Original Research
AI Buyer Prompt Study: 300-Prompt Pilot In Progress
Follow the study design before the data is published; statistics will appear only after real rows are complete.
0 of 300 required observations are complete. This page is noindex and omitted from the sitemap until the study data is ready. No statistics are published before the dataset exists.
Study design
The planned pilot covers 10 industries, 5 buyer prompts per industry, and 6 AI systems: ChatGPT, Claude, Gemini, Perplexity, Grok, and DeepSeek. That creates a target of 300 observations.
Required row schema
A complete observation must include industry, prompt, AI system or signal, tested brand or category, answer summary, brands surfaced, citation state, citation URLs when visible, recommendation status, observed gap, scan date, and reviewer notes.
Metrics collected
The study is designed to summarize brand mention rate, competitor recommendation status, citation/source presence, answer confidence labels, category clarity gaps, evidence gaps, scan date, and reviewer notes only after the rows exist.
Publication threshold
The page becomes indexable only when the local study data file contains at least 300 complete observations. Until then, the page explains the design, stays noindex, and remains out of the sitemap.
Why the page is gated
Publishing invented statistics would weaken trust. The safer authority move is to show the method now and publish aggregate numbers only when the underlying data supports them.
What will be published
Once complete, the page should show aggregate tables by model, industry, prompt type, citation presence, competitor pattern, and observed action gap, plus any limitations that affected interpretation.
Field Note For This Buyer Question
Only count the result when the prompt, AI system or signal, surfaced brands, citation state, scan date, and reviewer notes are complete.
Use this as a diagnostic result, not a guaranteed ranking claim. The scan should show what the answer said at a specific time.
Next supporting fixes: Collect complete observations, Review outliers, Publish aggregate tables.
What a Saved Result Should Show
Illustrative report structure. Treat it as educational unless a page explicitly labels the evidence as a completed scan or approved customer result.
What To Review Before You Fix It
A useful AI buyer prompt study review should connect the buyer prompt, answer status, competitors named, source gaps, and the next action. The goal is not a prettier dashboard. The goal is knowing which evidence could move the answer.
How We Test AI Visibility
Use the question that would hurt most if a competitor won, not a generic keyword.
Save the AI system, date, recommendation status, who AI named instead, and citation/source notes.
Separate direct prompt evidence from directional or indirect signals.
Turn the result into a content, proof, review, citation, or positioning action.
Research pages should publish numbers only after the data file contains complete rows. Until then, the page explains the design and stays out of the sitemap.
What a Useful Report Includes
The exact buyer question tested.
The model or signal reviewed.
When the evidence was captured.
Visible, weak, missing, or competitor-led.
Brands or alternatives surfaced in the answer.
Citations, reviews, or pages to improve.
The first fix to test before the next run.
What To Fix First
- Collect complete observations
- Review outliers
- Publish aggregate tables
- Turn findings into industry and model guides
Sources
Conclusion
AI Buyer Prompt Study: 300-Prompt Pilot In Progress should help a team make a better marketing decision. Start with one buyer prompt, record the evidence, choose a specific fix, and rerun the same prompt after the update is live.
Frequently Asked Questions
Are the study statistics live yet?
No. The study is still collecting real observations, so no statistics are shown yet.
Why use noindex while collecting data?
Noindex prevents an unfinished research page from being treated as a finished public study.
What counts as a complete row?
A complete row includes industry, prompt, AI system or signal, tested brand/category, answer summary, recommendation status, observed gap, scan date, and reviewer notes.