> ## Documentation Index > Fetch the complete documentation index at: https://docs.peakmark.cc/llms.txt > Use this file to discover all available pages before exploring further. # Prompt-Level Evidence: Exact AI Answers Behind Scores > Peakmark stores the complete AI responses behind every metric so you can read what ChatGPT, Claude, Gemini, and Perplexity said about your brand. A visibility score tells you how often you're mentioned. Prompt evidence tells you everything else: what the AI actually said, how it framed your brand, which competitors it placed above or below you, and what sources it leaned on to reach that conclusion. Every percentage, every trend line, every share-of-voice figure in Peakmark is backed by the verbatim AI responses that produced it — and the prompt evidence view lets you read any of them, for any model, on any day. This is how GEO moves from intuition to evidence: you stop guessing what AI thinks about your brand and start reading it. ## What prompt evidence is When Peakmark runs your daily tracking, it doesn't just record a binary yes/no for each prompt. It stores the complete AI response — the full text the model returned, exactly as it was generated — alongside the prompt that triggered it, the model that answered, and the date it was captured. Every metric in your dashboard is derived from this corpus of stored responses, which means every metric is auditable down to the sentence level. Prompt evidence is what separates a GEO platform from a GEO score. A score tells you whether you're winning. The evidence tells you why, how, and what to change. ## What you can see in the prompt evidence view ### The exact prompt You see the precise wording of the question that was sent to the AI model — not a paraphrase, not a category label, but the actual text. This matters because slight variations in phrasing can produce meaningfully different answers, and understanding which prompt triggered which response helps you interpret the results accurately. ### The full AI response, verbatim The complete model output is stored and displayed without editing. You read exactly what ChatGPT, Claude, Gemini, or Perplexity said in response to that question on that day. ### Highlighted brand mentions Your brand and the competitors you track are highlighted inline within the AI response. You can see at a glance whether you appear, where you appear relative to competitors, and how many times you're referenced within a single answer. ### Citations and sources (where available) When a model includes citations — as Perplexity and certain ChatGPT modes do — those source links are captured alongside the response. You can see which web pages the model referenced when forming the answer that either did or didn't include your brand. ### Date and model metadata Every stored response is tagged with the date it was captured and the model that generated it, so you can filter, compare, and track changes across time and across models with full context. ## How to use the prompt evidence view ### Spot negative framing A visibility score can look healthy while the underlying framing works against you. Read the actual responses for prompts where you're mentioned and look for consistent qualifiers: "though it's more expensive than alternatives," "better suited for enterprise," "some users report a steep learning curve." If the same caveat appears repeatedly, you've found something specific to fix — and you know it in AI's own words, not a customer survey. ### Identify winning language The prompts where you're mentioned as a top recommendation contain something valuable: the exact phrasing and claims AI associates with your brand. Look for recurring themes — the features called out, the use cases highlighted, the comparisons where you come out ahead. These are the signals AI trusts, and amplifying them in your content and earned media is a direct lever on future visibility. ### Verify competitor claims If a competitor publicly claims strong AI visibility, the prompt evidence view shows you whether that claim holds up and what it actually looks like. You can see whether they're framed as a first pick or a footnote, which qualifiers follow their name, and whether that positioning is consistent across models or concentrated on one. ### Audit score changes When your visibility score shifts — up or down — go to the prompt evidence view and filter for the days around the change. You'll be able to read the responses that drove the movement and understand exactly what changed in how AI describes your category. This turns a graph inflection into an actionable insight rather than a mystery. Prompt evidence is the source of truth behind every metric in Peakmark. Every percentage you see in the dashboard has a full AI response backing it. ## Navigating prompt evidence Isolate responses from a single AI model to compare how ChatGPT and Gemini frame your brand differently for the same question. Select a specific date range to audit the responses behind a score change or examine the impact of a content or PR action. Show only prompts where you were mentioned, only where you weren't, or both — to study wins and losses separately. From any response with citations, navigate directly to the Sources & Citations view to see which pages the model referenced.