To track your law firm in ChatGPT and other AI assistants, write a fixed list of the questions your clients ask, by case type and city. Ask each one several times in each assistant, every month, under the same conditions. Record whether your firm is named, who else is named, and which sites are cited. Then track the share of answers that name you, not a single "position," because the order of names changes almost every time.
Partners ask us: "Are we showing up in ChatGPT?" Usually someone typed one question on a phone and saw a competitor. One screenshot can't answer that. A repeatable test can, and a firm can run it with a spreadsheet. This is the deeper version of the measurement section in our GEO guide: the method, the columns to log, and how to read them.
Why "ChatGPT ranking" is the wrong thing to track
Firms want to track ChatGPT AI rankings over time, the way they track Google rankings. But AI answers don't have stable positions.
In a study published in January 2026, SparkToro and Gumshoe.ai had about 600 volunteers run 12 prompts through ChatGPT, Claude and Google's AI Overviews or AI Mode, collecting 2,961 answers. They found the assistants returned the same list of brands less than once in 100 runs, and the same list in the same order less than once in 1,000. Their conclusion: a brand's ranking position in an AI answer is close to random, but the percentage of answers that include the brand, across many prompts run many times, is a reasonable thing to measure.
Gumshoe.ai sells AI tracking and the study wasn't about lawyers, but anyone who has asked ChatGPT the same question twice has seen this. So when you think about your brand ranking in ChatGPT, think in rates:
- Mention rate. Of all the answers to your questions, what share name your firm?
- Share of answer. Of all the firm names across those answers, what share are yours compared with each competitor?
- Citation rate. What share of answers link to a page on your website?
Position is still worth noting, because being named first in a list of three is better than being named fifth. Just treat it as a rough average over many runs, never as "we're number 2 in ChatGPT."
Step 1: Build the prompt list
Your prompt list is your scoreboard. Build it from the way your clients talk, not the way lawyers talk. Intake notes and call recordings are the best source.
Cover four kinds of question for each case type you want, in each city you serve:
| Type | Example | What it tells you |
|---|---|---|
| Who to hire | "Who's the best truck accident lawyer in Dallas?" | Whether you're named when someone is ready to call a firm. |
| Situation | "I was rear-ended in Phoenix and the other driver's insurer keeps calling me. What should I do?" | Whether a firm is suggested at all, and whether your pages are cited as the explanation. |
| Your name | "What do you know about [your firm] in Atlanta? Are they good?" | Whether the facts about you are right: phone, address, practice areas, reviews. |
| Comparison | "[Your firm] or [competitor] for a wrongful death case?" | How the assistant describes you next to the firms you lose cases to. |
Start with 20 to 40 prompts per practice area. Keep the wording natural: people write "I got hurt in an Uber, who do I call," not "rideshare accident attorney Miami."
Then freeze the list. If you reword prompts every month, you can't tell whether a change came from your work or from the wording. Add new prompts as a separate group and keep the core set unchanged for at least a year.
Step 2: Keep the test fair
AI answers depend on who is asking and from where. If the test setup drifts, your numbers drift with it.
- Put the city in the prompt. OpenAI says ChatGPT uses an approximate location from your IP address for local results, and that a VPN can change it. Naming the city keeps the test from depending on where the tester sits.
- Turn off personal history. OpenAI also says saved memory can influence ChatGPT's search results. Use ChatGPT's Temporary Chat, which doesn't create memories, or a logged-out browser.
- Use the same assistants every time. We use ChatGPT, Gemini, Perplexity and Google (AI Overviews and AI Mode). Note the model or mode if the assistant shows one.
- Run each prompt several times. Three runs per prompt per assistant is a practical minimum.
- Test on a schedule. Run the full set in the same week each month.
30 prompts × 4 assistants × 3 runs = 360 answers a month. By hand that is a few days of someone's time. That's why most firms that keep this up automate it or hand it off.
Step 3: What to log for every answer
One row per answer. A shared spreadsheet is enough. These are the columns we'd use:
| Column | What goes in it |
|---|---|
| Date, assistant, run | When, where and which repeat (1, 2 or 3). |
| Prompt ID | A short code from your frozen list, like DAL-TRUCK-01, so rows can be grouped by city and case type. |
| Firm named? | Yes or no. This drives the mention rate. |
| Position | Where your firm appeared in the list, if it did. Average it later; don't read single values. |
| Other firms named | Every firm in the answer. This drives share of answer and shows who you're really up against. |
| Cited URLs | Every link the answer shows as a source. This is the most useful column. |
| Your site cited? | Yes or no, plus which page. This is ChatGPT website rank tracking in practice: is your site one of the sources? |
| Errors | Wrong phone, old address, practice areas you don't handle, closed offices. |
| Full answer text | Paste it in. When a number moves, you'll want to read what was actually said. |
From those rows you get the monthly figures: mention rate by assistant, city and case type; share of answer against your top competitors; citation rate for your site; and the domains cited most often in answers where you weren't named. That last list is where the work comes from.
Our free AI Visibility Check asks ChatGPT, Gemini and Perplexity ten questions an injured person would ask in your city. We walk you through all thirty answers.
Step 4: Choose a method: by hand, by API, or a ChatGPT SEO rank tracking tool
| Method | Good for | Watch out for |
|---|---|---|
| By hand | A first baseline, a small prompt list, and seeing exactly what a client sees in the app. | Time. People also drift: they reword prompts, forget to log sources, or skip runs. |
| By API, with a script | Running hundreds of prompts on a schedule, with the sources captured automatically. | Someone has to build and maintain it, and API answers may not match the consumer app exactly. Spot-check by hand. |
| A paid tracking tool | Teams that want dashboards, competitor views and alerts without writing code. | Most are built for national brands. Check how they handle local questions and city settings before you buy. |
The API route. The main assistants all offer developer access with web search, and each returns the sources it used, which is what makes automated tracking possible. OpenAI's web search tool accepts an approximate user location (city, region, country) and returns cited URLs plus the full list of sources it consulted. Google's Gemini API, when grounded with Google Search, returns the searches it ran and the source links behind each part of the answer. Perplexity's Sonar API returns a list of citation URLs with each answer. A script loops through your list and writes every answer and source to your sheet.
The tool route. ChatGPT SEO rank tracking tools fall into two groups:
- SEO suites that added AI tracking. Semrush's AI Visibility Toolkit tracks prompts in ChatGPT search, Google's AI Mode and Gemini, along with mentions and cited pages. Ahrefs' Brand Radar tracks mentions and citations across AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot. Useful if your firm or agency already pays for one of these suites.
- Dedicated AI visibility trackers. Tools such as Profound, Peec AI and Otterly.AI were built for this job: you load prompts, add competitors, and they report visibility and the sources the assistants cite.
We don't get paid for mentioning any of these companies, and their coverage changes often, so check the current list of assistants on their sites. Whatever you choose, ask three questions before you sign up: Can it run my prompts as if from a specific city? Will it show me the raw answers, not just a score? Does it report a rate across repeated runs, or a single "rank"?
Step 5: Add the signals your analytics already has
Answer tracking shows whether you're named. Analytics shows whether anyone clicked. Use both.
- Google Analytics 4. GA4 now has an "AI Assistant" channel in its default channel group, for visits from sources like ChatGPT, Gemini, Deepseek, Copilot and Grok. Google notes that it excludes its own AI Overviews and AI Mode. OpenAI adds utm_source=chatgpt.com to links from ChatGPT search, so you can also filter by that source.
- Google Search Console. Clicks and impressions from AI Overviews and AI Mode are counted inside the normal "Web" search type. There is no separate AI filter, which is why Google's AI answers still need checking by hand or with a tool.
- Bing Webmaster Tools. Microsoft's AI Performance report, a public preview since February 2026, shows how often your pages are cited in Microsoft Copilot and Bing's AI summaries, which pages are cited, and a sample of the queries that led to them. It's the closest thing to a "Search Console for AI" so far, and it's free if you verify your site.
- Your intake. Analytics only sees clicks. Someone who reads an AI answer and then calls you, or searches your firm's name on Google, never shows up as an AI visit. Add "AI assistant (ChatGPT, Gemini...)" to the "How did you hear about us?" options and ask it on every call.
Put these numbers next to the answer tracking in one monthly report, so mentions are always read alongside visits and calls.
Step 6: What to do with the results
Tracking is only worth the effort if each pattern leads to a specific fix. These are the ones we see most often:
- You're not named anywhere
Start with the basics: make sure AI crawlers can read your site, your key pages are indexed, and your name, address, phone and practice areas match everywhere. Our ChatGPT guide covers the order.
- A competitor is named and you're not
Look at the cited URLs in those answers. If the same directory, review site or news story keeps coming up and you're missing from it, that's your next task.
- You're named in one assistant but not another
The assistants read different sources. Compare the cited domains assistant by assistant. Google's answers lean on Google's own index, so our Google AI Overviews guide is the place to start there.
- Your site is cited but your firm isn't named
Your page explains the law well but doesn't make it clear who you are, where you practice and which cases you take. Add that plainly, near the top.
- The facts about you are wrong
Find the source the assistant cited and fix it there. The assistant repeats what it reads.
Re-run the same prompts the next month and check whether the fix moved the number. Watch the three-month trend, not one month.
How we run it for law firms
At Bookedly this is the method behind our free AI Visibility Check: real prompts by case type and city, run across the AI assistants. Our monthly report covers where AI names you, calls, leads and signed cases, so mentions are always tied back to the cases they bring in. The tracking feeds the AI SEO work, and it shows where a paid spot under the answer, through ChatGPT Ads, would help while the organic work catches up. Your firm approves every ad and page before it goes live, and we work month to month, with no long-term contract.
For injury firms, see personal injury marketing and case pages like truck accident lawyer marketing. For the wider plan, see law firm marketing.
Questions
How do I track my law firm's ranking in ChatGPT?
Write a fixed list of questions your clients ask, by case type and city. Ask each one several times in ChatGPT every month, in a Temporary Chat or logged out, and record whether your firm is named, who else is named and which sites are cited. Track the share of answers that name you over time, not a single position.
Why does ChatGPT give a different answer every time I ask?
AI assistants write a new answer each time and run fresh web searches, so the firms they name and the order they name them in change from run to run. Research from SparkToro and Gumshoe.ai found the same list of brands came back less than once in 100 runs. That's why one screenshot tells you very little.
Is there a Search Console for ChatGPT?
No. OpenAI doesn't offer site owners a report of where ChatGPT names them. You can see ChatGPT visits in Google Analytics, through GA4's AI Assistant channel and the utm_source=chatgpt.com tag. Microsoft's Bing Webmaster Tools has an AI Performance report for Copilot and Bing's AI summaries.
How often should a law firm check AI answers?
Monthly is enough for most firms, with each prompt run several times per assistant. Checking daily mostly measures random variation.
Do I need a paid AI rank tracking tool?
Not to start. A spreadsheet and a few hours a month will give you a baseline. A tool or a script makes sense once you track more than a few dozen prompts or several cities. Before buying, check that the tool handles local questions and shows the raw answers.
What should I do if ChatGPT gets my firm's details wrong?
Look at the sources it cited for that answer and correct the information there: your Google Business Profile, directory listings and your own website. Then re-run the same prompt the next month to confirm the fix.
Sources (checked October 22, 2026)
- Rand Fishkin, SparkToro, "New Research: AIs are highly inconsistent when recommending brands or products," January 2026 (with Gumshoe.ai). sparktoro.com
- OpenAI Help Center, "ChatGPT search" (location from IP address, VPN, memory). help.openai.com
- OpenAI Help Center, "Temporary Chat FAQ." help.openai.com
- OpenAI Help Center, "Publishers and developers FAQ" (utm_source=chatgpt.com). help.openai.com
- OpenAI API docs, "Web search" (user location, citations and sources). developers.openai.com
- Google AI for Developers, "Grounding with Google Search." ai.google.dev
- Perplexity docs, "OpenAI compatibility" (citations and search results in Sonar responses). docs.perplexity.ai
- Google Analytics Help, "[GA4] Default channel group" (AI Assistant channel). support.google.com
- Google Search Central, "AI features and your website" (Search Console reporting). developers.google.com
- Microsoft Bing Webmaster Blog, "Introducing AI Performance in Bing Webmaster Tools Public Preview," February 2026. blogs.bing.com
- Semrush Knowledge Base, "Getting Started with the Semrush AI Visibility Toolkit." semrush.com
- Ahrefs, "Brand Radar." ahrefs.com
- Profound tryprofound.com, Peec AI peec.ai, Otterly.AI otterly.ai (product pages).