I Audited My Own Business. AI Kept Introducing Me As Someone Else

I've just rebuilt my site at reddogstrategies.co to reflect what I offer now. It went live late last week. Shameless plug -- please go check it out and tell me what you think. And if you have an owner-operated business, check out the Free AI Visibility Score tool I offer. Before the rebuild, I ran the audit on myself. Apparently 25 years of helping eBay, HP and LinkedIn tell their stories did not exempt me from checking whether my own was being told correctly. Fair enough.
October 26, 2026
Hani Durzy

On October 18, I asked Google, "What is Red Dog Strategies?" Its AI Overview opened with a political consulting firm in Sacramento. My business showed up as a parenthetical footnote.

That political firm is Red Dog Inc. There's nothing wrong with it appearing in an answer about that firm. There is something wrong with it leading an answer about mine.

Mine is Red Dog Strategies LLC. I help owner-led businesses outside tech with their marketing, their content and what AI says about them. That made the result a little awkward -- I sell audits that uncover exactly this sort of thing.

I've just rebuilt my site at reddogstrategies.co to reflect what I offer now. It went live late last week. Shameless plug -- please go check it out and tell me what you think. And if you have an owner-operated business, check out the Free AI Visibility Score tool I offer. Before the rebuild, I ran the audit on myself. Apparently 25 years of helping eBay, HP and LinkedIn tell their stories did not exempt me from checking whether my own was being told correctly. Fair enough.

The short version: the machines could find my name. They had a much harder time finding me.

This isn't a case study about how I fixed it. I haven't earned that one yet. It's the before-picture -- including the parts I'd rather not put in a launch announcement.

Found by name. Described as someone else.

I took an initial reading on September 22. On October 18, I ran a larger test using the same 18 questions, with three runs each across ChatGPT, Perplexity and Claude: 162 answers. Those were collected through the companies' software interfaces, not their consumer chat apps; I checked Google's AI Overviews separately in a browser. That distinction matters. This is a defined test, not a claim about every answer every customer gets.

The questions did different jobs. Some asked for a business like mine without supplying my name. Others asked directly about Red Dog or me. A few tested comparisons. Mixing those together would give me a number, but not a useful one.

On the discovery questions, sometimes called "unbranded" -- the ones that didn't name me -- Red Dog appeared in zero of 63 answers. That's a finding about these questions and these tools on that date. It isn't proof that nobody can find me anywhere.

When the question did name the firm, the tools had no trouble producing an answer. The problem was which firm they described. Fifteen of 36 answers led with the Sacramento political business: nine from Claude, four from Perplexity and two from ChatGPT.

Then there was the description of my actual work. Across a separate set of 63 answers to questions naming Red Dog or me, none described the current positioning around AI visibility and owner-led businesses. Thirty-two still used "fractional," the language of the work I was beginning to move away from.

Being named. Being described correctly. Being recommended. Three different things, and I'd been treating an updated description as though it would take care of all three.

The correction was there. It didn't make the answer.

One result was more instructive than the zeros.

I'd updated my LinkedIn headline to describe the work I do now. In 12 of 21 Perplexity answers to questions naming Red Dog or me, the source material included that updated headline. Not one of the 21 answer texts used the new positioning.

The tool could find the correction. It didn't carry it into the answer.

I can't tell you why from that test alone. It doesn't show how Perplexity weighed each source, or prove that a stack of old descriptions outvoted the new one. It shows something simpler, and useful enough: finding your updated information isn't the same as using it. Checking whether your page was cited doesn't finish the job. You have to read what the answer actually says.

And I can't pin this entirely on the tools. My old website was still live. My newer LinkedIn headline and older descriptions were telling different stories. Asking a machine to resolve that contradiction isn't much of a communications plan.

Why this belongs in a website announcement

The new Red Dog site went live last week. It describes the work I do now: find out what AI is saying about a business, fix the information and content behind it, and build the systems that keep that work going. Reputation and crisis readiness belong in that same picture, because the answer that introduces a business can also misrepresent it.

A private club, a wealth management firm, a dental group...none of them need to share my name problem to share the business problem. Is AI recommending the right location? Describing a service the practice still offers? Introducing the business through something an owner thought had been corrected? Those are questions to investigate, not findings from my self-test. But the distinction matters for any owner: the same answer can be your introduction to a new customer and the customer's first impression of whether to trust you.

I explained the move toward owner-led businesses last month. This is what checking my own business uncovered. And as the Los Gatos dental test showed, being recommended is a different question again. I'm not going to pretend a new homepage answers it.

The rebuild gives me a clearer account of the business on a site I control. Whether the tools adopt that account, separate me from the other firm, or start recommending me is what the next measurements have to establish. None of those outcomes is guaranteed by a launch date.

What I'll report next

I began putting this plan into practice last month. Early next year, three to six months into that work, I'll report what changed. It doesn't happen overnight.

I'll use similar measures, chosen before seeing any after-results: whether Red Dog gets named when the question doesn't supply my name; whether answers about the firm lead with the right business; whether they describe the work I do now; and whether Google's AI Overview identifies the right firm when someone asks about Red Dog Strategies.

If the description improves and discovery doesn't, I'll report both. I'll give the collection dates, too. An answer collected in December doesn't become a six-month result because I publish it later.

Other work will be happening in that window, including the San Jose Dental Index, planned for mid-November release, with accuracy checks covering 46 practices. That work may affect what the machines find about Red Dog. This won't be a clean experiment that isolates the website. It will be a documented account of what changed while I worked on the business.

If you own a business, start by asking an AI tool what it says about you. Save the answer, the question and the date. Then check the facts against something you can verify. One answer isn't a verdict, but a wrong one is a reason to look harder.

That's what I should have done sooner. Not another announcement about what I wanted the business to be. A check on what was already being said about it.

If you'd like a starting point for your own business, request the free Red Dog AI Visibility Score. It's a smaller snapshot check, delivered within one business day -- not this full baseline exercise and not the paid audit.

Method and limits

The October 18, 2026 test used 18 fixed prompts, three runs per prompt, across ChatGPT with web search, Perplexity and Claude through their APIs (software interfaces): 162 answers. Google AI Overviews were checked separately in a browser on three prompts. The September 22 reading used the same 18 prompts once each on ChatGPT and Perplexity: 36 answers, plus separate browser checks. October's denominators are used above. The discovery and branded sets each contain 63 answers but are separate sets. The firm-identity subset contains 36 answers; the Perplexity branded subset contains 21.

The October repeated runs agreed on the recorded outcomes; that doesn't establish stability on another day. These tests do not establish revenue loss, customer behavior or the causes of the errors. They are not the engine lineup or scope of today's free Score or paid audit. The follow-up will preserve the locked questions and collection method; website, LinkedIn, publishing and third-party changes will be logged separately. Red Dog Strategies LLC is not affiliated with Red Dog Inc. of Sacramento. The questions and underlying responses are available on request.

‍

‍

case study
visibility check
1400
One piece a week. no sponsors, no affiliate links, no recycled SEO advice.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
More Like this
No items found.
Next step
Let’s find out what AI says about you.
Thirty minutes, no deck. I'll run the questions your customers ask while you watch, show you what comes back, and tell you honestly whether anything is worth fixing. Sometimes the answer is "not yet."