Specimen deliverable · fictional subject
Sample AI Visibility Baseline report
This is the deliverable you receive, reproduced in full: the frozen prompt set, per-platform scores across four dimensions, the findings behind them, and the remediation roadmap in the order it is worth executing.
This is a specimen report. The subject is a fictional company invented to show the structure, scoring, and roadmap format of a real AI Visibility Baseline. The figures are examples of layout, not measurements of any organization.
Report header
- Subject
- Northbridge Compliance (fictional)
- Category
- SOC 2 readiness consulting, US mid-market
- Measurement window
- Illustrative window: 5 business days
- Prompt set
- 52 prompts agreed with the subject and frozen before testing
- Platforms
- ChatGPT, Google AI Overviews, Gemini, Claude, Perplexity
- Runs
- 3 runs per prompt per platform, reported separately, never averaged into one score
- Scoring
- 0–100 per dimension per platform, derived from the recorded answers only
- Attribution
- No claim that any single change caused an answer to change; re-measurement reports what moved, what did not, and what cannot be attributed
- Re-measurement
- The same frozen prompt set re-run 30 days after implementation begins
Scores by dimension and platform
Each platform is scored separately from 0 to 100 on each of the four dimensions. There is no single blended number, because a brand that resolves cleanly on one platform and not another has a different problem from one that fails everywhere.
| Dimension | ChatGPT | Google AI Overviews | Gemini | Claude | Perplexity |
|---|---|---|---|---|---|
| Identification | 72 | 58 | 64 | 55 | 81 |
| Retrieval | 66 | 71 | 62 | 48 | 77 |
| Citation | 34 | 22 | 29 | 18 | 46 |
| Description accuracy | 51 | 44 | 47 | 40 | 59 |
Findings behind the scores
Identification
Does the system resolve the correct organization from the name alone?
Two platforms conflate the subject with a similarly named UK accountancy practice. Identification is the binding constraint: nothing downstream is reliable until the entity resolves.
Retrieval
Can the subject's own pages be fetched, parsed, and quoted?
Service detail lives inside a client-rendered tab component, so the substantive copy is absent from the delivered HTML. Third-party directory pages are quoted instead.
Citation
Is the subject named, and linked, on the prompts its buyers actually ask?
On 41 of 52 prompts the answer names competitors and not the subject. The cited evidence is almost always a review aggregator or a vendor comparison post, not the subject's site.
Description accuracy
Is what the system says about the subject current, correct, and complete?
Three platforms state a service line that was discontinued, and four omit the audit-readiness offer entirely. Both errors trace to stale third-party profiles.
The prompt set, row by row
Six of the 52 rows are shown here. In the delivered report every prompt appears with the same columns, plus the verbatim answer recorded on each run so any conclusion can be checked against what the platform actually said.
| Prompt | Intent | Subject named | Competitors | Evidence cited |
|---|---|---|---|---|
| Who are the best SOC 2 readiness consultants for a 200-person SaaS company? | Vendor shortlist | Perplexity only | 6 | Review aggregator listicle, two competitor blog posts |
| How long does SOC 2 Type II readiness take? | Informational | No platform | 4 | Competitor guide, auditor association page |
| Is Northbridge Compliance a real company? | Entity verification | All five | 0 | Subject homepage, LinkedIn, one directory profile |
| What does Northbridge Compliance charge for SOC 2 readiness? | Commercial detail | All five | 2 | None on subject site; two competitor pricing pages quoted |
| Northbridge Compliance vs a Big Four readiness engagement | Comparison | Three of five | 5 | Competitor comparison post, forum thread |
| Which compliance consultancies work with fintech startups in Texas? | Geographic shortlist | No platform | 7 | Regional directory, two competitor location pages |
Prioritized remediation roadmap
Ordered by the constraint each item removes, not by effort. Items 1 and 2 are sequenced first because nothing below them can be relied on while the entity resolves incorrectly and the copy is missing from the delivered HTML.
01 · Identification
Publish a canonical entity record with a disambiguating definition, NAP block, and sameAs set.
Identification — two platforms resolve the wrong organization.
Scope: 1 page, plus structured data on every template
02 · Retrieval
Server-render the service detail currently hidden behind client-side tabs.
Retrieval — the substantive copy is not in the HTML crawlers receive.
Scope: Template change across 9 service pages
03 · Description accuracy
Correct the discontinued service line on the four third-party profiles that platforms quote.
Description accuracy — errors originate off-site.
Scope: 4 outreach requests, 2–6 weeks to propagate
04 · Citation
Publish stated pricing bands and a written timeline for the readiness engagement.
Citation — commercial questions are answered from competitor pages.
Scope: 1 page
05 · Citation
Build answer pages for the 11 buyer questions where no platform names the subject.
Citation — no retrievable evidence exists on the subject's domain.
Scope: 11 pages
06 · Citation
Publish location coverage pages for the two regions where geographic prompts return competitors.
Citation — geographic intent has no matching evidence.
Scope: 2 pages
Recommendation
Proceed with implementation, in roadmap order. Identification and retrieval are the binding constraints here and both are inside the subject's control. If the identification finding had been clean and the citation gap had traced entirely to third-party publishing the subject cannot influence, the recommendation in this section would have been to stop after the baseline.
Every baseline ends with a recommendation that can be “stop here”. Reporting standards for this document are set out in our editorial policy, the fact tagging follows the provenance standard, and anything we get wrong is corrected in public on the corrections page.
To get this report for your own brand and category, tell us the questions your buyers ask. Scope, schedule, and price are confirmed in writing before any testing begins.