We Tested 100 Prompts on ChatGPT, Claude, and Perplexity — Here's What We Found About Brand Visibility

We ran an experiment. We took 100 prompts across three categories — category-level queries, feature-specific queries, and direct brand queries — and systematically tested them across ChatGPT, Claude (Anthropic), and Perplexity. Then we analyzed the patterns.

What we found illuminates exactly how AI decides which brands to surface, and more importantly, what kinds of brands consistently appear versus which ones consistently don't.

The Prompt Categories

We organized prompts into three buckets:

Category-level queries (e.g., "What's the best [category] tool for a startup?"): These are the highest-volume, most competitive prompts. Every brand in the category theoretically wants to appear here.

Feature-specific queries (e.g., "What's the best [category] tool with [specific feature]?"): These are mid-funnel, higher-intent queries. Buyers who know what they need are asking.

Sentiment and trust queries (e.g., "Is [brand] trustworthy?", "What are the downsides of [brand]?", "[Brand] vs [competitor]?"): These are late-funnel, decision-stage queries. The user is close to committing.

Finding 1: Category Queries Are Won by External Authority, Not Content Quality

For broad category queries, the brands that consistently appeared were not necessarily those with the best product pages or the most comprehensive website content. They were the brands most frequently mentioned across external sources — Reddit, industry publications, comparison sites, and review platforms.

Brand A, with an exceptional website but minimal external mentions, barely appeared. Brand B, with a mediocre website but extensive Reddit presence and 15 backlinks from DR 40+ publications, appeared in 70% of category queries.

The takeaway: For category queries, external citations dominate. Your own content is almost irrelevant if nobody else on the web mentions you.

Finding 2: Feature Queries Are Won by Structured Content

For feature-specific queries, the pattern reversed. Brands with clearly structured content that explicitly called out their features in extractable formats (feature tables, bulleted feature lists, explicit FAQ answers like "Does [product] integrate with Slack?") were significantly more likely to appear.

Brands that described their features only in paragraph prose, with no structured extraction points, were frequently skipped — even when the feature existed and was technically superior.

Finding 3: Sentiment Queries Surface Reddit and Review Sites Almost Exclusively

For trust and sentiment queries, AI drew almost exclusively from Reddit, G2/Capterra reviews, and industry media coverage. No brand's own website appeared as a source for sentiment-type responses.

Brands with active, positive Reddit communities were described in warm, specific terms. Brands without Reddit presence were either not mentioned or described in generic, non-differentiating language.

Finding 4: Claude and ChatGPT Show Different Biases

ChatGPT showed a stronger preference for Wikipedia and long-form content from established publications. Claude (Anthropic's model) showed more willingness to synthesize from a broader range of sources, including smaller but well-structured websites. Perplexity consistently surfaced Reddit content more heavily than the other two.

This means your GEO strategy should not be designed exclusively for one platform. Multi-platform presence matters.

Finding 5: Negative Mentions Are Sticky

Brands with significant negative Reddit threads or critical reviews on G2 carried that negative framing into AI responses. When asked "what are the downsides of [brand]?", AI would often reference specific Reddit complaints almost verbatim.

Negative information is sticky and difficult to dislodge once it's embedded in AI training data. The implication: don't ignore sentiment issues. Address them proactively, before they become part of how AI describes you.

What This Means for Your Brand

These five findings map directly onto the dimensions that Show Your Brand measures in its GIO audit. We send 100 prompts — across all three query categories, across multiple AI platforms — and report on exactly these patterns for your specific brand.

Are you winning category queries? Feature queries? How does AI describe your sentiment? Are negative mentions surfacing? How do you compare to your top two competitors across all three query types?

You can know all of this within days of booking your audit.