We wanted to see what those answers actually look like at scale. So we ran a structured test. Twenty questions a founder would ask, put to all three engines, four separate times each. Two hundred and forty answers in total. Seven of the twenty questions named a brand, so the honest read comes from the thirteen that did not: open questions like which platform is best for a small store, where 52 answers per engine decided who got named and who got picked.
This article uses Shopify as the test subject, because it is the biggest name in the category. The full sample report is free further down this page, behind a one-field email form. Here is what the answers showed.
Shopify was named in every single open answer
On all three engines, Shopify appeared in 100 percent of the unbranded answers. If a founder asked an AI which platform to use, Shopify was in the room every time.
Being named is not the same as being picked, though. Answers name platforms to dismiss them, to price them, and to compare against. So we also coded whether each answer actually put the platform forward as a pick for the person asking. On that measure Shopify was shortlisted in 92 percent of ChatGPT answers, 87 percent of Gemini answers, and 92 percent of Claude answers. One question, the monthly cost question, produced price tables and zero recommendations for any brand on any engine, so 92 percent is effectively the ceiling on this panel.
If you sell software, that gap between named and picked is the number that matters. Buyers rarely read the full answer. They take the shortlist.
The engines do not agree with each other
The most useful finding was not about Shopify. It was that the three engines behave like three different reviewers.
Claude is generous. It shortlisted Shopify in 92 percent of open answers, but it also shortlisted WooCommerce in 83 percent and BigCommerce in 81 percent. Shopify's share of all platform mentions on Claude was 22 percent, exactly tied with WooCommerce. On ChatGPT and Gemini, Shopify held the largest mention share outright, 30 percent and 27 percent. A business that checks only one engine is seeing one reviewer's taste and calling it the market.
Gemini is decisive, and occasionally blunt. It told dropshippers to avoid one tracked platform entirely and described another as something stores outgrow in months. It also produced the only negative answer about Shopify in all 240 responses, which brings us to the next point.
One question where Shopify nearly disappeared
Asked which platform is best for SEO and content-led growth, Gemini shortlisted Shopify in only one of four runs, while shortlisting WooCommerce and BigCommerce in all four. Its reasoning, repeated across runs, was that Shopify's blog and URL structure are weak for content-led growth.
ChatGPT and Claude answered the same question differently. Both shortlisted Shopify and WooCommerce side by side and split the verdict: WooCommerce if content is the product, Shopify if selling is.
One question is a signal, not a pattern, and our audit rules treat it that way. But it shows the shape of the risk. A platform can be the default answer almost everywhere and still lose a specific buying question on a specific engine, for a stated reason a business could actually address.
The question where nobody gets recommended
The monthly cost question is worth its own paragraph. Asked what the platforms cost, all three engines stopped recommending entirely and produced price tables instead. No brand was picked in any of the twelve answers.
If your buyers ask AI assistants about price, and they do, the assistant is not selling for you or against you in that moment. It is laying out numbers. What those numbers say, and whether your pricing page gives the engines clean figures to work with, is the whole game on that question.
What this means if you sell anything at all
Shopify is the strongest brand in its category, and even it has a question where one engine nearly drops it and a rival it ties with on another engine. Most businesses are not Shopify. They have real gaps: categories where they are never named, engines where a competitor is the default, questions where the answer gives buyers a reason to pick someone else.
You cannot fix what you have not read. The answers your buyers are getting exist right now, word for word, and they can be collected and counted.
How we ran this, and the limits
Twenty prompts, three engines (ChatGPT, Claude, Gemini via their APIs), four runs each, 240 answers. Thirteen prompts were unbranded and carry the headline figures. Recommendations were coded by a written rubric: a brand counts as shortlisted when the answer presents it as a pick for the asker's situation, including conditional picks. Price-table listings and avoid-framings do not count.
Two limits, stated plainly. First, API answers are not identical to what a buyer sees in the consumer apps; they are the closest controlled way to run the same question many times, and every report we write says this. Second, 52 unbranded answers per engine is a structured sample, not the universe. Treat the rates as directional. The gaps we call real passed significance tests; the ones that did not, we report as leads, not proof.
See the full sample, or check your own business
The complete Shopify sample report, with the scorecard, per-engine breakdowns, and the action plan format our clients get, is free. We ask for an email so we can send it, nothing more.
And if you want to know what the AIs say about your business, the free check runs five buyer questions for your category and sends you the exact answers by personal email within 24 hours. The full audit runs 40 questions across all three engines, maps where the gaps are, and recommends actions for each one.