The problem this solves
Most teams know their share of voice in search and their share of shelf in ads, and have no number at all for the surface buyers increasingly trust first: AI answers. The anecdotes exist - someone asked ChatGPT and we were not there - but anecdotes cannot be budgeted against. Without a denominator, "we have no AI visibility" and "we are at 12 percent and rising" sound the same in a planning meeting, and the work that would move the number never gets prioritized.
The denominator has to be built comparatively. Your citation count alone says little; the same count next to five named rivals, across the same questions, on the same engines, says everything. That is a benchmark, and building one by hand means hundreds of prompts, careful and consistent scoring, and a spreadsheet nobody wants to maintain past week two. A mission does the whole sweep in one run, scores it one way, and leaves the method written down where the next quarter can repeat it exactly.
How the mission runs
- Fix the question set and the field. The mission locks 25 buyer questions that map your funnel - problem questions, category questions, comparisons - and the five competitors to score against. Both lists are written into the deliverable, because a benchmark is only as credible as its stated method.
- Sample the engines. AI Visibility puts every question to the live answer engines and captures the full responses: who is mentioned, who is cited as a source, in what order, with what framing. Captures are timestamped so the benchmark has a defensible date.
- Score the matrix. Every answer is scored into a brand-by-question-by-engine matrix: mentioned, cited, lead recommendation, absent. Prominence is weighted, so being the default answer counts for more than trailing an also-mentioned list.
- Deliver the Sheet and the read. The matrix lands in Google Sheets with per-brand share rolled up, plus a one-page read: where you win, where you trail a specific rival, and the questions where nobody owns the answer yet - the cheapest ground to take.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A Google Sheet holding the full share-of-citation matrix - six brands, 25 questions, every engine sampled - with per-brand totals, the scoring method on its own tab, and a summary of the winnable gaps. It is the number your AI-visibility work gets measured against from now on.
Make it yours
- Re-run monthly with the same panel and add a trend tab, turning the benchmark into a time series.
- Split the matrix by funnel stage to show exactly where in the journey the assistants stop recommending you.
- Swap one rival for an adjacent-category leader to see whose content strategy the engines are actually rewarding.
Frequently asked questions
Why 25 questions and five competitors?
Large enough that one odd answer cannot swing the totals, small enough that every cell in the matrix gets scored carefully and the whole run stays affordable. The panel is yours to change; the mission just insists the panel be stated so the number stays comparable.
How is a citation different from a mention?
A mention is your brand named in the answer text; a citation is your page used as a source. Both are scored separately in the matrix, because they respond to different work: mentions follow brand presence in the training corpus, citations follow having a page worth sourcing.
Can I audit any number in the Sheet?
Every cell traces to a stored capture: the question, the engine, the full response, the timestamp. Nothing in the matrix is an impression; it is all replayable evidence attached to the mission run.