# Story angles and interview guide

These are reporting directions, not claims that a funding event, partnership or new launch has occurred. Reporters remain free to test the premise and interview independent sources.

## 1. What does an AI visibility score actually count?
**Angle:** A single score can hide several different measurements. Omni Impact's public [How we measure](https://omniimpact.net/docs/methodology/how-we-measure) guide gives a concrete starting point for examining mention rate, citation rate and share of voice.

**Questions:** What is the denominator? What happens when a model fails? How are duplicates and branded questions treated? Why might two tools disagree? What uncertainty remains after aligning their inputs?

**Evidence to request:** A privacy-safe answer-level export, the prompt list, metric definitions and a comparison with recorded exclusions. The public release note is a lead for investigation, not independent proof of accuracy.

## 2. The agency workflow after the AI-answer screenshot
**Angle:** An agency's work does not end with showing a client a model response. Explore the handoff from evidence to an actual website improvement.

**Questions:** Which finding is actionable? Who approves the wording? What goes to a CMS, what becomes a pull request, and what remains manual? How is the change recorded and checked later?

**Evidence to request:** One end-to-end demonstration using a permitted demo site, with the proposed change and the observed outcome kept separate.

## 3. A small Rochester software company working with AI agents
**Angle:** A local founder building an AI-related product while also using coding agents in the development process.

**Questions:** What does Ben still decide himself? How are generated changes reviewed? Where do agents save effort, and where do they create rework? What was the actual origin of the company?

**Evidence to request:** An attributable founder interview and a demonstration of a non-sensitive development workflow. This kit supplies no headcount productivity multiplier or invented personal anecdote.

## 4. A website can be readable without being recommended
**Angle:** Explain the gap between technical readiness, information quality and what an AI system ultimately says.

**Questions:** What does a crawl prove? What does it fail to prove? Can a highly readable site still be absent from an answer? Why might a brand be mentioned without its website being cited?

**Evidence to request:** A defined diagnostic, a separate answer sample, and an independent search or information-retrieval specialist's interpretation.

## 5. The sources behind the answer
**Angle:** Brand teams may discover that other sites describe their business more often than their own pages do. Reporting can examine how teams inspect those sources without treating every citation as trustworthy.

**Questions:** Are the sources current? Do they agree? Can the underlying claim be verified? Is the source a review, the company's own page or another kind of evidence?

**Evidence to request:** A small set of attributable, publicly accessible answers and the pages they actually cited, recorded with date and execution context.

## 6. What responsible AEO should avoid promising
**Angle:** Distinguish useful measurement and website work from claims of guaranteed AI placement.

**Questions:** What would falsify an optimization claim? When is a before/after chart misleading? Which platform guidance contradicts a proposed tactic? Can a vendor explain missing data without turning it into zero?

**Evidence to request:** The external context sources, a vendor demonstration and a skeptical independent perspective. No competitor allegations are supplied by this kit.

## A practical 20-minute interview

1. **Minutes 0-3:** Ask Ben to describe the company and the original problem in his own words.
2. **Minutes 3-8:** Follow one buyer question through an answer, mention/citation distinction and source inspection.
3. **Minutes 8-12:** Examine one website/content finding and its proposed change.
4. **Minutes 12-16:** Ask what a score, comparison and confidence interval do and do not establish.
5. **Minutes 16-19:** Discuss the business model, intended users and Ben's development process.
6. **Minute 19-20:** Confirm spellings, facts, attribution, image captions and any promised follow-up evidence.

## Optional recording shots

- Founder speaking to camera, with name and role confirmed in the actual recording.
- A clean demo workspace showing the question, sample conditions and one answer.
- A source page beside the answer that cited it.
- A proposed content/code change before approval.
- The answer behind a counted number, with its engine, date and question visible.

The included screenshots are suitable as labeled illustrations. They are not video footage or an independently observed customer session. Arrange recordings through hello@omniimpact.net.