How to Read Your AI Visibility Report and Decide What to Fix First

The AI Visibility Report is the longest screen in ApexaRank.ai, and the one most people read in the wrong order. They open it, look at the number at the top, feel briefly good or briefly bad about it, and close the tab.

The number at the top is the least useful thing on the page. It is a summary of several different measurements, and summaries are where the actionable detail goes to hide.

This guide reads the report the way it is meant to be read — from the sections that tell you what to do, not the one that tells you how you feel.

What the Report Is Actually Measuring

Every section of the report is built from the same underlying exercise: real questions from your category are put to the major AI platforms, and the answers are checked for whether your brand appears, how it is described, and who appears instead of you.

That produces four distinct measurements, which the report keeps deliberately separate:

Confusing these four is the single most common misreading. A brand can be present everywhere and still lose, because a competitor is named twice as often. Another can have a modest score but be described so well that the mentions convert.

ChatGPTGeminiClaudePerplexityGoogle AI Overviews

Every section of the report is measured separately for each of these platforms — the differences between them are usually more useful than the average across them.

Read It in This Order

The report is laid out top to bottom for orientation, not for decision-making. This is the order that produces an action:

The Overall Score, and Its Three Bands

At the top of the report sits the Overall AI Visibility Score out of 100, alongside the domain it belongs to and the analysis run it came from.

The score is colour-banded, and the bands are what you should read:

Visibility by AI Platform

Below the score, the report breaks the same measurement out across ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude and Microsoft Copilot. Each gets its own card with its own score.

Read this row for the spread, not the average. Three patterns show up repeatedly, and each points somewhere different:

Whatever is happening is happening to your content generally. If the scores are evenly low, the problem is the content or the site, not any one platform's preferences.

Usually a freshness-versus-authority split. Perplexity rewards recent, explicitly citable pages; ChatGPT leans on established, widely referenced sources. This pattern says your content is good but your brand is not yet a known entity.

AI Overviews draw heavily on what already ranks. Being weak there while doing fine elsewhere usually points at an ordinary SEO problem rather than an AI one.

Opening a platform card gives you its Summary and a Key Insight — a plain-language explanation of why that platform is or is not citing you.

Testing a Prompt Yourself

Inside each platform view there is a Run Custom Prompt box, with the output shown under Prompt Test Results. This is the most underused control in the product.

Use it when you disagree with the report, or when you want to check a question nobody tracked. The single most valuable prompt to run is the one your sales team hears most often — typed exactly as a customer would say it, not as you would write it.

Share of Voice

The Share of Voice section answers a different question from the score, and it is the one most likely to change your plans.

The section splits three ways. Overall Share of Voice across all platforms, by Platform, and by Topic — with a YOU LEAD marker on the topics where you are the most-cited brand.

Read "by Topic" first. It tells you which parts of your category you already own, and those are the areas where more content compounds fastest.

Brand Sentiment

The Brand Sentiment Analysis section reports an overall positive / neutral / negative split, then two lists that are worth more than the split itself:

Concerns Mentioned is the most directly actionable list in the whole report. Each item is an objection a buyer is being handed before they ever reach your site. If AI keeps saying you are expensive, or that setup is complex, or that you lack an integration, those are content briefs — a pricing explainer, an onboarding page, an integrations directory.

Neutral sentiment is not a failure. Assistants describe most brands neutrally by default. Sustained negative language, though, usually traces back to a specific review site or old article the models keep reading.

Historical Trends

Once you have more than one run, the Historical Trends section shows AI Visibility Score Over Time and Platform Score Trends, with four summary figures: Score Change, Citations Gained, Citations Lost and New Platforms.

Citations Lost deserves more attention than Score Change. A stable score can hide real churn underneath — twelve citations gained and eleven lost is not the same as a quiet month, even though the number barely moves.

Auto-Discovered Competitors

On the dashboard, below your sites, ApexaRank surfaces Auto-Discovered Competitors — marked AI-Detected, because they are not the competitors you named. They are the domains competing for the same citations, found through query overlap.

Each row shows the domain, how many shared queries you have, an overlap percentage, and a threat tier of high, medium or low.

Expect at least one surprise here, and take it seriously.

The list regularly includes review sites, marketplaces, industry publications and adjacent products — sources that answer your buyer's question without selling what you sell. They are not competitors in your market, but they are competitors for the answer, and that is what is being measured.

Deciding What to Fix First

The report ends with a Know what to fix next prompt for good reason. Here is how to get there from what you have just read:

Do not start with content. Start with whether AI can use your site at all — accessibility, structure and entity signals. The report's own research notes make the point plainly: URL accessibility is the single biggest citation factor, and brand mentions carry roughly three times the weight of backlinks.

Fix for that platform's bias rather than rewriting everything. Weak on Perplexity usually means stale content. Weak on AI Overviews usually means weak rankings. Weak on ChatGPT usually means weak entity presence off your own site.

Take the topics where a competitor leads and you already have pages. Strengthening an existing page that nearly qualifies beats starting a new one from nothing.

Answer it directly, on its own page, in the words the concern is phrased in. This is the fastest-moving fix in the report because the question is already being asked.

Common Mistakes

The score is an index. It tells you whether things are moving, not what to do. Every instruction in the report lives in the sections underneath it.

Refreshing a report you have not acted on produces a slightly different number and no information. Change a page, then re-run.

A review site taking your citations is costing you the same customers as a direct rival would — and is usually far easier to displace, because you know your product better than they do.

Six platform scores averaged into one number throws away exactly the detail that tells you where to work.

Frequently Asked Questions