Worst Performing AI Conversations

Worst Conversations identifies the monitored questions where Your Brands appears least consistently across AI platform responses.


Use it to find gaps in brand visibility, compare weak performance between platforms and decide which underlying answers, content themes or positioning issues deserve closer investigation.


What the table measures

For each monitored question, Wheremore calculates:


latest monitored responses containing Your Brands ÷ all latest monitored responses × 100


The calculation uses responses matching the current dashboard filters.


The overall percentage combines every measured platform and monitoring date. Each platform row applies the same calculation to that platform only.


For example, if a question has 40 latest monitored responses and Your Brands appears in 8 of them, its overall visibility is 20%.


If ChatGPT supplied 10 of those responses and Your Brands appeared in 1, the ChatGPT row shows 10%.


Why these are called “worst” conversations

The table orders questions from the lowest overall visibility to the highest and displays the first ten.


“Worst” means lowest measured brand visibility under the selected filters. It does not automatically mean:


  • the AI answer is factually wrong;
  • the answer has negative sentiment;
  • competitors are recommended above your brand;
  • the answer lacks citations;
  • the question is commercially important; or
  • your website content caused the result.

Treat the table as a prioritised investigation list rather than a final diagnosis.


What counts as a brand appearance

Wheremore compares structured provider names in each response with the company and product names saved under Your Brands in AI Visibility Settings.


Matching is case-insensitive after spaces and formatting have been normalised. Use specific aliases that reflect the provider names AI platforms are likely to return.


The selected Provider Position determines the appearance required:


  • Mentioned Anywhere — Your Brands can appear at any valid provider rank.
  • Top 3 Recommendations — Your Brands must appear at rank 1, 2 or 3.
  • #1 Recommendation — Your Brands must appear at rank 1.

The denominator remains all latest monitored responses for the question. Provider Position changes what qualifies as a successful appearance.


A question can therefore have moderate visibility under Mentioned Anywhere but very low visibility under #1 Recommendation.


Latest responses and retries

Provider requests can be retried or reprocessed.


For each configuration, question, platform and monitoring date, Wheremore uses only the newest processed run. Older attempts do not add extra observations.


This prevents retries from inflating either the number of eligible responses or the number of brand appearances.


Before data appears

The selected project needs:


  1. An AI Visibility monitoring configuration.
  2. One or more monitoring questions.
  3. Selected AI platforms.
  4. Completed and processed monitoring responses.
  5. At least one company or product alias under Your Brands in AI Visibility Settings.
  6. Structured provider records for responses where brands or products are returned.

Legacy configurations can fall back to their saved company name when project-level Your Brands settings are absent. Configure Your Brands explicitly for consistent reporting across the dashboard.


Quick start

  1. Open the AI Visibility Dashboard.
  2. Select a project.
  3. Choose a date range.
  4. Leave Journey Phase, Configuration and Platform set to all values for the broadest view.
  5. Set Provider Position to Mentioned Anywhere.
  6. Open Worst Conversations.
  7. Review the overall visibility and evidence count for each question.
  8. Compare the platform rows to identify where the gap is concentrated.
  9. Compare Mentioned Anywhere with Top 3 and #1 Recommendation.
  10. Open AI Platform Conversation Responses to read the answers behind an important result.

Apply the dashboard filters

Project

Project selects the monitoring configurations, questions, responses and Your Brands settings used by the table.


Only projects available in the active workspace can be selected.


Date Range

Choose a preset period or a custom start and end date.


Dates are interpreted as UTC calendar dates and both custom boundaries are included. Last 30 Days means today and the preceding 29 UTC dates.


Journey Phase

Choose all phases or restrict the table to:


  • Awareness;
  • Consideration; or
  • Decision.

Use this to distinguish visibility gaps while buyers discover a problem, compare solutions or approach a decision.


Configuration

Choose all configurations or one monitoring configuration.


This can isolate an audience, market context, product or question set. Disabling a configuration stops future scheduling but does not remove its completed historical responses.


Platform

Choose all platforms or one platform.


Supported platform labels are:


  • ChatGPT;
  • ChatGPT Search;
  • Claude;
  • Claude Search;
  • Gemini;
  • Gemini Search;
  • Perplexity; and
  • Perplexity Search.

When one platform is selected, both the overall result and platform breakdown reflect that platform.


Provider Position

Use Mentioned Anywhere, Top 3 Recommendations or #1 Recommendation to change the required provider rank.


Start with Mentioned Anywhere to identify complete visibility gaps. Then compare Top 3 and #1 to find questions where the brand is mentioned but rarely recommended prominently.


Read each row

Each row contains:


  • the monitored question;
  • an overall visibility percentage with its brand-mention and eligible-response counts;
  • the monitoring configuration;
  • the customer journey phase; and
  • a visibility bar with percentage and counts for each represented AI platform.

The overall badge shows 10% overall alongside 1 of 10 responses.


Each measured platform row shows 10% alongside 1 of 10. The percentage is visually stronger, while the evidence count is smaller and muted.


The first count is the number of latest monitored responses that mentioned Your Brands and the second is the total eligible responses. Hover text retains the full explanation: “1 of 10 latest monitored responses mentioned Your Brands.”


The weakest measured platform percentage in a row is visually emphasised. This makes it easier to see whether low overall visibility is broad or concentrated on one platform.


Inline counts remain available on touch devices where hover text may not be accessible.


How conversations are ranked

The table returns up to ten questions.


Questions are ordered by:


  1. lowest overall visibility percentage;
  2. largest eligible-response count when percentages are tied; and
  3. conversation text alphabetically when both values are tied.

The second rule gives priority to a tied result supported by more evidence.


The overall percentage is response-weighted. It is not a simple average of the displayed platform percentages.


Understand 0% and “—”

These states have different meanings:


  • 0% — the platform produced eligible monitored responses, but Your Brands did not appear at the required provider position.
  • — — no monitored response was available for that platform under the selected filters.

Do not interpret “—” as poor visibility. It is unmeasured data.


A 0% result based on one response is less reliable than a 0% result based on many responses. Always inspect the underlying count.


How to investigate a weak question

  1. Confirm the result has a useful evidence base.
  2. Compare platforms to see whether the gap is widespread.
  3. Change Provider Position to determine whether the brand is absent or merely ranked lower.
  4. Read the underlying responses in AI Platform Conversation Responses.
  5. Check which competitors and alternatives appear.
  6. Review sentiment and citations separately.
  7. Check whether Your Brands aliases correctly match the returned provider names.
  8. Review the monitored question and audience context.
  9. Compare the question with related content, product pages and first-party evidence.
  10. Record the intended change and compare later monitoring periods.

Historical and archived questions

Completed responses remain available when a configuration is disabled.


If historical run data remains but the original question record is unavailable, the table can display Archived or unavailable conversation. The observation remains in the calculation rather than being silently discarded.


Permanently deleting historical response data is separate from disabling a configuration or removing a question from future monitoring.


How settings changes affect history

The table applies the current Your Brands aliases to stored structured provider records.


Adding, removing or correcting an alias can therefore change historical visibility percentages without changing the underlying responses.


Changes to questions, audience context, platforms or processing days affect future monitoring. Completed historical responses remain available within the relevant filters.


What the table does not measure

Worst Conversations does not directly measure:


  • competitor Share of Voice;
  • whether a competitor ranks above your brand;
  • sentiment;
  • citation frequency;
  • citations to your own domain;
  • answer accuracy or hallucination;
  • source authority;
  • search rankings, traffic or conversions;
  • the commercial importance of a question; or
  • causation.

It measures how consistently Your Brands appears in the latest monitored responses for each question.


What should I do next?

  • Open AI Platform Conversation Responses to read the underlying answers.
  • Use Competitive Win Rate to see whether selected competitors rank above your brand.
  • Use Share of Voice to compare your overall appearance share with competitors.
  • Use Conversation Sentiment Analysis to understand how the brand is presented.
  • Use Citation Rate Analysis to check whether brand appearances are supported by sources.
  • Use Key Item Tracker to identify recurring features, objections and positioning terms.
  • Review AI Visibility Settings if aliases are missing, too broad or incorrectly grouped.
  • Review Manage AI Visibility Tracking if questions, phases, platforms or processing days need changing.
  • Compare Worst Conversations with Top Conversations to identify themes associated with stronger visibility.

Data timing

The table is not real time.


Monitoring runs on each configuration’s selected processing days. Results appear after provider calls finish and raw responses have been processed into structured runs and provider appearances.


The newest monitoring date may be incomplete while jobs remain pending, running, saved as raw responses or waiting for processing.


Troubleshooting

Select a project appears

  • Select a project in the global controls.
  • Confirm the active workspace is correct.
  • Confirm you can access the project.
  • Allow the project selector to finish loading.
  • Reload the page if the table loaded before the shared controls.

Your Brands is not configured

  1. Select Configure on the AI Visibility Dashboard.
  2. Add the required company and product names under Your Brands.
  3. Save the settings.
  4. Reload the table if it does not refresh automatically.

No conversations found

  • Increase the date range.
  • Return Journey Phase, Configuration and Platform to their all-values options.
  • Set Provider Position to Mentioned Anywhere.
  • Confirm the configuration has completed monitoring runs.
  • Confirm Your Brands aliases match the structured provider names.
  • Check AI Platform Conversation Responses for completed answers in the same period.

A platform shows “—”

No eligible response was measured for that platform under the current filters. Confirm the platform is enabled and its jobs completed in the selected period.


A platform shows 0%

Responses were measured, but no structured provider matching Your Brands appeared at the required position. Try Mentioned Anywhere before concluding the brand was completely absent.


A question with a high percentage appears

The table always returns the lowest available results, up to ten. When every monitored question performs strongly, the least strong questions can still have high percentages.


The table differs from Competitive Win Rate

Worst Conversations checks whether Your Brands appears. Competitive Win Rate compares the position of Your Brands with selected competitors. A brand can appear frequently but still rank below a competitor.


The table differs from Share of Voice

Worst Conversations uses all eligible responses for each question as its denominator. Share of Voice compares brand appearances with appearances by selected competitors. The metrics answer different questions.


The percentage changed after editing Your Brands

This is expected. Current aliases are applied to historical structured provider data.


Conversation visibility could not be loaded

  • Select Retry.
  • Confirm the project still belongs to the active workspace.
  • Check that your session is still signed in.
  • Reload the page if the shared controls did not initialise.
  • Ask an administrator to inspect the runs and provider-data request if the error continues.
Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.