Which buyer question should we test first to see whether AI recommends us?

Why it matters.

Buyers already use AI while they buy. Forrester’s State of Business Buying 2026, published on 21 January 2026, reports that 94% of business buyers use AI during their buying process. So the question that matters is not whether AI mentions your company somewhere, but whether it names you when a buyer asks it for suppliers.

A brand question (“What is Acme?”) and a commercial question (“Which firms should a mid-sized logistics company consider for warehouse software in the UK?”) test different things. The first checks whether the engine knows you. The second checks whether it shortlists you. Most companies test the first, feel reassured and stop.

The 10-minute self-check.

Write one sentence a buyer would type when they need a supplier like you. Then check it against these four points:

• It names what is being bought, in the buyer’s words, not your product name.
• It names the buyer: company size, industry or role.
• It names the market or region you actually serve.
• It asks for options (“which firms”, “who should we consider”), not a definition.

If it fails two or more, rewrite it before you test anything.

Steps.

1. Ask two salespeople how recent prospects described their problem on the first call. Write down their exact phrases.
2. Draft three versions of the commercial question from those phrases. Keep the one a buyer would most plausibly type.
3. Add the qualifiers that change the answer: market, company size, industry. Remove your own jargon.
4. Save the final wording in a shared document with today’s date, so every later test uses the same words.
5. Run it once on ChatGPT, Perplexity and Google’s AI. Record every company named and every source shown. This is your baseline, not a verdict.

If you want a second view of the wording, our AI search visibility services start from the same commercial question.

What we saw on prolinkage.com.

How to measure it.

• Named rate: runs that name you, divided by total runs, per engine. Check monthly.
• Position is not a metric: record it, but judge progress on the named rate across runs (see play 03).
• Wording log: the exact question and the date it was fixed, so later results compare like with like.

Related plays.

• Play 02: How many buyer questions do we need to test AI visibility properly? (Measure first)
• Play 03: Why does ChatGPT give a different answer every time we ask? (Measure first)

Sources.

• Forrester, The State of Business Buying, 2026, 21 January 2026.
• Prolinkage AI self-test (ChatGPT, Perplexity, Google AI Overview, one run each), 6 November 2026.