What does a completed AI visibility case show?
Broadcastwell measures and fixes whether AI search engines name B2B software companies.
Run the free checkFictional sample. A completed AI visibility case shows the company, the agreed buyer questions and their base, the capture dates, the five engines, the named count before and after with each base and its Wilson 95 percent interval, the sources the engines cited, the commercial fields the client supplies, and the client's written permission to publish. Every value below comes from the fictional Kalvenor sample (Sample data, fictional company).
Updated 11 November 2026.
Kalvenor Systems is a fictional field service software company. Its figures are published in our fictional sample to show the format, and they describe no client and no real result. This page describes no client case.
1. The company and the category
| Field (fictional sample) | Value |
|---|---|
| Company | Kalvenor Systems, a fictional company |
| Category, in the buyer's words | field service management software |
| Own domain | kalvenor.example |
| Rivals recorded for the sample | Northvale FSM, Trakwell, Orbit Field, Asterforge, Cindermark and Verano FSM |
2. The buyer question set and its base
| Field (fictional sample) | Value |
|---|---|
| Questions | 35 brand-free buyer questions, frozen before capture |
| Engines | ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode |
| Runs | 3 scheduled runs of every question on every engine |
| Scheduled answers | 525 |
| Scored answers, the base | 518 |
| Excluded answers | 7, listed with their reason and kept out of every base |
| Method | Method 1.1 |
3. Capture dates
| Capture (fictional sample) | Dates |
|---|---|
| Baseline | 8 to 12 July 2026 |
| Baseline run used for the comparison | 12 July 2026 |
| Monthly run | 12 August 2026, under the same method |
4. Named counts at baseline
Each count is shown with its base and its Wilson 95 percent interval; the three measured rates defines each one.
| Measure (fictional sample) | Count and base | Rate | 95% interval |
|---|---|---|---|
| Kalvenor Systems named | 125 of 518 | 24.1% | 20.6% to 28.0% |
| Most named rival, Northvale FSM | 329 of 518 | 63.5% | 59.3% to 67.5% |
| Kalvenor's own pages cited | 66 of 518 | 12.7% | 10.1% to 15.9% |
| Questions with at least one Kalvenor appearance | 21 of 35 | Not a rate | Not computed |
5. Before and after
The comparison uses only the engine and questions both runs asked. A change is reported only where the two 95 percent intervals separate.
| Fictional sample, ChatGPT | Count and base | Rate | 95% interval |
|---|---|---|---|
| Before: baseline run, 12 July 2026 | 15 of 29 | 51.7% | 34.4% to 68.6% |
| After: monthly run, 12 August 2026 | 0 of 10 | 0.0% | 0.0% to 27.8% |
Result (fictional sample): moved down on ChatGPT; the two intervals do not overlap. Both bases are below 30 answers. Our client reports print counts only below 30 observations; the rates and intervals here are the ones the fictional sample publishes. A case records a fall the same way it records a rise.
Provider releases between captures: from our model release tracker, OpenAI published "Updates to GPT-5.6 in ChatGPT" on 6 August 2026. That release is real and the Kalvenor counts are fictional, so the mark shows only where the field sits; under our annotation rules a release never explains a change.
6. The source chart
| Domain | Answers citing it | Base |
|---|---|---|
| hvacoperations.example | 105 | 518 |
| servicetechforum.example | 99 | 518 |
| veranofsm.example | 97 | 518 |
| trakwell.example | 93 | 518 |
| orbitfield.example | 92 | 518 |
| asterforge.example | 89 | 518 |
| northvalefsm.example | 85 | 518 |
| softwarecompare.example | 85 | 518 |
| cindermark.example | 84 | 518 |
| servicebuyersguide.example | 81 | 518 |
| kalvenor.example, the sample company | 66 | 518 |
A citation is an address an answer referenced, not a visit, and it does not show why an engine chose the source.
7. Commercial metrics a real case carries
A real case reports these fields from the client's own systems, each with its source and date range, and keeps each one apart from the answer counts. Our method states: "Referral traffic, self-reported discovery and pipeline remain separate." The fictional sample has no commercial values, so none are shown.
| Field | Where a real case gets it | Fictional sample |
|---|---|---|
| AI-referral sessions | The client's analytics, under the AI-referral channel definition agreed at onboarding | No value: fictional |
| Search Console generative AI performance | The client's Search Console, checked at onboarding | No value: fictional |
| Leads that name an AI assistant as their source | The source question on the client's forms | No value: fictional |
| Brand-search and home page changes | Read monthly alongside the supplied analytics | No value: fictional |
| Pipeline and revenue | The client's own records, under a rule the client states in writing | No value: fictional |
"A crawler visit is not a citation, and a citation is not a lead," as our Fix Sprint page puts it.
8. The client-permission checkpoint
A real case is published only after this checkpoint passes. Our method states: "Client baselines, findings, question sets and outcomes require written approval before publication."
| Field | What a real case records | Fictional sample |
|---|---|---|
| Approver | Name and role of the person approving on the client's side | Not applicable: fictional |
| Written approval | Date and copy of the approval | Not applicable: fictional |
| What may be published | Company name, question set, counts, quoted answers and commercial fields, each approved or withheld | Not applicable: fictional |
| What stays private | Every field the client withholds | Not applicable: fictional |
| Withdrawal | How the client withdraws approval, and what we change when they do | Not applicable: fictional |
Related pages
The content strategy template is filled on the same fictional sample, our operating framework gives the stages and clocks behind each field, and the fictional findings sample shows a single report in full.
Sources
- Broadcastwell, fictional Kalvenor sample, 35 questions, baseline 8 to 12 July 2026, monthly run 12 August 2026: https://app.broadcastwell.com/sample (read 9 November 2026).
- Broadcastwell, Methodology, method v1.1, on intervals, exclusions, disclosure and separate outcomes (read 9 November 2026).
- Broadcastwell, Fix Sprint, on measurement setup (read 9 November 2026).