¶01 Why asking ChatGPT once tells you very little
A single ChatGPT question is not a reliable test. Artifexa tests UK NetSuite and ERP partners with buyer-style prompts, asked fresh, repeated across 7 AI assistants (counting the two Duck.ai models separately), and scores whether the firm is named, linked and which rivals appear instead. This test is separate from the engines our plans track, which are listed on our engagements page.
Ask an AI assistant the same question twice and you can get two different answers. Ask a different assistant and you may get a third.
So if you asked once and saw your firm, that is good news, but it is not proof. If you asked once and saw a rival, that is worth a closer look, but it is not the full story either.
¶02 How we test it
We test the way a buyer would ask, not the way a marketer would.
Buyer-style prompts. Questions a finance lead or operations director might type when looking for a partner.
A fresh chat each time. So earlier questions do not steer the answer.
Repeated runs. Each prompt is asked more than once, because one answer can be a fluke.
7 assistants. 5 AI services (7 models), such as ChatGPT, Gemini, Perplexity and Google’s AI answers, because each one draws on different sources.
The aim is a result you can trust and repeat later, not a one-off screenshot.
¶03 What we score
Named. Does the answer mention your firm at all?
Cited. Does it link to your site as a source?
Rivals. Which other firms appear instead, and how often?
We also record barriers rather than guess around them. For example, some assistants need a sign-in before they answer, and we note that instead of filling the gap.
¶04 What a result looks like
You get a short table for each assistant: how often your firm was named, how often your site was linked, and which rivals appeared instead.
Every result carries the date and the prompts used, so the same test can be run again later and compared like for like.
¶05 What we do with the result
A score on its own does not win work. We use it to decide what to fix first.
Visibility Review. We set the AI results next to your site, search presence and proof. About the review.
Fixes to your existing site. Clearer service pages, proof buyers can check, and pages that are easy to read and quote.
Test again. The same prompts, so you can see what changed.
For why an assistant picks one firm over another, see why an AI answer names one firm and not another. For the wider search side, see SEO for NetSuite partners.
As an example of the delivery work, we rebuilt the website of a UK NetSuite consultancy. Read the Next Layer case study.
¶06 Want to see where you stand first?
Our playbook explains the method in more detail. Our free AI Visibility Checker reads your homepage and up to five key pages and runs our public checks. No email needed.
Or start with the Enquiry Leak Finder. It takes about three minutes and shows where enquiries may be slipping away.
Basis for this piece
Our own testing method and one anonymised review from our work. We do not name the firms involved, and we make no claim about how any AI company ranks firms.