AI visibility analytics is the foundation of Narrashift. When a buyer asks an AI assistant which software to use, the answer usually names only a handful of products. If your SaaS is not among them, you lose the opportunity before the buyer ever visits your website. Traditional analytics cannot see this moment, because it happens inside a chat window. Narrashift makes it measurable by asking the same questions your buyers ask, on a fixed schedule, and analyzing every answer in the same, consistent way.
What Narrashift measures
Every monitoring run sends each active prompt to each configured provider: ChatGPT through the OpenAI API, Gemini and Claude through their official APIs, with web search enabled where the provider supports it. Each answer is stored exactly as it was returned, together with the model that produced it and the sources it cited. An analysis step then extracts which brands appear, whether each one is recommended or only mentioned, the sentiment of the description and any strengths or weaknesses stated about it.
From those observations Narrashift calculates the KPIs on your overview. Brand visibility is the share of analyzed answers that mention your product. Recommendation rate counts only explicit recommendations, so a passing mention does not inflate the number. Domain citation rate looks at answers that cite sources and checks whether your own website is among them. The AI Visibility Score combines the three, weighted 60/30/10, into a single number you can follow over time.
Share of voice that is not inflated by your own name
Share of voice tells you how much of the conversation belongs to you compared with competitors. Narrashift calculates it on unbranded questions only — the prompts that do not already contain your product name. A question like “Is Acme good for agencies?” will almost always mention Acme, so including it would make your share look better than it is. Restricting the calculation to questions where buyers have not yet chosen gives you an honest picture of spontaneous discovery.
Change over time, not just a snapshot
Every KPI shows its change against the monitoring cycle from at least a week earlier, as percentage points for rates and as plain numbers for counts. Hovering a change shows the earlier value and warns you when the comparison is not like-for-like, for example because questions, providers or models changed between the two runs. The visibility chart scales itself to your data so that small but real movements remain visible instead of being flattened against a fixed 0–100% axis.
You can filter the whole overview by provider and by date range. This matters because the assistants behave differently: a product can be well represented in one model and almost absent in another. Seeing the providers side by side helps you decide where to focus and explains why a change appears in one channel before the others.
Evidence behind every number
A metric is only useful if you can trust it. In Narrashift every number can be traced back to the answers it is based on. Recent answers are listed on the overview with the provider, whether your brand appeared, how many competitors were named and the sentiment, and each one opens the full response with its citations. When a KPI moves, you can read the actual answers that moved it rather than guessing.
Narrashift also tells you what it does not know. Answers that failed or are still being analyzed are excluded from rates instead of counting as absences, and the overview states how many analyzed answers each figure is based on. API answers can differ from what consumer apps show, so the platform labels its measurements accordingly and never presents them as identical to every user's experience.