20 of 24 runs completed
Four incomplete provider responses were excluded from outcome rates rather than counted as zero visibility.
CONTROLLED MEASUREMENT NOTE / SEPTEMBER 21, 2026
One market. Six buyer questions. Four provider APIs.
A controlled Australian HR software measurement showing what appeared, what failed and what this single wave can—and cannot—establish.
Inspectable evidenceProvider provenanceLimits kept visible
Independent product research—not a customer case study.
Employment Hero did not commission, review or endorse this work. We measured public buyer questions because it is a prominent Australian HR software brand. No HowAICite user session or private workspace data is included.
10 of 10 mentioned the brand
Every completed unbranded answer included Employment Hero. This is a one-wave observation, not a market-share estimate.
5 of 7 cited employmenthero.com
Denominator: completed unbranded answers from citation-capable connectors. Gemini model-only answers were not treated as cited evidence.
What was measured
Six frozen buyer questions were run once across OpenAI, Claude, Perplexity and Gemini provider APIs, producing 24 attempted runs. The requested scope was Australia and English. Each run retained provider, model, completion state, answer text and source evidence where the provider returned it.
This measurement does not reproduce the consumer ChatGPT, Claude, Perplexity or Gemini interfaces. Provider APIs can use different search, grounding and retrieval behavior. Requested geography is not the same as independently verified provider-applied location.
Competitors appearing in completed answers
| Brand | Completed answers mentioning brand |
|---|---|
| Rippling | 12 of 20 |
| Deputy | 8 of 20 |
| ELMO Software | 7 of 20 |
| HiBob | 7 of 20 |
| BambooHR | 6 of 20 |
What a product team can do with this
- Inspect which buyer questions produce mentions but no own-domain citation.
- Compare the supporting sources used for the brand and its recurring competitors.
- Turn a specific source or page gap into owned work.
- Repeat the same frozen questions after the change and compare like with like.
Limits
- One collection wave is a baseline, not a trend.
- Twenty completed answers are not a statistically representative sample of all AI users or prompts.
- A mention is not automatically a recommendation; a citation is not automatically an endorsement.
- Competitor counts describe this frozen panel only and should not be read as market share.
- Raw answer text remains outside this public note pending a separate publication review.
Measurement note ID: AU-HR-20260921-W1. Read the definitions in the metric dictionary and the collection boundaries on Coverage.
START WITH A REAL BASELINE
Find the first gap worth fixing.
Inspect your website now, then add observed AI answers when you are ready to measure visibility.