints, and NSS plus driver clustering surfaces the difference cleanly.

Five tactics for using NSS well:

1. Track NSS per engine, not just per brand. A 75% NSS on Perplexity and a 48% NSS on AI Overviews for the same prompt set is normal — different engines weight different sources. Treat each engine as its own scorecard.

2. Benchmark against tracked competitors. Absolute NSS numbers are less interesting than relative ones. A 62% NSS is good if competitors are at 50%, weak if they're at 75%.

3. Watch the trend more than the level. Monthly NSS shifts of a few points are normal; sustained 5%+ drops or jumps are the signal worth chasing back to drivers.

4. Pin NSS shifts to drivers. Use the driver clustering attached to NSS to know which topics — pricing, support, product gaps — are responsible for movement. NSS without drivers is a headline without a story.

5. Alert on threshold crossings. A 5%+ NSS drop on any tracked engine, or a competitor passing your NSS, is worth paging the brand team for. Set thresholds per engine, not globally.