Brand, website, product, and campaign scenarios
The Target Graph lets teams measure a parent brand and its individual websites, pages, products, services, campaigns and claims without collapsing every signal into one score. Each scenario uses the same evidence rules but different targets, prompts, metrics and actions.
Scenario selection
| Question | Primary targets | Primary intelligence | Typical action |
|---|---|---|---|
| Is our brand present and accurately described? | Brand, aliases, people, claims and competitors | Visibility, SOV, entity accuracy, narratives and citations | Knowledge, authority and message corrections |
| Can AI systems understand and cite our website? | Website, pages, content and machine artifacts | Discoverability, citability, source influence and crawler activity | Technical/content updates and structured data |
| Are our products or services recommended for relevant needs? | Product line, product/service, pages, audience and competitors | Product visibility, prompt demand, comparison and factual accuracy | Product content, evidence and feed changes |
| Did a campaign improve generative visibility and outcomes? | Campaign, message, claim, channels, market and audience | Campaign lift, narrative, citation, referrals and conversions | Message/channel iteration and budget decisions |
Brand baseline
Use this scenario when leadership needs a defensible baseline across AI systems, markets and competitors.
Required setup
- One owned brand target with legal, public and localized aliases
- Direct competitor targets such as category peers
- Markets and languages
- Category, comparison, factual, recommendation and safety prompts
- At least one active provider; multiple providers for cross-engine claims
Decision dashboard
Review visibility and SOV with sample/confidence, entity ambiguity, cited domains, unsupported claims, narrative themes, provider differences, freshness and top actions. A 100% SOV result with only one eligible brand is not a competitive conclusion.
Acceptance outcome
The team can identify where the brand is missing, confused or inaccurately described; see evidence; assign an owner; and verify whether the next observation changed the result.
Website and page citability
Use this scenario when the question is whether owned web properties provide discoverable, trustworthy and citable evidence.
Measure separately
- Domain availability, canonical URLs and indexability signals
- Page identity, title/headings, answer clarity and source evidence
- Structured data validity and target applicability
- Sitemap/robots/
llms.txtreachability without treating them as ranking guarantees - Citation presence and source influence
- AI crawler activity and AI referral sessions
Common actions
Create an evidence-backed FAQ, clarify claims, add appropriate JSON-LD, correct canonical identity, improve product/service page specificity, generate a reviewed llms.txt artifact, or publish through a CMS/repository connection.
Product or service visibility
Use this scenario when a product line, product or service must be evaluated independently from the parent brand.
Guardrails
- Do not infer product availability, price or suitability from stale evidence.
- Separate product-family visibility from individual SKU/product visibility.
- Model competitor products explicitly and use the same prompt denominator.
- Treat recommendation occurrence as visibility evidence, not guaranteed purchase intent.
- Link claims to approved product evidence and validity windows.
Campaign measurement
Use this scenario for a launch, seasonal campaign, sponsorship, content program or message change.
Campaign model
Create a campaign target connected to messages, claims, audience, market, landing pages, social/channel assets and products. Store the launch window and annotations so dashboards do not confuse campaign timing with provider/model changes.
Success measures
- Change in eligible prompt visibility and citation presence
- Narrative/message adoption and factual consistency
- Landing-page citation and source influence
- AI referral sessions and downstream outcomes
- Experiment result when a valid control/candidate design exists
- Cost, publication success and time-to-impact
Campaign lift remains directional without a suitable control and sufficient sample. Attribution must state its model and tracking coverage.
Multi-market program
Do not compare markets unless prompt intent and metric applicability are compatible. Localized names should be aliases or localized target metadata rather than duplicate unrelated brands.
Custom agent, MCP, or RAG system
Register each component and dependency so a result can be traced through the agent, model, router, MCP tool and RAG source that influenced it.
Operating cadence
| Cadence | Activity |
|---|---|
| Daily/weekly | Monitor failed/stale runs, high-risk narratives, provider health and urgent actions |
| Weekly/biweekly | Review competitive visibility, citations, prompt gaps and action progress |
| Campaign window | Capture pre/post observations, annotations, publications and tracked outcomes |
| Monthly | Review targets, competitors, metric confidence, dashboards and scheduled reports |
| Quarterly | Revalidate strategy, provider coverage, knowledge, integration scopes and retention |