Methodology · CiteSurge
How CiteSurge measures, improves, and verifies AI visibility.
The five stages are plain: see what AI says, identify the gap, choose the work, make the change, and measure again. Each stage keeps facts, decisions, and limitations distinct.
Five stages
What does CiteSurge record at each stage?
CiteSurge records four things separately: what it found, what it recommended, what changed, and what later measurements showed.
Observe
See what AI says. Record the pages, buyer questions, markets, configured AI systems, available answers, mentions, citations, and technical evidence in scope.
Diagnose
Identify the gap. Separate what was observed from the interpretation, then connect supported content, source, entity, technical, or public-coverage gaps to the affected page or source.
Prioritize
Choose the work. Decide which supported pages, content, press materials, or technical changes should move first, who owns them, and what success would look like before implementation begins.
Implement
Make the change. Help content, brand, communications, web, and engineering teams create, review, approve, publish, or release the work. Keep the recommendation separate from what the teams completed.
Verify
Measure again. Compare later observations within their own AI system, market, prompt, timing, and availability context. Comparable change can be verified; causal attribution requires additional evidence.
The proof behind a finding
What does CiteSurge keep with each finding?
The proof behind a finding preserves enough context for another reader to understand what was checked and what the observation means. It keeps the measured AI system, prompt, market, time, page or source, availability state, and relevant response evidence together rather than presenting a score without its measurement boundary.
- URL
- Market
- AI system
- Prompt
- Timestamp
- Availability context
- Keep the URL, market, AI system, prompt, timestamp, and availability context attached where supported.
- Distinguish owned sources, independent sources, mentions, and citation links.
- Tie specific changes to the affected page or source gap and record completed work separately.
- Publish an outcome only when its baseline, intervention, later observation, limitations, and permission form a complete evidence chain.
Further reading
- AI mentions vs citationsThe core measurement distinction
- Why a coding agent cannot supply GEO measurement by itselfThe implementation boundary
- How AI is changing enterprise technology vendor selectionThe buyer-side application
- The agentic internet needs proof, not more mentionsThe claim-to-source standard
- Martin Lange's prompt-injection researchThe security boundary around AI workflows
- The Claude and Grok freshness studyOriginal citation research
The editorial and research standards explain how the same sourcing, review, disclosure, and correction rules apply when CiteSurge turns the record into public analysis.
Uncertainty and causation
How does the method handle uncertainty and causation?
AI systems change, configured coverage varies, and repeated prompts can produce different responses. CiteSurge therefore reports observations inside their engine, market, prompt, time, and availability context. A later improvement can be real and useful evidence. Verification establishes comparable change; causal attribution requires additional evidence.
Technical readiness is also distinct from answer inclusion. Crawlable pages, coherent schema, clear passages, and reliable source references make evidence accessible, but they do not compel an AI system to cite a page. Similarly, a mention without a citation does not identify the source behind the answer. The method retains those distinctions so a team can act on supported gaps without presenting probability as certainty.
Verification uses the most comparable later record available and states material differences in scope. When provider availability, prompt wording, market, or evidence collection changes, that limitation stays attached to the comparison. Incomplete comparability is labeled and narrows the supported conclusion.
See Work & Evidence for the public proof standard and the four capabilities for the work it governs.
