How we write your buyer questions

The short answer: we read your website, work out what you sell and who buys it, and write the 12 questions a real UK buyer would ask an AI — eleven of which never mention your name. Then we ask ChatGPT, Claude, Gemini, Perplexity and Grok all twelve, live, and record who gets recommended.

Step 1 — We read your site, not a form

The audit starts from your page's actual readable content — the same extraction browsers use for reader mode — plus what our free technical scan detected about your sector and structured data. Nothing is templated: a Cardiff bridging-loan broker and a Manchester SaaS company produce completely different question sets.

Step 2 — Buyer intents, not keywords

Every business is matched to the intents its buyers actually have:

• Brand — someone typing your name. Exactly one question, as a baseline.
• Researching — buyers who don't know solutions exist yet, searching the problem in plain English.
• Comparing — “X vs Y”, “alternatives to”, weighing options.
• Ready to buy — “best X uk”, “cheapest X”, strong purchase signal.
• Local — geography-anchored, using the real places your site serves.
• Support — existing customers looking for help.

Step 3 — Written the way people actually type

Eleven of the twelve questions are brand-blind: phrased as if the searcher has never heard of you, because that's where being recommended actually matters. Every draft question is then scored for realism — would a human type this? — and anything that reads like SEO keyword soup gets rewritten before it's used. All questions are UK-localised.

Step 4 — You can steer, not script

The optional field at checkout — “what do you sell, and where?” — is treated as authoritative over your page copy. If your homepage is thin or generic, one sentence from you fixes the targeting. What you can't do is hand-pick the questions, and that's deliberate: an audit you can script isn't an audit.

Step 5 — Locked, so rescans are honest

The moment your 12 questions are written, they're frozen to your audit. Rescan after fixing things and we ask exactly the same questions — so when the ✕s turn to ✓s, it's because your visibility changed, not the test.

What we don't do

• No cherry-picked questions designed to make you look good (or bad).
• No training-data answers counted — if a model didn't actually search the web, that answer is excluded from your score.
• No reshuffled rescans — same questions, every run.

See it on your own site: the free check takes about a minute, and the full citation audit is £19 — 12 questions × 5 AI models, live while you watch.