AI Conversation Insight Analysis

Conversation Insight analyses the saved AI platform answers for one monitored conversation. Use it when you want a concise assessment of how your tracked brand appears for a particular buyer question, which competitors and sources matter, and which evidence-based actions to take next.

Conversation Insight is part of AI Visibility → AI Conversation Responses. It is different from AI Visibility Analysis on the main dashboard, which analyses a selected persona or compares several personas.

Before you start

You need:

• an AI Visibility configuration with completed monitoring results;
• a conversation with answers matching the filters you want to study;
• access to the conversation's project and workspace; and
• at least one tracked name under AI Visibility Settings → Your Brands.

Conversation Insight uses credits. Enhanced analysis can use additional credits for public research.

Choose the analysis scope

1. Open AI Visibility → AI Conversation Responses.
2. Use Project, Configuration and Phase to find the question.
3. Apply AI Platform, Date Range, Sentiment, Has citations and Cites your domain as needed.
4. Expand the configuration and select the conversation.
5. Select Analyse with AI.
6. Check the scope under Analyse this conversation. Close the panel if you need to change the page filters.

The analysis uses the selected conversation and all visible response-page filters. It analyses the newest 30 matching monitoring dates at most. A custom date range can cover no more than 366 days. These limits do not shorten Response History, which remains paginated separately.

Choose Standard or Enhanced

Enhanced research is selective. It does not search merely to repeat saved Claims Verification work or collect background information that would not change the recommended action. If useful research cannot be completed, the report can continue from the existing analysis and saved evidence with a limitation.

Run the analysis

1. Select Standard or Enhanced.
2. Select Analyse with AI.
3. Follow the progress message while Wheremore prepares the selected evidence, analyses the conversation and, for Enhanced mode, researches and refines useful recommendations.

The work runs on the server. You can close the analysis panel or leave the page without stopping it. Return to the same conversation to reconnect to active work. The completed report is saved under Recent.

Wheremore reuses the same retained operation when a browser connection is interrupted. It does not create another analysis simply because the page lost the start response or was reopened.

Understand the report

An open report has two views:

• Analysis contains the summary and prioritised Recommended Actions.
• Evidence contains the supporting analysis behind those conclusions.

The supporting analysis can include:

• key differences in the selected answers;
• visibility, rank and competitor findings;
• tracked-brand and competitor sentiment;
• brand positioning and perception;
• Tracked USP Visibility;
• citation rate and cited-page coverage; and
• Claims and evidence.

Each recommendation can explain what Wheremore observed, why it matters, the target, the specific change, supporting evidence and any information still required. Priorities are Immediate, Near term and Can wait.

The performance figures use all eligible answers in the selected scope. If the raw answer text is too large for one safe analysis request, Wheremore can select bounded examples for the AI input while retaining full-cohort headline metrics. Treat the narrative as a synthesis of the retained evidence and use the headline figures for the complete eligible cohort.

Understand claims and cited-page support

When published Claims Verification findings match the selected answers, Conversation Insight can show two separate judgements:

• Is this claim accurate? reports the saved public claim-verification result.
• Do the pages cited in the AI answer support it? reports whether the saved answer sources support that particular claim.

A claim can be accurate while its cited page does not support it, or a cited page can support a narrower statement while the wider claim remains only partly accurate. Not checked yet means there is no completed public verification result for that claim.

Tracked USPs are handled separately. The report can assess whether a USP is consistently present, inconsistently present, distorted or absent from the selected answers. A tracked USP is not evidence that a product claim is true.

If the Claims Context service is unavailable, the report does not turn tracked USPs into claim verdicts. Claims findings may be unavailable until that service is restored.

Send an action to the Action Board

Select Add to Board on a recommendation you want to manage as a task. Wheremore sends the recommendation with its signed analysis evidence and scope.

Adding an item to the Board does not publish content or make an external change. Review the recommendation, evidence, target and any required inputs before carrying it out.

Return to saved reports

Open Analyse with AI and select Recent. Wheremore keeps the five newest completed Conversation Insight reports for the selected configuration and conversation, including reports created with different filters, dates or research modes.

Each card shows its completion time, scope and whether it used Standard or Enhanced. A New badge is personal to each user and clears when that user opens the report.

From a saved report you can:

• switch between Analysis and Evidence;
• select Chat about this analysis;
• select Export report;
• add individual recommendations to the Action Board; or
• delete the saved report.

Chat about this analysis starts a chat with the saved report and its scope as context. A reopened report does not reconstruct expired raw monitoring evidence or another user's earlier chat.

Export report downloads a plain-text .txt  file containing the conversation, analysis scope and visible report text. Links are included beside their labels where available.

Deleting a report removes it from Recent. It does not delete Action Board items already created from that report.

If the analysis takes longer than expected

The browser checks progress for up to 12 minutes at a time. If the operation is still running, the page offers Resume analysis. Selecting it reconnects to the retained operation; it does not start another analysis.

Some provider timeouts can be retried automatically. During that recovery, the status says the analysis is taking longer than expected and Wheremore is trying again. A second timeout or an uncertain result can stop the operation for review rather than risk repeating paid work.

If completed analysis or research was saved but the final report could not be prepared, the interface may offer Retry analysis. The confirmation explains that saved work will be reused and that normal analysis credits apply when the final analysis runs.

If a report does not appear:

1. Check Recent.
2. Reopen the same conversation and select Resume analysis if it is offered.
3. Broaden the filters if Wheremore reports that no monitored answers match.
4. Add a tracked brand under AI Visibility Settings → Your Brands if setup is required.
5. Contact support if the retained operation says it needs review.

Do not repeatedly start another analysis for the same scope while an operation is active or retained for review.

Interpretation limits

• The report applies to one monitored conversation and the selected filters.
• The newest 30 eligible monitoring dates are the maximum analysis cohort.
• Recommendations are interpretations of saved answers and evidence; review them before acting.
• Mention rate, rank, sentiment and citation rate measure different things.
• A citation records an evidence link attached to an answer. It does not prove that every statement in the answer is supported.
• Public research can narrow uncertainty, but it cannot supply private product, customer or approval information.
• Provider answers, citations and sentiment can change on later monitoring dates.