What does Athena publish about data accuracy?
The AthenaHQ accuracy FAQ describes collecting AI answers and tracking visibility, then points to customer outcomes as evidence of reliability. For a truth check, you need the underlying claim and its reference instead. Wrong About You classified 1,924 of 2,922 statements as true using 22 September and 23 September 2026 captures; its vendor-reference check took place on 8 October. Those dated classifications are ours, not a test of AthenaHQ.
What we measured
Reported. We read AthenaHQ's data and accuracy FAQ on 29 October 2026. It describes collecting responses and reporting mentions, citation rate, share of voice, recommendation rate and sentiment. Its reliability section discusses customer results. Those reported outcomes are not a claim-by-claim validation table on that page.
The Answer Signal's method, version 1.0 from August 2026, uses aggregated AthenaHQ records. The site is labelled Supported by AthenaHQ. It defines citation share over eligible answers, counts a root domain once per answer and distinguishes entity mentions from recommendations. Unless a study states otherwise, flagged responses and deleted records are excluded. We read this separate research method on 29 October too; it should not silently become the specification for every AthenaHQ product metric.
Measured. Our Wrong About You classification method extracted statements and checked them against saved vendor pages. Its complete base includes true, wrong, stale and unverifiable classes. Team review of flagged records remains in progress. The study did not test AthenaHQ's collection or classification system.
What this means for you
Reasoned. The FAQ's content-gap description concerns areas where AI misunderstands or lacks business information. That is a different label from The Answer Signal's entity-mention definition, which counts resolved references in eligible answers. A named company can therefore satisfy that mention definition without any test of the surrounding product claim. A mention is not a truth check.
Read the eligibility rule before comparing percentages. The Answer Signal method excludes flagged and deleted records unless its particular study says otherwise. Wrong About You keeps unverifiable statements visible in its complete statement base. A percentage calculated from retained answers and one calculated from all extracted statements describe different units, even when both appear beside the word accuracy.
The root-domain rule addresses another particular issue. Repeated links to the same source in one answer count once under the research method. That prevents repeated links from becoming extra answer-level citation events; it does not establish whether a product claim on the cited page is correct. Keep a citation-count reconciliation separate from a factual-claim review.
The FAQ cites Grüns with a reported sixfold share-of-voice lift in 60 days. This concerns a customer's visibility measure; the FAQ does not turn it into the number of factual statements checked or the number agreeing with references. For an AthenaHQ export containing a disputed product claim, retain that claim's source passage and classification separately from the Grüns outcome.
Also asked as
What does Athena publish about data accuracy?
Related
- AthenaHQ data and accuracy FAQ
- The Answer Signal research definitions
- Wrong About You method and records
- Valid measurement in the Manual
Next step
Use Free 10-question check (one engine) to inspect a defined set of buyer questions. No email is needed; runs are limited per network and per day, and by the site's daily allowance.
For one documented gap, Fix one answer, $490. Keep the question and source receipt with the requested change.
Our answer-study method, v1.1.
Prepared by the Broadcastwell team.
Page owner: Vikram Patel, AI Architect. Corrections: hello@broadcastwell.com