A measurable visibility gap—not a verdict on the whole market.
HowAICite was absent in all 20 answers. At least one tracked competitor appeared in 9 of them. The same ten questions were measured on two dates.
Valid observed answers
Links to howaicite.com
Ten questions × two waves
Scope: Perplexity Sonar API (sonar-pro) · requested US market · English · September 23–24, 2026. See collection method.
Two waves. No observed mention.
Denominators include valid completed answers only. The two dates are repeated observations, not 20 independent buyer questions.
| Date (UTC) | Valid answers | Brand mentions | Own-domain citations |
|---|---|---|---|
| September 23, 2026 | 10 / 10 | 0 / 10 | 0 / 10 |
| September 24, 2026 | 10 / 10 | 0 / 10 | 0 / 10 |
The absence held in both waves. We did not count a missing, failed or empty response as a negative answer; none of these 20 runs failed. All 20 observations had a citation-capable provider source basis, but no first-party URL appeared.
Tracked competitors appeared in 9 of 20 answers while HowAICite did not.
The most useful gap is question-level: four selected questions mentioned a tracked competitor in both waves; a fifth did so once.
Valid answers containing at least one of the four tracked competitor names. This is not nine distinct questions or a market-wide share-of-voice estimate.
| Tracked name | Sep 23 | Sep 24 | Total valid answers |
|---|---|---|---|
| Profound | 5 / 10 | 4 / 10 | 9 / 20 |
| Peec AI | 3 / 10 | 3 / 10 | 6 / 20 |
| Scrunch AI | 2 / 10 | 3 / 10 | 5 / 20 |
| Otterly.AI | 0 / 10 | 0 / 10 | 0 / 20 |
These are exact-name matches in saved answer text, excluding URLs. They measure mentions—not recommendations, competitor-domain citations, rank, sales or exclusive share of voice. Variants outside the four configured names may be missed; a zero for Otterly.AI means no exact match in this panel, not absence everywhere.
A fixed panel with visible provenance.
Each saved buyer question was run once per wave. Answer, source basis and timestamp remain in the private workspace.
- Provider and model
- Perplexity Sonar API ·
sonar-pro - Requested market
- United States · English
- September 23 window
- 10:51–10:53 UTC
- September 24 window
- 03:27–03:29 UTC
- Measured outcomes
- Own-brand mention; own-domain citation; exact-name mentions of four tracked competitors
- Excluded attempts
- None in these two waves
“United States” is the requested market, not independently verified provider-applied geolocation. This API response is not a capture of the consumer Perplexity interface or another chatbot’s consumer experience.
Exactly what we asked.
Ten questions, each observed twice. The panel was selected by us; it is not a search-volume sample.
- How can agencies create client-ready AI visibility reports?2 valid · 0 mentions · 0 own-domain citations
- How can marketing teams measure whether AI assistants cite their website?2 valid · 0 mentions · 0 own-domain citations
- How do I track first-party citation rates in AI-generated answers?2 valid · 0 mentions · 0 own-domain citations
- What are the best AI visibility monitoring tools for B2B SaaS companies?2 valid · 0 mentions · 0 own-domain citations
- What are the best generative engine optimization tools for agencies?2 valid · 0 mentions · 0 own-domain citations
- What is the best AI citation monitoring software for small businesses?2 valid · 0 mentions · 0 own-domain citations
- What platform can monitor citations across ChatGPT, Claude, Perplexity, and Gemini?2 valid · 0 mentions · 0 own-domain citations
- What tools can verify whether GEO changes improved AI citations?2 valid · 0 mentions · 0 own-domain citations
- Which AI search visibility tool provides evidence-backed recommendations?2 valid · 0 mentions · 0 own-domain citations
- Which GEO software tracks brand mentions and citations in AI answers?2 valid · 0 mentions · 0 own-domain citations
Download the aggregate question-level CSV ↗ The CSV includes competitor answer counts by question, not raw AI answers or customer data.
The work shipped. The outcome has not been remeasured.
A page edit is an intervention to test, not evidence that visibility improved.
Original questions, timestamps and source basis retained privately.
No post-change answer has been collected for this note. We will compare the same question, provider, market and language later.
Three internal fix tasks retain their baseline observations. A later mention or citation would be an observed change; it would not, by itself, prove the page edits caused it.
What not to conclude.
- Twenty answers across two dates do not represent all US buyers, all prompts or all AI products.
- Provider APIs can retrieve and cite differently from consumer interfaces; this note makes no claim about ChatGPT, Claude, Gemini or the consumer Perplexity app.
- A mention is not necessarily a recommendation; a citation is not a sale or endorsement.
- Retrieval and model behavior can change over time, so a before/after difference is not automatically causal.
- Raw third-party answer text and private workspace evidence are not republished here.
For definitions and connector boundaries, see the metric dictionary and coverage matrix. We will update this note only when a comparable, valid post-change wave has been reviewed, keeping the original baseline visible.