Most teams find out how AI assistants describe them by accident - a prospect mentions it on a call, or someone tries a search in ChatGPT at lunch. This post lays out the deliberate version: a repeatable workflow for measuring your brand's AI visibility, using Threadly's actual workflow as the example.
Step 1: Track questions, not keywords
AI assistants are conversational, so the unit that matters is the full buyer question. Good tracking questions sound like a real person asking for advice: "What are the best async video messaging tools?", "Loom vs Vidyard for a remote sales team", "Affordable Loom alternatives for startups". Aim for a mix of category questions, head-to-head comparisons, alternatives queries and problem questions.
In Threadly you add these as tracked questions on your brand profile, along with your domain and your competitors. If you are not sure which questions your buyers ask, the app can suggest a starting set from your brand description - then you edit it down to the ones that matter.
Step 2: Scan across all three providers
A scan sends every tracked question to each configured provider - ChatGPT, Claude and Gemini - and reads the full answer text. Scanning one provider alone is misleading, because the three assistants build their answers from different model knowledge and different retrieval sources. A brand can lead in one assistant's answers and be invisible in another's.
For every question and provider, Threadly records whether your brand was mentioned, whether your domain was cited as a source, and your position among the brands named. The full answer text and citation list are kept, so you can always read exactly what the AI said rather than trusting a score.
Step 3: Read the four headline metrics
- Mention rate tells you how often you appear at all. If this is low, the problem is awareness in the model's sources, not wording.
- Citation share tells you how often your own site is the evidence. Low citation share with decent mentions means the models know you but rely on third parties to describe you.
- Average position tells you whether you lead the shortlist or trail it.
- Share of voice compares your mentions to your competitors' across the whole question set - the single best summary of who owns the category conversation.
Step 4: Compare competitors question by question
The aggregate numbers tell you where you stand; the per-question table tells you what to do. Sort for questions where a competitor is mentioned and you are not. Each one is a concrete gap - usually a missing comparison page, an unanswered question on your site, or a third-party conversation you are absent from.
Step 5: Watch the trend, not the day
AI answers drift. Models are updated, retrieval indexes refresh, and new content appears. A single scan is a snapshot; the value is in the trend line. Threadly keeps every scan in your history, so mention rate, citation share and share of voice become curves you can act on - and so you can see whether the content you shipped last month actually moved the answers.
The whole loop - questions, scan, metrics, gaps, trend - takes minutes to set up. Try it on the live demo first if you want to see the output before connecting anything.