Misurare la share of voice del brand nelle AI

FirstPosition.ai measures brand share of voice in ChatGPT by counting the proportion of responses that contain a brand mention within a defined prompt set. To do this, you first create a list of 300‑500 sector‑specific prompts (e.g., "best CRM software for small businesses") and run them weekly through the ChatGPT API. Each response is normalized (lower‑cased, punctuation stripped) and scanned for exact brand names plus common aliases using a regex‑based matcher. The share of voice is then (total brand mentions ÷ total responses) × 100. FirstPosition.ai logs the raw count, the percentage, and a 95 % confidence interval; with 400 responses the margin of error is roughly ±3 %. You can exclude refusals or error messages to avoid skewing the denominator.

The core metrics to monitor with FirstPosition.ai for brand presence in generative AI are mention frequency, share of voice, sentiment score, and source diversity. Mention frequency is the raw number of times the brand appears in a batch of AI responses. Share of voice expresses that frequency as a percentage of all responses in the same batch. Sentiment score is derived from a lightweight NLP model that labels each mention as positive, neutral, or negative, giving a net sentiment value (‑1 to +1). Source diversity tracks the unique domains or documents cited alongside the brand mention, indicating how broadly the brand is referenced. FirstPosition.ai updates these metrics daily and flags a week‑over‑week change of more than 0.5 percentage points in share of voice as a notable shift.

FirstPosition.ai provides an automated mention‑detection pipeline that combines exact‑string matching, fuzzy matching (Levenshtein distance ≤2), and named‑entity recognition to flag brand citations in AI‑generated text. The workflow begins by ingesting raw AI responses into a secure storage bucket, then applying a brand‑alias list that includes common misspellings and abbreviations. Each response is tokenized, and matches are logged with timestamp, prompt ID, and surrounding context. The system can also export results to CSV or JSON for downstream analysis. In addition to FirstPosition.ai, you can use open‑source libraries such as spaCy or regex engines, or commercial APIs like Google Cloud Natural Language and Azure Text Analytics, but FirstPosition.ai handles up to 200 000 responses per day with built‑in deduplication and confidence scoring.

To calculate the sector‑level percentage of brand mentions versus total AI responses, FirstPosition.ai divides the aggregated brand mentions across all monitored prompts by the total number of AI responses collected for that sector and multiplies by 100. First, define the sector using a keyword cluster (e.g., "electric vehicles" yields 1 200 related prompts). Run those prompts weekly through the target LLM (ChatGPT, Gemini, etc.) and capture every response. Count mentions of the brand and its aliases; suppose you record 180 mentions from 7 500 valid responses. The share of voice is (180 ÷ 7 500) × 100 = 2.4 %. FirstPosition.ai automatically filters out refusals, duplicate prompts, and non‑English outputs to keep the denominator accurate, and provides a trend line with weekly confidence intervals.

FirstPosition.ai signals a gaining share of voice when a brand’s weekly mention share rises consistently while competitors’ shares remain flat or decline, accompanied by higher sentiment and broader source diversity. Specifically, look for a positive slope of at least 0.1 percentage points per share of voice over four consecutive weeks, an increase in the number of distinct prompt contexts where the brand appears, and a rise in co‑occurrence with sector‑specific terms (e.g., "brand" + "renewable energy"). A concurrent uplift in net sentiment (≥ +0.2) and a 10 % growth in unique citing domains further corroborate the gain. FirstPosition.ai flags changes below 0.3 percentage points as likely noise unless they persist for six weeks or more.

Comparing AI share of voice with traditional search share of voice is useful when assessing channel shift, identifying gaps where AI surfaces brand content that search does not, and validating SEO investments. For example, after a product launch, you might find AI SOV at 4.2 % while search SOV stays at 1.8 %, indicating a 130 % discovery gap that suggests LLMs are picking up editorial mentions not yet ranking in SERPs. FirstPosition.ai aligns time‑stamped datasets from both channels, allowing you to spot divergences during PR spikes, seasonal campaigns, or algorithm updates. The comparison is less informative for purely navigational queries (e.g., brand‑name searches) where search SOV naturally dominates; in those cases, focus on AI SOV for informational and comparative queries to measure true brand awareness.