AI answers
The same 30 questions put to six AI engines, latest check 1 October 2026 and previous 1 September 2026. The audit is re-run monthly.
Questions where at least one engine named or cited Harbour & Key, 20 of 30.
How this is measured, and the engine-by-engine counts
| Engine | Named | Cited | Checked |
|---|---|---|---|
| Gemini | 5 → 6 | 8 → 16up | twice |
| Perplexity | 1 → 2up | 7 → 10up | twice |
| Google AI Mode | 2 → 5 | 4 → 7 | once |
| Claude | 1 → 3up | 5 → 5 | twice |
| Google AI Overview | 3 → 1 | 8 → 4 | once |
| ChatGPT | 0 → 0 | 1 → 1 | twice |
Counts of questions, out of 30. Only a change larger than an engine's own run-to-run disagreement is marked, so unmarked figures have not moved meaningfully.
Named means the answer text itself names Harbour & Key. Cited means a Harbour & Key page appears in the sources behind the answer, whether or not the answer names the firm. An engine can do either without the other, so the two are always shown apart.
ChatGPT, Claude, Gemini and Perplexity answer every question twice, so a single flip in one run does not read as a change. Google AI Overview and AI Mode answer once. A question that flips on one of those two is still shown in the list above, but one answer can swing it, so treat those arrows as a prompt to look rather than as proof of a move.
Google AI Overview does not appear on every search. Questions where it showed nothing are excluded from its own percentage rather than counted against it, and marked with a faint dash in the list.
Being named in AI answers is tracked and aimed at, never promised.