AI Conversation Responses

AI Conversation Responses lets you inspect the answers returned by each monitored AI platform. Use it to understand whether your brand appears, how it is described, which competitors are recommended, and what sources support search-grounded answers.

The report is built from the questions and platforms selected in Manage AI Visibility Tracking. It is a response viewer, not a general test of everything an AI platform knows about your brand.

What the report shows

The report has two views:

Conversation list — browse monitored questions, filter the results and see how much response history is available.
Conversation detail — compare individual platform responses across monitoring dates and run an AI analysis of the current filtered history.

Results are grouped by AI Visibility configuration and then by monitored conversation.

Before you begin

You need:

an AI Visibility configuration;
at least one monitored conversation;
at least one selected AI platform; and
one or more completed monitoring runs.

Monitoring is processed in the background. A newly launched configuration may appear before its first responses are ready.

Quick start

Open AI Platform Conversation Responses.
Select the relevant Project and Configuration.
Start with Last 30 days and All Platforms.
Expand the required configuration.
Select a conversation.
Compare the answers returned on the latest monitoring date.
Review tracked-brand sentiment, competitors, Web Sources and Response Citations.
Use Load earlier responses to compare previous monitoring dates.
Select Analyse with AI when you want a summary of patterns across the current filtered history.

Filter the conversation list

The filters work together. A conversation must match all active filters to appear.

Project

Limits the report to configurations belonging to one accessible project.

Configuration

Limits the report to one audience and monitoring setup.

Configuration options respond to the selected project. Clear the project filter to see configurations across all accessible projects in the active workspace.

Phase

Filters conversations by buyer-journey phase:

Awareness — discovering or defining a problem.
Consideration — comparing approaches, products or providers.
Decision — selecting a provider or solution.

AI Platform

Limits results to one provider or provider mode. Supported labels can include:

ChatGPT;
ChatGPT Search;
Claude;
Claude Search;
Gemini;
Gemini Search;
Perplexity; and
Perplexity Search; and
Google AI Mode.

Standard and search-enabled versions are separate monitoring platforms. Search-enabled answers may use live web search and return grounding sources or citations.

Date range

Choose:

Yesterday;
Last 7 days;
Last 30 days;
Last 90 days;
Last 6 months;
All time; or
a custom start and end date.

Preset periods use UTC calendar dates. Last 7 days, for example, means today plus the previous six UTC dates. Custom start and end dates are both included.

Sentiment

Filters by the sentiment recorded for the configured tracked brand:

Positive;
Neutral;
Mixed;
Negative; or
Unknown.

The filter does not use an overall average of every company mentioned in the answer.

Unknown means Wheremore could not find a usable tracked-brand sentiment value in the structured response. It does not mean Neutral.

Has citations

Shows conversations with responses that contain one or more extracted citations.

Cites your domain

Shows conversations with a response that cites the configured website or one of its subdomains.

The match uses the configured domain, not a text search for the company name. Once a matching response is opened, Wheremore continues to display all its citations so the answer can be understood in context.

Understand the conversation list

Configurations are shown as expandable groups. Each conversation can show:

the monitored question;
its phase;
the number of platform responses;
the latest monitoring date; and
an indicator when the tracked domain is cited.

Select a conversation to open its full response history.

Response-count definitions

The report separates several measurements that can otherwise look similar:

Monitoring dates — distinct scheduled dates with matching saved results.
Platform responses — saved responses returned by individual monitored platforms across those dates.
Platforms represented — distinct AI platforms present in the matching history.
Web Sources — provider-supplied sources used to ground a search-enabled response.
Citations — extracted cited evidence attached to the response.

A single monitoring date can contain several platform responses. Retries do not intentionally create extra visible responses; the latest saved response for the same platform and monitoring date is used.

Review conversation detail

The detail view displays the original monitored question and its phase, followed by Response History.

History is grouped by monitoring date. Within each date, the report displays the responses returned by the matching platforms.

Each response can contain:

the platform name;
tracked-brand sentiment;
the answer text;
provider or competitor sentiment detail;
a Google search-grounding attribution block, where supplied;
Web Sources Used by Search; and
Response Citations.

The contents vary by platform. A standard model response may have no web evidence, while its search-enabled equivalent may show several sources.

Interpret tracked-brand sentiment

The sentiment badge describes the configured company or domain in that response.

Positive — the brand is clearly recommended or presented favourably.
Neutral — the brand is mentioned without a clear positive or negative judgement.
Mixed — the answer contains both favourable and limiting statements, trade-offs or caveats.
Negative — the brand is clearly criticised, rejected or presented unfavourably.
Unknown — no reliable tracked-brand sentiment was extracted.

Competitor sentiments can still appear within the detailed response. They do not control the tracked-brand badge or sentiment filter.

Read the underlying answer before acting on a badge. A positive mention can still be weak, conditional or limited to a narrow use case.

Web Sources and Response Citations

These sections are deliberately separate.

Web Sources Used by Search

These are sources the provider reports using for search grounding. They help explain where a search-enabled answer may have obtained its information.

Some providers return redirect or opaque source URLs. Where Wheremore has a usable HTTP or HTTPS address, it is clickable. An unresolved or unavailable source address can still be shown by domain or title without an unsafe link.

Response Citations

These are the citations extracted from the saved platform response. They may include a title, URL and supporting snippet.

A response can have Web Sources without separately extracted Response Citations, or Response Citations without a visible provider grounding block. A citation also does not prove that the cited page supports every statement in the answer.

Google search attribution

When Gemini Search supplies its own search-attribution block, Wheremore displays that block separately from the answer and source lists.

The block is isolated from the surrounding page and shown only when its saved payload passes safety checks. It may be absent when:

the provider did not return one;
the response was not search-grounded;
the provider changed the markup; or
the saved block failed validation.

The absence of the visual block does not necessarily mean the response has no Web Sources.

Google AI Mode sources can support particular answer sections rather than belonging neatly to a named provider. Wheremore therefore stores them as answer-level citations, with a provider relationship only when one is genuinely available. Do not interpret a source as endorsing your brand or a competitor merely because both occur in the same answer.

Load earlier response history

The newest monitoring dates load first.

Select Load earlier responses to request the next page of matching dates. This is server-side pagination: older history is not downloaded until requested.

Loading earlier responses does not delete, replace or truncate the newer responses already displayed.

Analyse a conversation with AI

Select Analyse with AI from the conversation detail view to identify patterns across the currently filtered history.

The analysis respects:

the selected configuration and conversation;
the selected AI platform, if one is active; and
the current date range.

To keep the request reliable and appropriately sized, the analysis uses at most the 30 newest matching monitoring dates. The response viewer itself remains fully paginated and can display older history.

The analysis runs asynchronously. Its status appears in the AI Analysis section. If it stalls, Wheremore can retry the processing request; if it fails, run the analysis again.

Treat the analysis as a synthesis of the saved evidence, not a replacement for reading important provider answers. Check any recommendation against the underlying responses and citations.

Where supported, useful recommendations can be added to the Action Board with their analysis evidence attached.

Recommended review workflow

Select one project and configuration.
Review all platforms over the last 30 days.
Compare the standard and search-enabled version of the same provider.
Check whether your brand appears in unbranded Awareness and Consideration questions.
Review Positive, Mixed, Negative and Unknown tracked-brand results.
Use Cites your domain to find responses grounded in your own website.
Inspect which competitor sites are used as Web Sources.
Compare the newest response with earlier monitoring dates.
Run AI Analysis to identify repeated patterns.
Use the relevant specialist content or search tool to correct weak evidence, missing coverage or unclear positioning.
Keep the configuration stable long enough to measure whether subsequent answers change.

How to interpret the evidence

A brand mention applies to that question, platform and monitoring date.
No mention does not prove the platform never knows about the brand.
A positive sentiment does not automatically mean the brand is the leading recommendation.
A citation to your domain does not prove the answer interpreted the page correctly.
A competitor citation can reveal which external evidence currently shapes the category.
Different platforms can reasonably return different answers because their models, search systems and sources differ.
Provider answers and citations can change between monitoring dates.

Deleted questions and historical responses

Removing a question from a monitoring configuration stops future scheduling and preserves completed records in storage.

The current response viewer does not show soft-deleted questions. An archived-history view and a separate permanent-delete control are planned improvements. Until that is available, review any history you need before removing a question.

Troubleshooting

No configurations appear

Confirm the active workspace.
Confirm you have access to the relevant project.
Clear the Project and Configuration filters.
Confirm the AI Visibility configuration has not been deleted.

A configuration appears but has no conversations

Clear Phase, Platform, Sentiment and citation filters.
Broaden the date range.
Confirm the configuration contains active questions.
Confirm monitoring has completed for at least one selected platform.

No matching responses

Start with All Platforms, All Sentiments and Last 30 days.
Check that the configuration was active on a selected processing day.
Allow the background provider and response-processing queues to finish.
Remember that a newly edited question starts building its own later history.

The report could not be loaded

Select Retry.
Check that the workspace has not changed in another tab.
Refresh the page if your session has expired.
If the problem continues, record the project, configuration, filters and time of the failure.

Cites your domain returns no results

Confirm the correct website is attached to the AI Visibility configuration.
Check whether the response actually cites the website rather than only mentioning the company name.
Try Has citations first.
Search providers may cite third-party sources without citing your own domain.

Sentiment is Unknown

The saved response did not contain a usable sentiment for the exact tracked brand identity. Read the answer directly and compare it with a later run. Unknown should not be interpreted as Neutral.

Web Sources have no clickable URL

The provider may have supplied an opaque redirect or an unresolved address. Wheremore only makes valid HTTP and HTTPS addresses clickable. The displayed domain can still identify the source used by the provider.

Google attribution is not shown

Confirm the result came from Gemini Search rather than standard Gemini.
Check whether Web Sources are still present.
The provider may not have returned an attribution block.
An unexpected or unsafe saved block is intentionally not rendered.

Response totals look different

Check which total you are comparing:

monitoring dates;
platform responses;
platforms represented;
Web Sources; or
Response Citations.

They measure different parts of the history and are not expected to be equal.

AI Analysis covers fewer dates than the viewer

The analysis is deliberately limited to the 30 newest matching monitoring dates. Use the response history to inspect older dates individually.

A removed question is no longer visible

The current viewer excludes soft-deleted conversations even though completed records can remain in storage. Archived access is a planned improvement.

What to do next

Use the evidence to choose a specialist next step:

improve missing or inaccurate topic coverage in the Content tools;
compare weak pages with stronger sources;
strengthen original evidence and clear brand positioning;
review internal links and page structure;
monitor search performance for the underlying queries; and
rerun the same stable AI Visibility configuration to measure the result.