Intro
Without evidence, AI visibility becomes opinion.
The Capston Core evidence layer is the traceable, dated, source-attributed record behind every score, every diagnosis, and every verification. It is what turns AI visibility from a marketing conversation into a board-level conversation.
This page explains what the evidence layer captures, how it is structured, and why it is non-negotiable in every Capston Core engagement.
Request an evidence-backed baseline
What the evidence layer captures
For every prompt, every engine, every capture:
- The exact prompt text
- The engine and model version
- The date and timestamp
- The full AI answer (text and, where relevant, screenshots)
- The citation set (URLs, titles, anchor positions)
- Brand mentions and their position in the answer
- Competitor mentions and their position
- Factual claims and their accuracy status
- The source category for each citation (own site, OTA, travel media, review site, destination board, partner page, outdated listing)
Nothing is paraphrased. Nothing is summarised away. The original capture is preserved.
Why traceability matters
Three reasons.
1. AI engines change. Models update silently. Citation behaviour shifts. Without dated captures, “we moved” becomes “we think we moved.”
2. Internal stakeholders need proof. Marketing claims a win. Tech claims a fix. Revenue claims a loss. Evidence lets each function look at the same record.
3. Partner work needs verification. When a certified partner executes part of the plan, the evidence layer is what confirms the work actually moved the dimension it was meant to move.
Without this layer, AI visibility devolves into anecdote. Capston Core treats anecdote as not-yet-evidence.
How the data is structured
The evidence layer is structured around four objects:
- Prompt — text, intent bucket, target market, assigned engines
- Capture — engine, model version, timestamp, raw answer, citation list
- Source — URL, domain class, trust tier, freshness, accessibility
- Claim — fact about the brand, accuracy status, citation source
These objects compose into the dimension scores, the diagnosis categories, and the action plan. Every recommendation in a Capston Core report can be traced back to specific captures.
What partners and clients see
Capston Core clients receive:
- The full prompt library, locked at baseline
- A dated record of every capture
- The citation map per prompt
- A delta view between baseline and verification
- A monitoring feed for quarterly retests
Certified partners receive the same evidence layer for the engagements they execute, with QA standards applied identically.
How this fits into Capston Core
The evidence layer powers the AI visibility scoring system, enables the Capston Core methodology to produce comparable before-and-after results, and underpins the Capston QA standards.
→ Back to Capston Core
FAQ
Do clients own their evidence?
Yes. The captures, citations, and scores produced for a client belong to that client. Capston Core retains aggregate, anonymised statistics for benchmarking only.
Can captures be re-run on demand?
Yes. Any prompt in the locked set can be re-captured between scheduled retests, with the result added to the timeline.
What about engines that change rapidly?
Model version is recorded with every capture. When engines update, prior captures stay valid as historical records.
Is screenshot evidence stored?
Yes for any answer where the visual format carries meaning (AI Overviews, structured cards, embedded citations). Pure text answers are stored as text plus model metadata.
Final CTA block
See the evidence behind your visibility.
Request a baseline
Read the methodology