How do AI trackers sample answers and count share of voice?
Broadcastwell measures and fixes whether AI search engines name B2B software companies.
Run the free checkSuede reports persona-based capture; Profound defines several denominators; AthenaHQ-supported research publishes eligible-cohort and source-counting rules. 13 checked-name appearances in three Perplexity returns show why an answer total and a mention total need different labels. That is the scaling-small-sales-team CRM (customer relationship management) row in our 22 September 2026 Index.
Observed answers on stated dates, not a ranking.
What we measured
Measured. The Index method and data retain three Perplexity returns to one CRM question. HubSpot CRM, Pipedrive, Salesforce Sales Cloud and Zoho CRM each appear in all three; Freshsales in one: 13 checked-name appearances across three answers, counting each supplier once per return. The panel does not establish provider sampling weights, coverage or account-specific totals.
Claude’s same-release total is 2,039 checked-name appearances across 420 valid answers: 14 categories, ten questions each, three returns per engine on 22 September 2026. Several vendors may appear per answer. This is neither a brand’s answer percentage, a UK sample nor evidence of account access.
Reported. Suede’s Platform page, read on 30 October 2026, describes daily persona-specific questions, captured responses, stance, sentiment and sources, but gives no share-of-voice denominator or sampling weights. Profound’s AEI (Answer Engine Insights) guide, read on 29 October 2026, defines Visibility against responses containing a brand, and Share of Voice against total brand mentions across responses. AthenaHQ’s FAQ (frequently asked questions), read that day, lists mentions, citations, share of voice and content-gap functions. The separately Athena-supported Answer Signal method, version 1.0 of August 2026 and read on 29 October 2026, specifies observation windows, eligible cohorts and normalized root domains. Unless otherwise stated, it excludes flagged responses and deleted records, and uses like-for-like cohorts for longitudinal comparisons.
Reasoned. The 13/3 row shows vendor-name presences exceeding answer counts. For GEO (generative engine optimization), also called AEO (answer engine optimization) or AI search visibility, compare each provider’s share-of-voice definition with that unit.
Two historical answer fragments
For the scaling-small-sales-team question, a Perplexity answer captured on 22 September 2026 includes “Pipedrive, Pipeline CRM, Zoho CRM, and Salesforce Starter Suite”. A ChatGPT return to that question on the same date reads “Pipedrive, HubSpot, Zoho CRM, and Close”. The ChatGPT table includes product-pricing links. Both are historical answer fragments, not verified functionality or explanations of aggregate counts.
What this means for you
Semrush’s toolkit guide, read on 30 October 2026 UTC, lists UK coverage and daily rolling updates for Visibility Overview, Competitor Research and Prompt Research. Brand Performance uses configured location/language and updates weekly; Prompt Tracking updates daily; AI Search Site Audit runs on demand. Worldwide/regional Overview filtering differs from a configured UK brand report. The toolkit FAQ requires a paid subscription and allocated corporate-user limit; its no-trial rule differs from the generic footer invitation.
One comparison in the seoClarity article, updated on 23 October 2025 and read on 29 October 2026, covers UK desktop Overviews from September 2024 to September 2025. It describes an Overview filter in SERP Features (search engine results page features), referenced-site checks and returned content in SERP Previews. These concern Google source links, not Claude brand mentions. Keep each engine, unit, location and date distinct, as in the Profound/seoClarity comparison.
Measured. The CRM CSV records Close in 3 of 30 ChatGPT answers on 22 September 2026, across ten questions with three repeats. All three belong to one scaling-team question; the other nine questions have no Close observations. Each ChatGPT repeat shortlisted Close alongside Pipedrive, HubSpot and Zoho. Valid non-naming answers remain in the base. Across five engines, Close appears in 12 of 150 answers. Those observations do not establish vendor popularity or product quality.
A single-question service could reproduce the same three mentions without covering the thirty-answer panel. Specify questions, repeats, engines, countries/languages, retained text, source links, exports and review work when comparing scopes. The release records no supplier fees, staff time or repair costs. Product-pricing citations require their own billing-unit and dated-source checks; naming counts cannot price that work. Measurement noise is a separate consideration.
Also asked as
I'm a competitive intelligence analyst. methodologically, how do suede, profound, and athenahq differ in sampling ai answers and measuring share of voice?
Additional source: The Answer Signal research method.
How do Suede, Profound and AthenaHQ differ in sampling AI answers and measuring share of voice?
Which are the best ai-powered tools for analyzing online visibility currently available in the uk?
Which AI tools for analyzing online visibility are available in the UK?
What does tracking brand visibility in ChatGPT cost?
chatgpt brand visibility cost?
Related
- Suede Platform fields
- Profound AEI metric definitions
- AthenaHQ product FAQ
- The Manual on valid measurement
- More measurement answers
- CRM records in the Index
Next step
Free 10-question check (one engine). No email is needed; runs are limited per network and per day, and by the site's daily allowance.
Claiming or correcting a vendor record is free and never depends on a purchase.
Every figure comes from the published method, v1.1.
Answer Repair: $490.
Prepared by the Broadcastwell team.
Page owner: Vikram Patel, AI Architect. Corrections: hello@broadcastwell.com
Related evidence
Published . Last updated .