Why we look at buyers first
We started CheckMarque from both sides of the same table. I sell technical products, and I watched a buyer challenge my team using an AI's summary of our company. My co-founder, Avi Sancheti, works in procurement, and he watched his own team shortlist suppliers on what a model said about them before a single vendor conversation had started.
Neither moment was about technology. Both were about how people now decide. Buyer behaviour is the first of the three lenses we read everything through, and the one most businesses skip.
What has changed in how people buy
They ask before they can name it. A buyer used to need the right term before they could search. Now they describe the situation in their own words, and the assistant supplies the vocabulary, the rules and the options.
The shortlist arrives ready-made. In G2's March 2026 survey of 1,076 B2B software buyers, 54% ranked AI chatbots as the biggest influence on which vendors made their shortlist. 69% said a chatbot led them to a different vendor than they had planned, and a third chose one they had not been considering.
They check the names they were given, not the ones they were not. Trust is low. In Alchemer's August 2026 survey of 1,002 US shoppers, 48.5% had used AI to research a purchase and only 35.4% trusted its recommendations mostly or completely. Gartner found 69% of B2B buyers prefer to validate AI insights with a sales rep. Yet when an AI answer leaves a brand out, G2 found only 24% go looking for peer feedback, while 21% go with the AI's pick and 14% fall back on a brand they already know.
More than one person asks. Forrester's State of Business Buying 2026 counts an average of 13 internal and 9 external people in a B2B purchase, and about twice as many where generative AI features are involved. At home, a couple compares answers from two different apps.
They start to hand the task over. Google's information agents hold a buyer's intent and watch for a match. In retail, Google's Universal Commerce Protocol already lets agents in AI Mode and the Gemini app fill a cart and read live prices and stock.
What our study shows about buyers
In our study of 240 AI conversations about money, each simulated buyer had fixed details, asked a question in their own words, added their situation, and then asked either "who should I go to?" or only "what's my first step?". Three things stood out from the buyer's side.
The buyer who never asks "who?" meets very different markets. Asking only for a first step, they were shown a firm in 1 of 24 conversations on ChatGPT, 6 on Claude and 16 on Gemini.
Conversations out of 24 per engine
Language decides the market. Only 11% to 19% of brands appeared in both Dutch and English conversations.
The first number sticks. For one expat couple, the same engine put their maximum mortgage anywhere from €400,000 to €550,000 across ten runs. Their €475,000 flat was in reach in one conversation and out of reach in the next.
Six things that shape what a buyer hears
We call the context of any question a buyer moment. It has six parts, and changing any one of them can change the answer.
A buyer who is a decision-maker, just looking, asks ChatGPT in Dutch, in the Netherlands with a short question at the opening question.
Illustrative. Click a part to change it.
- Who is asking: a decision-maker, a specialist, a first-time buyer, a colleague in procurement.
- How close they are to deciding.
- Which engine they open.
- Which market and language they ask in.
- How they ask: a short question, a long description, a photo, a document.
- When in the conversation: the opening question, a follow-up, or the turn where they ask which one to choose.
Most businesses test one of these, if any. Their buyers vary all six.
The lens applied: what you can do yourself
Each row is a finding from the study, read as buyer behaviour, with something you can do this week. The last column names the gate and state in our framework: Found (does AI name you?), Understood (does it get you right?) and Actionable (where does it send a buyer who is ready?). The overview explains each state.
| Finding | What it means for buyers | What you can do yourself | Gate · state |
|---|---|---|---|
| Published prices get firms named | "What will it cost?" comes just before the choice | Ask your sales team which question buyers ask right before deciding, and answer it on your site | Understood · Undersold: the fact the buyer needs is missing |
| Two languages, two markets | Each language is a different customer with a different question | Write each market's question in that market's own words, not as a translation | Found · Overlooked: not named in that market |
| Figures move between runs | Buyers anchor on the first number they see | Add "what did an AI tell you?" to your intake form or first call | Actionable · arrives misinformed |
| Some engines wait, some volunteer | Most buyers ask how to start, not who to call | Test with "what's my first step?", not only "who should I go to?" | Found, and for buyers who never ask, not considered |
| Few brands appear on every engine | Partners and colleagues compare answers from different apps | Ask the same question on two engines as two members of the buying team, and compare | Across gates: how stable your position is |
| Engines send buyers to check different things | Buyers verify on registers, reviews and public sources | Find out where your buyers check, and make sure you are there and correct | Found · Unsourced |
| The textbook answer wins | Buyers accept the default, especially when it saves money | Ask about the problem, not your category, and see whether your kind of solution comes up | Found · Overlooked, framed out |
| Specialists win specific questions | The more specific the situation, the more a buyer wants a specialist | Collect the situations your best customers were in; each one is a question to answer | Understood · Out-positioned |
| Each engine shows different proof | The sources a buyer sees depend on the app | Ask recent customers which sites they read before deciding | Actionable · via aggregator |
| Pages used without the firm's name | Buyers learn from you without learning who you are | Ask a question one of your pages answers, and check whether your name appears | Found · Ghost-cited |
Try it yourself: a 20-minute buyer test
- Write down how your buyer would describe their problem without knowing your name. Use their words, not yours.
- Ask it on ChatGPT, Gemini and Claude. Then add their details, the way a real buyer would.
- Ask "what's my first step?" in one conversation and "who should I go to?" in another.
- Repeat in each language you sell in.
- Note which names come up, in which order, and what the buyer is told to check.
- Ask your next three new customers what an AI told them before they contacted you.
You will learn more about your buyer journey from that than from most dashboards.
Why one lens is not enough
Buyer behaviour tells you which questions matter. It does not tell you why an engine answers them the way it does, or what it says about you when it does. For that, read the other two lenses: how AI engines decide which companies to name and the version of your company AI tells. For all three applied to one industry, see One study, three lenses.
Frequently asked questions
How do buyers use AI to research purchases? They describe their situation, get an explanation of the options, ask for names and prices, and check the answer, often inside one conversation. In G2's 2026 survey, 51% of B2B software buyers start their research with an AI chatbot more often than with Google.
Do buyers trust AI recommendations? Less than they use them. About a third of US shoppers trust AI recommendations mostly or completely, yet when AI leaves a brand out, most buyers do not go looking for it.
Does this apply to B2B as well as consumers? Yes. The steps are the same. B2B purchases involve more people, and each of them may ask a different engine.
What is a buyer moment? The context of a question: who asks, how close they are to deciding, which engine, which market and language, how they ask and when in the conversation. Change any one part and the answer can change.
Where do your buyers stand?
If you would like to know what your buyers are told before they call you, we are happy to talk it through. In a 20-minute call we look at how a potential customer would describe the problem you solve, what ChatGPT, Gemini and Claude say when they do, and whether optimising for the AI buyer journey is worth your attention right now. Sometimes the answer is that it can wait, and we will say so.
Sources
- CheckMarque AI, "Three AI engines, eight Dutch money questions," 10 October 2026.
- G2, "The Answer Economy: G2's 2026 AI Search Insight Report," March 2026, n=1,076 B2B software buyers. https://learn.g2.com/g2-2026-ai-search-insight-report
- Alchemer, "Half of shoppers use AI to shop. Only a third trust what it tells them," August 2026, n=1,002 US shoppers. https://www.alchemer.com/content/industry/retail/ai-shopping-trust-gap-2026/
- Gartner B2B buyer survey, May 2026, via Demand Gen Report. https://www.demandgenreport.com/industry-news/news-brief/gartner-ai-is-reshaping-b2b-buying-but-human-sellers-still-close-the-confidence-gap/53046/
- Forrester, The State of Business Buying, 2026, via BIIA. https://www.biia.com/forrester-b2b-buying-groups-expand-as-they-question-ai/
- Google, "Universal Commerce Protocol updates improve AI shopping for retailers," 19 March 2026. https://blog.google/products-and-platforms/products/shopping/ucp-updates/
Raghav Shankar
Co-founder, CheckMarque AIRaghav Shankar is co-founder of CheckMarque AI. By day he leads European commercial operations at Krayden Europe (formerly CAPLINQ), an advanced-materials company selling into semiconductor, automotive and electronics manufacturing, where he also builds the AI agents his team works with. He holds a BTech in Computer Science and an MSc in Management of Technology from TU Delft.
How AI engines decide which companies to name
ChatGPT, Gemini and Claude name companies from memory and from search, and each mixes the two differently. What our study of 240 AI conversations shows about how the engines behave, and the technical checks you can make yourself.
Read →Analysis · · 12 minOne study, three lenses: what AI tells people about money in the Netherlands
We read 240 AI conversations about Dutch mortgages, tax, pensions and investing three ways: as buyer behaviour, as technology and as brand. Ten findings, what each means for financial firms, and five things they can do this month.
Read →