Accounting Firms
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Methodology
How to read the AI visibility snapshots
A limited observation of recorded answers, with the collection method and its limits stated alongside the data.
What each metric means
- Share of mentions: a name’s recorded mentions divided by all name mentions in that snapshot. An answer can name several businesses, so this is not the percentage of answers naming it.
- Mention count: the number reported by the source answer extraction. The displayed table is an extract, not the entire set of names behind the denominator.
- Average answer position: the source measurement of where the name appeared within answers. This is separate from the name’s rank in our table. Missing verified source values are shown as unavailable.
Collection records for the published snapshots
The compiled date records when the dashboard data was assembled; it does not establish a new measurement. Each measurement date below records when its source snapshot was collected.
Aesthetic Clinics
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Company Secretary
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Dental
Measured . 12 questions; 36 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Fertility
Measured . 12 questions; 36 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Hair Transplant
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Interior Design
Measured . 12 questions; 36 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Law Firms
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Plastic Surgery
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Private Schools
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Wealth Management
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Wedding Photography
Measured . 15 questions; 45 recorded answers across ChatGPT, Claude, Gemini. The question set includes general and named-business prompts, so the combined sample is not an unbranded recommendation rate.
Prompt and entity limitations
Published question examples include alternatives to named businesses. Those prompts can influence which names appear. The historical totals combine general and named-business questions; we cannot describe the combined result as purely unprompted recommendations. Separate branded and unbranded rates are not published because a verified split is unavailable in these extracts.
Extracted names are not a vetted list of service providers. The accounting snapshot includes MIA, Xero and QuickBooks, plus separate RSM name variants. That page is labelled as unclassified names and does not name a recommended accounting-firm winner. We have not added variant counts together: doing so could double-count answers.
What a snapshot does not prove
Answers change with prompts, engines, location, session and time. A mention is not an endorsement, a stable rank, traffic, an enquiry or market share. No current five-engine consumer-app claim is applied retrospectively to these three-engine records. Rebuilding the same data does not create a new observation, so trend comparisons require distinct, comparable measurement runs.
The revised export uses average positions from matching original snapshots. Source precision is retained; rounded percentages may not total exactly 100%. Entity variants remain explicit rather than silently merged.
Method explanation corrected 14 September 2026. This is an editorial correction, not a new AI visibility run.