We were named in none of 25 AI answers about our own field. That is the honest starting point for a young consultancy, and it is why we publish the baseline instead of a claim. The more useful finding is how little the answers agree with each other: one answer from one tool is a snapshot, not a ranking.
The short answer
Zero mentions in 25 answers. Sixteen of those answers were to the six questions closest to what we sell: SEO before a relaunch, white-label SEO, its price, SEO for a hair transplant clinic, Framer versus WordPress, and who checks AI visibility. GetBorvis appeared in none of them. We did not expect otherwise, and we would rather show the zero than a cherry-picked screenshot.
What we asked, and how
Nine fixed questions, three tools, one fresh conversation each, on 16 October 2026. Six questions were about our field. Three were controls from other industries, so we could see how the tools behave when the answer space is crowded with well-known names. Each question was typed exactly, in a logged-out or private session, with the tool's default mode, and the full answer and its links were saved with the time.
| Question | What came back | Named GetBorvis |
|---|---|---|
| Who can help a web design agency with SEO before a website relaunch? | Described roles, not providers, in all three | 0 of 3 |
| White-label SEO partner for a small web design agency | A shortlist in all three: 15 agencies | 0 of 3 |
| How much does a white-label SEO service cost for agencies? | Price ranges, no shortlist | 0 of 3 |
| SEO consultant for a hair transplant clinic | One shortlist, one general advice, one seen but not saved | 0 of 2 |
| Is Framer or WordPress better for SEO? | Platform comparison | 0 of 3 |
| Who checks whether a business shows up in ChatGPT answers? | Read as a question about OpenAI, or a process; one gave no answer | 0 of 2 |
| Best project management software for a small agency | A shortlist in all three: 8 products | Control |
| Best Italian restaurant in Manchester for a birthday dinner | A shortlist in all three: 8 restaurants | Control |
| Best international school in Dubai for a 10-year-old | A shortlist in all three | Control |
36 combinations were planned, including Claude. Claude asked us to sign in, so its nine were not run and are not counted as misses. One Perplexity question returned no answer, and one Gemini answer was seen but not saved, so both are left out. That leaves 25 completed answers. Network location was not controlled, and the tools did not show exact model versions.
Shortlists do not repeat
Asked for a white-label SEO partner, the three tools filled 16 shortlist places with 15 different agencies. No agency appeared in all three answers. One appeared in two. Of the three possible pairs of tools, two shared no agency at all.
The controls behaved differently. For project management software, the three answers named 8 different products, and three products appeared in all three. For an Italian birthday dinner in Manchester, they named 8 restaurants, and none appeared in all three. A crowded, well-known category produces more agreement than a fragmented one, though one answer each cannot tell us why.
If three tools asked the same question on the same evening agree this little, a single screenshot of being "recommended by ChatGPT" says less than it seems. A fair measurement asks the same questions on fixed dates, several times, and records what changes. That is the method in our prompt-set guide, and this is its first run on ourselves.
Many questions get no shortlist
Asked who can help a design agency with SEO before a relaunch, none of the three tools named a provider. All three described the roles instead: technical SEO, migration specialists, content and tracking, with a list of tasks such as redirect mapping and staging checks. Perplexity cited ten agency pages as sources while recommending none of the agencies.
That matters for how agencies think about AI visibility. For many real questions, the answer is advice, and the opportunity is being the source that advice draws on, not a name on a list. Our redesign checklist and redesign handoff exist for exactly that question.
What the tools said about price
The three answers to "how much does white-label SEO cost?" gave ranges, not names: roughly $300 to $3,000 per client a month, $300 to $1,500, and $300 to $900 for offshore local SEO. These are the tools' estimates, not quotes. When we read 13 real white-label offers ourselves, the six that published a price started between $178 and $1,299 a month. What those 13 offers actually sell is written up separately.
The Framer versus WordPress answers were similar in spirit: WordPress for flexibility, Framer for a simpler marketing site, and neither drew on measurements of live sites. We measured 133 live sites for our platform comparison, and the differences we found were about defaults, not rankings.
Where the answers got their sources
Across the 25 answers we saved 69 links: 48 cited as sources and 21 linking straight to a business. In the birthday-dinner answer, two of the five sources were restaurant websites and three were booking or venue platform pages. A business can be recommended on the strength of someone else's page about it, which is why we look at your profiles and listings as well as your website.
What it means for us
We are new, and the tools do not know us yet. We will not buy placements in "best agency" lists, and we will not write pages that pretend to be neutral reviews of ourselves. What we can do is publish evidence nobody else has, such as our studies of 81 design agency websites and AI rules on 474 websites in six industries, and be the source that answers draw on.
We will rerun the same nine questions on fixed dates, word for word, and publish the results whether or not they move. If we add market-specific questions for the UK, the US or Australia, they will be new questions with their own baseline, not a quiet replacement of these.
What it means for you
- Do not judge AI visibility from one answer. Ask a fixed set of real buyer questions, on fixed dates, several times.
- Separate three things: being named, being cited as a source, and being visited. They move independently.
- Check what the answers cite. If it is a directory or a booking platform, that page is part of your visibility too.
- Be wary of anyone who promises AI recommendations. Nothing in these 25 answers suggests that can be guaranteed.
Limits
- One answer per tool per question. This is a snapshot, not a rate of recommendation or a ranking.
- Logged-out sessions from one location. Personalised, signed-in answers may differ.
- Claude was not tested. Its nine answers are missing, not negative.
- We name the AI tools, not the businesses they recommended, because this is a test of the tools, not a review of anyone they named.
Our own recorded answers from ChatGPT, Perplexity and Gemini, 16 October 2026: 36 planned, 25 completed
Google: AI features and your website
OpenAI: crawler documentation