Insights/Tutorial
How to Check What AI Says About Your Firm (15-Minute Audit)
A step-by-step tutorial for auditing your firm's AI visibility across ChatGPT, Perplexity, and Google AI — no tools required, results in under an hour.
You can audit your firm's AI visibility in under an hour with no tools beyond free ChatGPT and Perplexity accounts. The output: a list of the questions your buyers ask, a record of who AI recommends today, and the source list driving those recommendations.
Step 1 — Write down 10 buyer questions (10 minutes)
List the questions a prospect would actually ask an assistant. Use plain buyer language:
- "Best [practice area] [lawyer/CPA/doctor] in [your city]"
- "Who should I hire for [the problem you solve] in [city]"
- "Is [your firm name] a good [category]"
- "[Your firm] vs [main competitor]"
- "How do I choose a [category] in [city]"
Two rules: phrase them the way clients talk on intake calls, and include at least one direct question about your firm by name.
Step 2 — Run them in a clean session (15 minutes)
Open a private/incognito window, logged out where possible — memory and history personalize answers and contaminate the audit. Run each question in:
- ChatGPT (enable search mode if it does not trigger automatically)
- Perplexity
- Google (note whether an AI Overview appears and what it says)
Run each query three times. Answers are probabilistic; one run tells you almost nothing.
Step 3 — Record three things per answer (15 minutes)
Build a simple spreadsheet with a row per query per engine:
| Column | What to record |
|---|---|
| Named firms | Every business the answer mentions, in order |
| Your status | Recommended / listed / mentioned with caveats / absent |
| Sources cited | The domains the engine cites (expand the citations panel) |
The sources column is the audit's real product. If the engine recommends competitors, the sources it cites for them — directories, review sites, local press — are the exact placements your firm is missing.
Step 4 — Read the results (10 minutes)
Score yourself honestly:
- Absent from all answers. The common case. Usually an entity problem (the engine cannot resolve who you are) plus a citation problem (no trusted source vouches for you).
- Mentioned but never recommended. The engine knows you exist but has no evidence you are the answer. Citation-source work moves this fastest.
- Recommended inconsistently. You appear in some runs, not others. This is the threshold moment — structured content and more source coverage consolidate it.
- Recommended consistently. Defend it. Track monthly, because these answers are volatile — ChatGPT's source mix has shifted dramatically in single quarters.
Step 5 — Decide what happens next
A 10-query manual audit is a snapshot, not a measurement program. Real measurement runs 100–300 queries, multiple samples each, monthly, with scoring — the methodology in our share of answers guide. But the snapshot answers the only question that matters today: when your next client asks AI who to hire, does the answer include you?
Every Result.st engagement starts with the full version of this audit across 200+ queries — contact us and we run it for you.
Frequently asked questions
Do I need paid tools to audit AI visibility?
No. A manual audit with free accounts on ChatGPT and Perplexity covers the essentials. Paid tracking tools matter later, when you need monthly measurement across hundreds of queries — not for the first look.
Why do I get different answers when I repeat the same question?
AI answers are probabilistic and can vary between sessions. That is why a credible audit runs each query at least three times and records how often the firm appears, not whether it appeared once.
What if ChatGPT recommends a competitor by name?
Record it — that is the most valuable data in the audit. The sources the engine cites for your competitor are the exact list of placements you are missing.