Skip to main content
A compact reference for the vocabulary used throughout Onsomble.

Core concepts

AI discoverability How well a business shows up when customers ask AI assistants for recommendations, comparisons, or answers. The industry calls the practice of improving it AEO (Answer Engine Optimisation), GEO (Generative Engine Optimisation), or LLMO (LLM Optimisation). See What is AI discoverability?. Site The workspace for a single business website or brand presence in Onsomble. All discoverability setup, scan history, and insights are scoped to a Site. One business with one website is one Site. See What is a Site?.

Discoverability terms

Scan A run of the discoverability check. Onsomble sends each active prompt to each enabled AI model, captures the responses, and refreshes the Site’s insights. Scans can run ad-hoc or on a recurring schedule. Prompt A realistic question a customer might ask an AI assistant. Prompts are the inputs each scan runs against. A scan usually includes ten to fifteen prompts covering different customer intents. Prompt library The set of prompts Onsomble runs during scans for a Site. The library can be edited at any time to reflect changes in what customers are asking or what the business wants to measure. Model An AI assistant Onsomble queries during a scan: currently ChatGPT, Claude, Gemini, and Perplexity. Each scan can target one model or multiple. Citation A mention of the business in an AI assistant’s response, or a source the assistant links when making one. See References for how citations are analysed. Visibility How often the business appears in AI answers across the prompts in a scan. The foundational discoverability metric, tracked over time on the Brand Scorecard. Share of voice When the business is mentioned, how much of the AI answer is about it, as opposed to competitors mentioned in the same answer. Distinct from visibility: you can appear often but have low share of voice if each mention is only a one-line reference. Gap The remaining visibility opportunity: how far the business sits behind its strongest competitor. Lower is better. Health A composite score across visibility, citations, share of voice, and competitor gap. The single-number summary of a brand’s AI discoverability. Sentiment Whether descriptions of the business across a scan are positive, neutral, or negative. Most descriptions sit in neutral; the outliers are what warrant attention. Competitor A business Onsomble measures yours against during each scan. Competitors drive citation benchmarks, share of voice, and sentiment comparisons. See Setting up a scan. Recommendation A specific, prioritised action Onsomble suggests based on scan results: a content gap to close, an inaccuracy to correct, a positioning improvement to make, and so on. See Acting on recommendations. Ingestion The process Onsomble uses to crawl and understand a business’s website. Ingestion informs competitor suggestions and prompt suggestions. See Ingesting your website.

Account and agency terms

Agency account The account type every Onsomble account uses, shown under Settings → Organisation. It holds Sites for your own business and, if you work for clients, a workspace for each client with that client’s Sites. See Agency setup. Portfolio A named collection on the Portfolios page of an agency account. It holds clients, plus Sites and topics added on their own; a client brings all of its Sites. See Portfolios. Portfolio dashboard The landing view that shows all clients at once, for triage across the portfolio. Free trial The period every new account starts with, before choosing a plan. It needs no card and includes a prompt execution allowance. When it ends, the account pauses until the owner chooses a plan; everything set up is kept. See The free trial. Plan The Onsomble subscription, which sets your monthly prompt executions and which AI models your scans can include. See Plans and entitlements. Prompt pack Extra prompt executions bought on top of a paid plan from Settings → Billing. Used only after the month’s allowance is gone; never expire. See Prompt packs. Prompt execution The unit your plan’s monthly allowance is counted in: one prompt run on one AI model in one region. Running it again counts again, so 10 prompts × 3 models × 2 regions is 60 executions.

Product-adjacent terms

Model provider The company operating an AI model: OpenAI (ChatGPT), Anthropic (Claude), Google (Gemini), Perplexity. Model behaviour and outputs change as providers update their systems. Model update A change released by a model provider that can alter how the model describes businesses. Sometimes produces scan-level shifts even without any action on the business’s side. Comparing movement across a portfolio is often the fastest way to tell a model update apart from something specific to a single business. Third-party signal Content about the business that lives outside the business’s own website: directory listings, press coverage, reviews, partner mentions. AI models draw heavily on this material, making it a long-horizon but high-impact discoverability lever.