How to Run a Full AI Visibility Audit (Step by Step)

A full AI visibility audit establishes, with numbers, where your firm stands in AI answers: which queries you appear for, who appears instead, and which sources drive those answers. It is the baseline every measurement program, and our 60-day guarantee, is judged against. Here is the complete methodology.

Step 1: Build the query universe (100–300 queries)

Start from buyer language, not keyword tools. Sources, in order of value:

1. Intake calls. The exact phrasings prospects use when they describe their problem.
2. Practice areas × locations. "Best [specialty] in [city]" for every combination you serve.
3. Problem-first questions. "What kind of lawyer do I need for X"; prospects who do not yet know the category.
4. Comparison and validation queries. "[Your firm] reviews," "[Your firm] vs [competitor]," "is [firm] good."

Aim for 100 queries for a single-market firm, 300+ for multi-market. Then freeze the set; comparability across months depends on it.

Step 2: Sample correctly

For each query, on each engine (ChatGPT, Perplexity, Google AI Overviews at minimum):

• Run in a clean session: logged out, no memory, no history.
• Sample three times minimum per cycle. AI answers vary between runs; appearance rate is the datum, not a single appearance.
• Record the full answer, not just your status; competitor data is half the audit's value.
• Capture the citation panel: every source the engine used.

A 100-query audit across three engines at three samples each is 900 runs. This is why real programs automate it, but the methodology is identical done by hand at smaller scale, as in our 15-minute version.

Step 3: Score every answer

Use a consistent rubric so the number means the same thing every month:

Appearance	Score
The recommendation, or first named	1.0
Named among options	0.5
Mentioned with caveats	0.25
Absent	0

Your share of answers is the average across all query-engine-sample combinations. Report it overall, per engine, and per practice area; the breakdowns reveal where the problem actually lives. Full scoring details in the share of answers methodology.

Step 4: Map the citation graph

For every answer where a competitor appears and you do not, log the cited sources. After a few hundred runs, a pattern emerges: a shortlist of 10–30 domains, directories, review platforms, local press, Reddit threads, that drive the recommendations in your market.

That shortlist is the audit's most actionable output. It is not a generic "get listed everywhere" recommendation; it is the specific, evidence-based set of placements that move answers in your market.

Step 5: Diagnose the failure mode

Audits surface one of four conditions, each with a different fix:

• Entity failure. Engines cannot resolve who you are: inconsistent names and addresses across the web, no structured data. Fix first; nothing else works without it.
• Citation absence. Engines know you but no trusted source vouches for you. The citation-graph shortlist from Step 4 is the work plan.
• Extraction failure. Your site is retrieved but never quoted; answers are buried in narrative prose. Fix with content engineered for extraction.
• Volatile presence. You appear inconsistently. Usually thin source coverage: one or two citations carrying you. Broaden the base.

Step 6: Set the cadence

Re-run the identical audit monthly. Report share of answers against baseline, per-engine movement, queries won and lost, and the citation changes behind each shift. Expect Perplexity to move in weeks, ChatGPT in months, and expect volatility, since engine source-mixes have shifted by double digits within single quarters.

Every Result.st engagement opens with this audit across 200+ queries, and the monthly re-measurement is the number our guarantee is judged on. Contact us to see your baseline.