What you will achieve
Produce a bounded library-information audit: the questions tested, sources inspected, gaps found and next actions with owners. The outcome is a practical improvement queue, not a universal readiness score or a new publishing plan.
Use available AdCyte analysis and agent capabilities to investigate saved material. Keep the audit in an ordinary working document where useful. The guide does not assume an automatic audit dashboard or a bulk repair action.
Before you start
Choose one campaign family or product range and a named internal recipient. Gather the relevant assets, saved analysis, approved references and existing writing. Record what is available and what is expected but missing.
The example below uses a cookware campaign family. Its records are illustrative and do not describe a customer result. Replace them with actual source locations and decisions when running the audit.
1. Select three real questions
Choose questions that represent different jobs: a product explanation, a creative comparison and a reuse decision. For example: “How is the handle removed?”, “How does the storage message differ between these campaigns?” and “Which approved answer already explains this model?”
Keep each question specific enough to inspect. A request to assess all brand knowledge is difficult to complete because neither the required information nor the decision is bounded.
Check: each question has a named user and a plausible next action.
2. Record the scope
List the selected campaigns and relevant versions. Identify longer edits, short cuts and updated-product films that may change interpretation. Record omissions explicitly.
If starting with a sample, explain how it was selected. Include contrasting material, such as a demonstration and a brand-led film, where that reflects the intended work. Do not describe a small convenience sample as proof of the entire library's condition.
Check: another reviewer can reproduce the selection and understand its limits.
3. Test identity and context
For each useful source, check product identity, campaign name, relevant version and known market. Mark unresolved fields rather than guessing from the creative.
| Example asset | Known context | Gap to resolve |
|---|---|---|
| Long storage advert | Campaign and source version known | Exact pan model needs confirmation |
| Short social cut | Related edit identified | Qualification absent from this cut; inspect longer source |
| Later demonstration | Revised product range identified | Current compatibility reference required |
Use the table to isolate identity problems before drafting explanations. Several similar files can produce a plausible combined answer that describes no actual product accurately.
Check: evidence is not combined across incompatible products or versions.
4. Inspect the evidence for each question
Open the relevant findings and supporting moments in AdCyte. Separate demonstrated actions, spoken claims, on-screen conditions and interpretation. Inspect surrounding context when a short moment appears important.
For the handle question, record the visible sequence and whether the mechanism is clear enough to explain. Do not infer compatibility, durability or safe use from the action alone. Identify the appropriate approved reference for those facts.
Check: each proposed finding has a source and a clear statement of what it establishes.
5. Look for existing writing
Where available, ask the agent to identify saved writing that addresses the selected questions. Inspect the actual records, versions and states. Keep drafts, approved content and published destinations distinct.
A useful prompt is: “Which existing writing addresses these three questions? Identify sources and review states, and tell me where the available records are incomplete.” The prompt investigates existing material; it does not authorise creation or publication.
Check: the audit records useful existing work before recommending new content.
6. Complete the gap register
| Question | What is usable | Main gap | Next action |
|---|---|---|---|
| How is the handle removed? | Demonstration shows the action | Exact model unresolved | Product owner confirms identity |
| How does the storage message differ? | Relevant scenes from two campaigns | Later film has limited saved context | Inspect available source and narrow comparison if needed |
| Which answer can we reuse? | Approved explanation found | Current compatibility not yet checked | Reviewer checks maintained reference |
The rows show different problems. None automatically requires a new article. A factual clarification or a targeted review may unlock a useful existing answer.
Record the source locations alongside the live audit rows. The example table is deliberately compact; your working document should contain enough evidence for each decision to be inspected.
Check: each gap is stated as something a person can resolve, rather than a vague instruction to improve quality.
7. Prioritise and assign
Consider the importance of the customer question, likely reuse and effort required. Start with a small number of improvements that unblock meaningful work. Give each a named owner and a review point.
Avoid adding numerical precision without a useful method. If a simple high, medium or low priority helps the team, define what it means. Do not publish an invented readiness percentage as a product measurement.
Check: the queue is short enough to act on and each action has an accountable recipient.
8. Retest after the changes
Once the owners resolve the gaps, repeat the same bounded questions and inspect the new evidence. Record what improved and which limitations remain. If the underlying scope changed, explain that difference.
Keep the internal audit outcome separate from publication and discovery. A more usable library may support better content work, but no external visibility result has been established merely by completing this document.
Troubleshooting
| Problem | Response |
|---|---|
| Too many gaps appear | Narrow the scope to the decisions that matter first |
| Product identity is uncertain | Assign confirmation before combining evidence |
| Agent finds no useful writing | Record the scope and access limits; verify before declaring none exists |
| Old approved content is found | Review current facts before reuse |
| The sample is unrepresentative | Describe the limitation and select another bounded sample |
| Nobody owns a reference | Resolve ownership before expanding dependent content |
Completion checklist
- Three useful questions and their users identified.
- Scope, selection method and omissions recorded.
- Product identity and source context checked.
- Evidence inspected and interpretation distinguished.
- Existing writing reviewed by version and state.
- Gaps translated into owned actions.
- Retest and maintenance responsibilities agreed.
Put the audit to work
Explore Video Intelligence to investigate the material behind your answers. For an audit involving several teams, discuss the workflow.
FAQs
Does this audit create new content?
No. It identifies what the team can use and what needs attention. Writing and publication are subsequent decisions.
Must we inspect every video manually?
Use analysis to support the work and prioritise human inspection around the questions and claims being reviewed. Clearly state whether the audit covers a sample or the full selected collection.
Is the result an AI-readiness score?
No. It is a record of tested questions, evidence and practical gaps. Do not turn it into a universal score without a defined, justified method.

