SEOritmo study – October 2026

Ask twice: what ChatGPT tells B2B buyers in regulated industries

We asked ChatGPT 360 B2B buying questions on 1,800 runs, observing 10,000 vendors in B2B and regulated fields like e-signature, CRM, HR software, data privacy, KYC and more, from the US and Spain, in English and Spanish.

This study shows what changes from one answer to the next, and whose evidence the buyer sees.

Cover of Ask twice, the SEOritmo study on what ChatGPT tells B2B buyers in regulated industries

1 in 3

vendors in a regulated shortlist change when the identical question is asked again

78%

of citations point to the recommended vendors’ own pages, even when price is not mentioned

52%

of regulated mentions go to local vendors a US buyer never sees, when a Spanish buyer asks in Spanish

73%

of answers name a challenger brand when the buyer asks who publishes prices, against 18% otherwise

The findings

Five findings from 1,800 ChatGPT answers

  1. Ask twice, get a different list: the same vendor comes first in all five runs for only 14% of regulated questions.
  2. Regulated categories are an open field: the top ten vendors hold 57–64% of mentions in data privacy, whistleblowing and payroll, against 90% or more in mainstream software.
  3. The vendors vouch for themselves: vendors’ own pages make up 78% of citations on plain questions and 93% on price questions. Review platforms appear in about 1 answer in 20.
  4. The market rewrites the list: country reshapes the shortlist, and the language of the question pushes it further toward local vendors and local sources.
  5. Price transparency is the challengers’ way in: smaller vendors in crowded categories appear mainly when buyers ask who publishes prices.

Only in the full study

  • Two action plans by each industry: one for B2B brands, one for SEO and GEO agencies
  • The list of vendors and brands appearing from the prompts
  • The data and charts behind each finding, by category, market and language
  • Real ChatGPT answers taken apart, with the external and internal sources it showed the buyer
  • The questions we asked, the limits of the study and the full methodology

Free · 21 pages · PDF in English. We email it to you once you confirm your address.

Methodology

How the study was built

1,800 runs

360 prompts, each run five times: 10 software categories (5 of them regulated), 3 question types and 4 wordings per question, for a company of 50–200 employees.

3 markets and languages

The US in English, Spain in English and Spain in Spanish. Spanish prompts were written natively, not translated, and each request came from the country named in the prompt.

Logged-out ChatGPT

The consumer web version, collected on 1 October 2026. ChatGPT searched the web before answering 99.8% of the questions. 1,797 complete answers were analysed.

Authors: André Pitì, Founder and CEO, and Giacomo Casolo, Founding Engineer and Developer, at SEOritmo.

What the data says

Does ChatGPT give the same vendor shortlist if you ask twice?

Not in regulated categories. About one vendor in three changes when the identical question is asked again, and the same vendor comes first in all five runs for only 14% of regulated questions.

Which sources does ChatGPT cite when it recommends B2B vendors?

Mostly the vendors’ own pages: 78% of citations on plain buying questions and 93% on price questions. Review platforms appear in about one answer in 20.

Does the buyer’s country or language change the shortlist?

Yes. Country reshapes the list, and asking in Spanish pushes it further toward local vendors: 52% of regulated mentions go to local vendors a US buyer never sees.

How does a smaller vendor get onto a ChatGPT shortlist?

Publish prices. When buyers ask who publishes prices, 73% of answers name a challenger brand, against 18% otherwise.

How often does your brand make the shortlist?

One answer is a sample. SEOritmo runs your category’s buying questions repeatedly, per market and language, and shows how often your brand makes the list and whose pages the buyer sees.

Or get the free study first