Understanding chats

Chats are the AI-generated responses we produce by running your prompts daily across different platforms. Understanding chat anatomy is crucial because these responses are the basis for all dashboard metrics, source data, and competitive analysis. Every visibility score, position ranking, and source classification starts with an analysis of these individual conversations.

​

Anatomy of a chat

You’ll find recent chats on your Overview dashboard, but you can view the last 100 chats for each prompt by clicking on the individual prompt. Each chat contains specific elements that we analyze to generate your metrics and insights:

• Status: Shows if the prompt ran successfully.
• Model: Which AI platform generated the response (ChatGPT, Claude, Perplexity, etc.).
• Location: Geographic location we prompted from (e.g., United States).
• Sentiment: How the AI model talks about your brand — positively, neutrally, or negatively (measured as a score between 0–100).

Each chat response contains:

• Main response: The actual AI-generated answer to your prompt.
• Brands mentioned: Shows which brands (including yours) were mentioned and where in the response.
• Fanout Queries: Shows the query fanouts performed by the model when generating the answer for a specific prompt.
• Sources sidebar: All URLs the AI model referenced or cited when creating the response.

​

Sources vs citations

Not all sources are citations, but every citation is a source. Understanding this distinction helps you interpret your data correctly.

• Citations: Sources explicitly referenced within the AI response text. These appear as direct mentions in-line in the response body. Citations indicate the AI used specific information from that URL to create particular sentences or sections.
• Sources: All URLs the AI model accessed during response generation. This includes citations plus additional sources the AI considered but didn’t explicitly reference. Sources appear in the sidebar and often at the bottom of a chat, even if not directly cited in the response.

Example: An AI response might cite 5 sources directly in the text but show 8 sources in the sidebar. All 8 contributed to the response, but only 5 were explicitly referenced. When a URL is cited multiple times, it will appear only once in the sidebar. Citations drive traffic to your website, while sources without citations still build your authority and influence the AI’s understanding. Both contribute to your AI visibility, but they require different optimization strategies. Track both metrics to get the complete picture of your performance.

​

How AI platforms behave

Different AI models handle sources and citations differently, which affects your data patterns:

• ChatGPT: Sometimes performs web searches and sometimes doesn’t. When ChatGPT doesn’t search the web, you’ll see responses with no sources listed. This is normal behavior, not a data issue.
• Perplexity: Shows many sources in the sidebar but tends to cite fewer of them directly in the response text. You’ll often see higher source counts but lower citation numbers.
• Claude and other models: Each has unique patterns for how they search, cite, and reference sources.

AI models have inherent randomness in source selection. A source might appear one day and be completely absent the next. Focus on trends over weeks rather than daily changes.

​

Reading chat position rankings

When multiple brands appear in a chat, we calculate position based on mention order — including all detected brands, not just your tracked competitors. Example: If a chat mentions Hyundai (1st), Chevrolet (2nd), BMW (3rd), BMW ranks in position 3. If tomorrow’s chat mentions Hyundai, Chevrolet, Ferrari, BMW — even though you haven’t added Ferrari as a competitor, BMW’s position becomes 4th.A higher position (closer to 1) indicates that AI models consider your brand to be:

• The most authoritative source for the topic.
• The go-to reference in your industry.
• Highly relevant to the user’s query.

This gives you a true competitive context by showing where you rank among all the mentioned brands, not just your selected competitors.

​

Chat Features

Chat features are structured flags attached to every chat, showing you at a glance what type of content an AI model included in its response. Whether a chat contains ads, maps, shopping results, or a live web search. Every time Peec runs a prompt, the AI model’s response can contain more than just text. It might serve a sponsored ad, pull up a local business map, or run a live web search before answering. These extras are what we call features, and they matter because they tell you what kind of result your prompt is actually triggering. For example: a prompt-driven ad placement means your topic is already being monetized by AI. A prompt that pulls maps means the AI is treating it as a local discovery query.

​

Available features

Each chat can carry any combination of the following features:

• Ads: The AI model included a sponsored placement in the response. Relevant for understanding monetization pressure on your topics.
• Maps: The AI returned a local business map result. Common for location-based or “near me” style queries.
• Web search: The AI performed a live web search to generate the response, rather than answering from memory.
• Shopping: The AI included product or shopping results
• Product Comparison: Chats in which the AI makes comparisons between two or more brands

​

Where to find features

Features appear across the platform in two places:

• All Chats table: Each chat row displays its features as visual flags. Use the dropdown filter at the top of the table to show only chats matching specific feature types (e.g., only chats that included ads, or only chats where web search was used)
• Prompts and Topics tables: Each prompt and topic now shows aggregated feature percentages. For example, you might see that 42% of chats for a given prompt included ads, or that 80% triggered a web search. This helps you quickly identify which prompts drive which response types without opening individual chats.

​

Filtering by features

In the All Chats table, click Filter, then select any feature to narrow the list. A few ways to use this:

• Filter to Ads to see every response where your brand appeared alongside paid placements.
• Filter to Maps to identify which prompts are driving local discovery results.
• Filter to Web search to focus on chats where the AI actively looked things up — these tend to have more recent, source-driven content.
• Combine multiple filters to find chats with overlapping features, such as shopping and web search.