About the Author
Corey is the owner and President/CEO of VOLTAGE. He is also founder and author of The Digital Marketing Success Plan® and the START Planning Process. Corey has spent 20+ years working in strategic and leadership roles focused on growing national and local client brands with award-winning, ROI-generating digital strategies. He's the recipient of the KCDMA 2019 Marketer of the Year award and his team at VOLTAGE has won nearly 100 local, national, and global awards for ROI-focused client work in the past decade.
Ask three AI tools who to hire for the same project and you will get three different shortlists, three descriptions your firm did not write, and uneven credit for the sources behind them. I know because on September 28 I ran one prompt through ChatGPT, Perplexity, and Google AI Mode (logged out, so no personal history shaped the results–and I didn’t include Claude since it is my go-to and will bias everything based on me) and compared the answers side by side.
The prompt was the kind a hospital executive, project stakeholder, or someone else involved in the very early stages might type: “Which architecture and engineering firms should I consider for a hospital expansion project in Columbus, Ohio? What makes each one a good fit?”
I shared this experiment on October 2 at the SMPS Heartland Regional Conference, in a session I co-presented with Jerry Gennaria of TOKY for architecture, engineering, and construction (A/E/C) marketing and business development leaders. The pattern applies to any B2B professional services firm where a buyer asks an AI tool which firms to consider for a specific project in a specific place, whether that is a healthcare consulting engagement, a complex commercial dispute, or a multi-state tax matter.
This article explores what came back and how to run the same test on your own firm. It was one prompt, one day, and one answer per tool, so I treat the results as observations worth testing further. I also kept every firm anonymous because I’m not comfortable naming real firms in an experiment like this.
Different AI Tools Recommend Different Firms
ChatGPT recommended seven firms and Perplexity recommended five. Only one firm appeared in all three tools, and two more showed up in both ChatGPT and Google AI Mode but never in Perplexity.
A buyer who opens one tool walks away with a noticeably different list than a buyer who opens another. A firm that looks well represented in Google AI Mode can be missing from Perplexity, so checking a single tool tells you very little about how the market sees you.
AI Tools Write Their Own Descriptions of Your Firm
Each tool wrote its own explanation of why each firm fits, and the firms had no hand in that wording. It may not match how they would position themselves.
What stood out to me was how familiar the language sounded. Phrases like “deep local experience,” “integrated design and engineering,” and “evidence-based design” appeared in the answers, and in my experience many professional services firms use similar words on their own websites (you can swap these examples out for your specific industry). Advertising and marketing agencies are equally guilty (“full service”, “integrated”, etc) in this. When the claims are that common, a model has little to separate one firm from another. Jerry’s half of our session covered this problem, and his fix is a Brand Truth, a point of view specific enough that a competitor couldn’t credibly claim it.
AI Tools Credit Sources Unevenly
ChatGPT’s only visible citation was the hospital system’s own newsroom. Its sources list, behind a click, included project pages from just three of the seven firms it recommended, so four firms were named without a page of their own in the sources. Perplexity’s top visible sources were directories, including Procore and AIA Ohio. Google AI Mode showed inline source chips citing a news story and one firm’s project page.
Comscore data reported by Search Engine Land shows the same dynamic at a larger scale. Tripadvisor was a source in 61% of lodging-related AI answers but a visible cited link in only 21% (U.S. PC activity, December 2025 through May 2026). That is a travel example and I haven’t seen an A/E/C equivalent yet, but the mechanism carries over to firms whose project pages inform an answer without earning a mention.
Run the Same Test on Your Own Firm This Week
Write the question a real buyer would type, including your service, the project type, and the location. Run it in two or three tools while logged out, and record which firms appear, how each is described, and which sources are cited.
Checking more than one tool matters because the market is shifting. In Comscore’s U.S. desktop data, ChatGPT’s share of AI assistant prompts fell from 70% to 50% between January and June 2026, while Gemini grew from 17% to 30% and Claude from 2% to 11%.
Then look at the results with a few questions. Do the descriptions sound like us? Which sources did the tools lean on? Are we present on them?
Give AI Tools Something Specific to Say About You
The most useful material is the kind only your firm has. That includes original research (proprietary data becomes the thing others cite), named frameworks (a method with a name becomes a reference point), a point of view your experience backs up, and first-hand proof such as project pages that describe real outcomes in plain language. From there, earn mentions on the third-party sources these tools trust, which in this experiment included directories, trade associations, and local news.
Our AI Search Field Guide and one-page Content Audit Checklist walk through the structure in more detail, and if you work in A/E/C, our built environment page shows how we approach it.
Frequently Asked Questions
How do I check what AI says about my firm?
Run a prompt a real buyer would type in ChatGPT, Claude, Perplexity, and/or Google AI Mode while logged out, then record the firms named, how each is described, and the sources cited.
Why do AI tools recommend different firms?
Each tool draws on different sources and generates its own summary, so shortlists vary. In my test only one firm appeared in all three tools.
Does this apply outside architecture and engineering?
Yes. The pattern applies to any B2B professional services firm where buyers ask AI tools which firms to consider for a project.
Wrap Up
Before a human evaluates your firm, a machine often already has. Rerunning this test with a different city, project type, or month is an easy habit to build into your reporting, and it shows you whether the work you put into content is changing what the tools say.