As a leader at your professional services firm, you’ve poked around with ChatGPT and Claude. In fact, AI has probably removed the tedium of writing an RFP or helped you think through an update to your sales deck.
But have you asked your favorite AI platform about your own firm yet?
Your curiosity has surely led you down that path, and it’s a good one… but it’s only a starting point. While asking ChatGPT about your firm will give you some data on how AI describes you, that test stops short of giving you the information you really need to know—whether you show up for potential clients who don’t even know you exist yet. This requires a different test, and it’s important if you want to make sure your firm doesn’t become digitally invisible in the next two years.
By July 2025, ChatGPT users were sending about 18 billion messages a week, more than five times the volume of a year earlier. About a quarter of that usage was people seeking information, up from 14%. They can get answers about everything (including your firm and the problems you solve) without ever going to your website. This is how people search the internet now. It’s your firm’s job to figure out how to get recommended by those AI platforms.
Branded Prompts and Non-branded Prompts Serve Different Purposes
To get the full picture of your firm’s AI visibility, track both branded prompts and non-branded prompts.
Branded prompts include your specific brand in the question. An example is, “What’s Scribewise’s approach to thought leadership?”
A non-branded prompt does not include brand names in the prompt. An example is, “Which firms help professional services companies publish smart thought leadership at scale?” These types of prompts are both important to your business, but they serve pretty different purposes.
Branded prompts offer insight into what AI platforms say about your firm. They show you what sort of messaging is being used and what the overall sentiment of your brand is in AI answers. Branded prompts can help you ensure your latest messaging and service offering are known and understood by AI platforms, which draw on sources they trust to build answers. Looking at branded prompts is what led us to realize that AI was getting our story wrong, and we were able to correct the misinformation. It can do the same for you. Think of it as a way to monitor your brand messaging and your brand health.
Non-branded prompts tell you whether your firm is being recommended when someone is trying to find a solution to a pain point. If a person doesn’t know your firm exists, will ChatGPT or Gemini recommend you? Non-branded prompts can help you gauge how well your service offering shows up and how clearly you state your value proposition. These prompts also show you which competitors AI compares you to and may surface some firms you aren’t aware of.
How to Run a Prompt Test in an Afternoon
Gathering initial data by running some prompts yourself is an excellent way to start, and can help you start to plan how you’ll tackle AI visibility.
- Write prompts: Write 10–15 prompts divided into four categories:
- Problems your firm solves (non-branded; these show whether your content shapes the answer)
- The industry category your firm serves (“best firm for X in the industry,” non-branded)
- A branded comparison between your firm and a competitor (“what’s the difference between firm X and firm Y,” branded) and a plain branded question (“tell me about firm X”)
- A shortlist request (“who should we hire for X,” non-branded; these show whether you’re named)
- Use prospects’ language: Rather than using your service page language, comb through emails, sales call transcripts and other prospect outreach to write prompts that more accurately match how someone would ask about problems you solve and firms in your industry.
- Run the prompts: Ask ChatGPT, Claude, Gemini and Perplexity all of your prompts, then search the same questions in Google to see whether an AI Overview appears (not every search triggers one). Sign out of your account or use incognito or a temporary chat, and start a fresh thread each time to get the best data. Some platforms limit signed-out use. Make sure web search is turned on in each tool, or you won’t see citations. Run your top prompts a couple of times since answers can differ, and log each result in a spreadsheet.
- Review the results: Check whether your firm was named. If you were named, is the information accurate? Look at the other firms named. And look for the sources behind all of those answers. Which websites is ChatGPT using as a source when it makes recommendations or answers questions?
What Your Prompt Run Results Mean
There are four possible results from your prompt run, and they all require different follow-up action.
If your firm is named in a non-branded response and all the information is accurate, congratulations. This is the best-case scenario, and your job is to protect those results by writing and publishing new content, updating content when necessary and promoting your firm through media outreach.
If your firm is named in a branded response but the information isn’t quite right, your next step is to review citations to see where the information is coming from and make adjustments. If your site is in the citations, it’s a straightforward fix to update or adjust content on your own site. If inaccurate information is published elsewhere on the internet, start by correcting any profiles, directories or bios you control. Beyond that, media outreach may help you adjust the narrative about your firm across the internet.
If competitors are named but you are not, review the citations behind those answers. Next, create a short list of third-party sites and trade publications that come up. You can create content from your own point of view to compete with the content AI already cites, or pursue media relations to get your point of view published in trade media.
If no one’s cited or named, try rephrasing the prompt and running it again. If the results hold, the category is wide open. This is a topic ripe for thought leadership and content. Plan your content calendar accordingly.
Where This Test Falls Short
While your one-afternoon test is a good indicator of where you stand and how you should plot out generative engine optimization (GEO), it’s not sustainable, and it isn’t the full picture. AI search is probabilistic, which means when you ask a question repeatedly, you’ll often get different answers. Brands need to track their share of voice over time across more than the 10–15 prompts you tested, and ideally across several platforms. Share of voice is how often your firm shows up in answers compared with competitors.
The best way to make sense of what you’re seeing and develop a strategy for GEO is to monitor AI visibility using a tool and come up with a plan to update and create content, backed by prompt development, a review of the citations behind the answers, and content and media relations work.
Asking ChatGPT the question your best client would ask, without your firm’s name, is a good foundation. It just isn’t a long-term plan for understanding AI search’s impact on your firm. If the answers surprise you, reach out.