How to Run an AI Answer Audit for Your Law Firm

What the audit tells you

An AI answer audit is a structured way to ask: when a client asks an AI tool about my practice area and my city, does the tool name my firm, what does it say, and where does it get the information? It will not give you a ranking. SparkToro's January 2026 research, covering 2,961 runs of 12 prompts across ChatGPT, Claude and Google's AI, found that the chance of two responses listing the same brands was under 1 in 100. A single answer is an anecdote. A few hundred answers can show a pattern.

Step 1: Build your prompt list

Write 20 to 40 prompts in the words a client would use. Mix the types.

• Problem prompts: "What should I do if my landlord won't return my security deposit in [state]?"
• Choosing prompts: "How do I choose a family lawyer in [city]?"
• Recommendation prompts: "Who are good estate planning attorneys near [neighborhood]?"
• Cost and process prompts: "How much does a DUI lawyer cost in [state]?"
• Brand prompts: "What is [Firm Name] known for?"

Pull the wording from intake calls, emails and your Search Console queries. Keep a fixed copy of the list so that you can repeat it each month.

Step 2: Choose the tools and settings

Pick the AI tools your clients are likely to use, such as ChatGPT, Google (AI Overviews and AI Mode), Perplexity, Claude and Microsoft Copilot. For each tool, record the date, whether you were signed in, your approximate location and the mode used. Results can vary with these factors, so keep them constant from month to month.

Step 3: Protect confidential information

Use only generic, public-style questions. Never paste client names, matter facts or documents into a tool. ABA Formal Opinion 512, issued July 29, 2024, says lawyers using generative AI tools must consider duties of competence, confidentiality, communication, supervision, candor, meritorious claims and reasonable fees. An audit prompt list should contain nothing you would not say in public.

Step 4: Run each prompt several times

Run every prompt three to five times per tool, in fresh conversations so earlier answers do not shape later ones. Copy the full answer into your sheet or save a screenshot. This is tedious, which is why it works: it gives you rates, not impressions.

Step 5: Record the results

Step 6: Calculate and read the results

1. Appearance rate = runs that name your firm divided by total runs, by tool and by prompt type.
2. Citation rate = runs that cite one of your pages divided by total runs.
3. Top outside domains = the domains cited most often across all runs. Those are the directories, news sites and bar pages that matter in your market. See third-party profiles and reviews.
4. Error list = every wrong fact about your firm, with the page or profile it likely came from.

Do not read too much into small differences. If your rate moves from 10% to 15% on 30 runs, that is two or three answers. SparkToro's authors advise treating visibility as a percentage across many prompts and many runs, and distrusting any "ranking position."

Step 7: Act on what you find

• If you never appear for problem prompts, you may lack a page that answers the question. Use the page-writing method.
• If tools describe you incorrectly, correct the source: your site, a profile or a listing. See attorney bios and firm identity.
• If tools cite competitors' pages for questions you answer, compare their page with yours for clarity, accuracy and identity details.
• If your pages are not cited at all, check the technical readiness checklist, since a page that cannot be indexed cannot be cited.

Step 8: Repeat and compare

Run the audit monthly at first, then quarterly. Keep every sheet. Change one thing at a time when you can, and note the dates of changes. Combine the results with your other measures from metrics that matter in AI search, including Bing's AI Performance citations and your GA4 AI channel.