# refd > Auditable AI search monitoring for businesses. See where a brand appears in AI answers, who appears instead, and the raw answer behind every metric. refd tracks buyer questions across ChatGPT, Perplexity, Gemini, Google AI Mode, and Google AI Overviews. It measures mentions, citations, first-mention position, sentiment, prominence, and share of voice while preserving the underlying answer evidence. Mentions and citations are independent signals, and a missing Google AI Overview is a valid observation rather than a failed fetch. The hosted product runs at refd.ai. The complete stack is MIT licensed and can be self-hosted with separate infrastructure and collection accounts. Every link below points to a markdown version of a page. Each one is served at the page's own URL with `.md` appended, so dropping the suffix gives the HTML original. ## Product - [Homepage](https://refd.ai/index.md): What refd measures, how it collects, and who it is for. - [Interactive demo](https://refd.ai/demo.md): A no-signup sample visibility report where every metric links back to the answer it was scored from. - [Agent access](https://refd.ai/agents.md): The read-only MCP endpoint, its OAuth model, the nine tools, and how to connect a client. ## Guides - [How refd measures AI search visibility](https://refd.ai/methodology.md): How refd collects repeated AI answers, detects mentions and citations, and calculates position, sentiment, and share of voice with auditable evidence. - [AI visibility tools compared, on facts you can check yourself](https://refd.ai/compare/ai-visibility-tools.md): Twelve AI visibility platforms compared on surfaces tracked, published pricing, raw answer evidence, open source status, and API access. Checked 5 September 2026. - [AI visibility checker: how to check whether AI answers mention your brand](https://refd.ai/ai-visibility-checker.md): There is no honest one-click AI visibility score. Here is what a real check requires, a free manual method you can run in about thirty minutes, and what it still cannot tell you. ## Documentation - [Documentation](https://refd.ai/docs.md): Setup and operating guides for hosted and self-hosted deployments. - [Getting started with refd](https://refd.ai/docs/getting-started.md): Create a workspace, define your brand and competitors, choose buyer questions and AI surfaces, and inspect the evidence in your first refd report. ## Surface trackers - [AI surface trackers](https://refd.ai/trackers.md): One measurement contract applied separately to each AI answer surface. - [ChatGPT visibility tracker with answer evidence](https://refd.ai/chatgpt-visibility-tracker.md): Track when ChatGPT names or cites your brand, which competitors appear first, and the answer evidence behind every visibility metric. - [Perplexity visibility and citation tracker](https://refd.ai/perplexity-visibility-tracker.md): Track brand mentions, cited domains, competitor position, and the answer evidence returned for your buyer questions in Perplexity. - [Gemini visibility tracker with repeated answers](https://refd.ai/gemini-visibility-tracker.md): Measure Gemini brand mentions, citations, competitor position, sentiment, and answer evidence across a stable set of buyer questions. - [Google AI Mode tracker with source evidence](https://refd.ai/google-ai-mode-tracker.md): Track brand and competitor visibility in Google AI Mode across scheduled buyer questions, cited sources, position, and sentiment. - [Google AI Overview tracker with coverage evidence](https://refd.ai/google-ai-overview-tracker.md): Track when Google shows an AI Overview, which brands and sources appear, and the answer evidence behind mentions, citations, and position. ## Research and analysis - [Research and guides](https://refd.ai/blog.md): Published analysis and field notes on measuring AI search visibility. - [What is AI search monitoring? Metrics, methods, and limits](https://refd.ai/blog/what-is-ai-search-monitoring.md): AI search monitoring measures whether AI assistants mention, cite, and recommend your brand. Here is what it measures, how it works, and where it stops. - [AI mentions and AI citations are not the same metric](https://refd.ai/blog/ai-mentions-vs-citations.md): A brand can be named in an AI answer without being cited, or cited without being named. The four combinations require four different responses. - [Why one AI answer is not a reliable visibility measurement](https://refd.ai/blog/one-answer-is-not-a-measurement.md): AI answers change between runs. Here is what a single answer can and cannot support, and the comparison rules that stop noise being reported as change. - [How to get cited in AI Overviews and AI answers](https://refd.ai/blog/how-to-rank-in-ai-overviews.md): What two live AI Overviews actually cited on 4 September 2026, what that implies about getting included, and how to measure whether any of it worked. - [How to build a prompt set for AI search monitoring](https://refd.ai/blog/how-to-build-an-ai-prompt-set.md): Your prompt set is the measurement instrument. Choose it badly and every number downstream describes a market you do not sell into. - [How to audit whether an AI visibility tool's numbers are real](https://refd.ai/blog/how-to-audit-an-ai-visibility-tool.md): Every tool in this category reports confident percentages. Here are the checks that separate a measurement from a claim, in about twenty minutes. - [AI visibility and SEO rankings measure different things](https://refd.ai/blog/ai-visibility-vs-seo-rankings.md): A page at position one can be absent from the AI answer above it. Understanding why changes what you measure and what you do about it. - [How to report AI visibility to your leadership team](https://refd.ai/blog/reporting-ai-visibility-to-leadership.md): The numbers in this category are new, non-deterministic, and easy to challenge. Here is how to present them so they survive the first hard question. - [How ChatGPT decides which brands to name, and how to track it](https://refd.ai/blog/tracking-brand-visibility-in-chatgpt.md): ChatGPT answers from model memory, from web search, or from both. Each path changes what a brand mention means and what you can do about it. - [Monitoring citations in Perplexity without overcounting them](https://refd.ai/blog/monitoring-citations-in-perplexity.md): Perplexity shows its sources, which makes it the best surface for citation tracking and the easiest one to measure wrong. - [Measuring brand mentions in Gemini, and why it is not Google Search](https://refd.ai/blog/measuring-gemini-brand-mentions.md): Gemini, Google AI Mode, and Google AI Overviews are three different products from one company. Treating them as one Google number loses the distinction that matters. - [Google AI Mode and AI Overviews are not the same surface](https://refd.ai/blog/google-ai-mode-vs-ai-overviews.md): Two Google products, two collection methods, two user postures, and one distinction that most AI visibility reporting collapses into a single number. - [Why AI surfaces disagree about your brand](https://refd.ai/blog/why-ai-surfaces-disagree.md): The same question, asked across five AI answer surfaces, routinely produces five different sets of recommended companies. That disagreement is usable information. ## Glossary - [Glossary](https://refd.ai/glossary.md): Every metric and term refd measures, read from the same definitions the product reads, plus the category vocabulary of AI search visibility. - [Mention rate](https://refd.ai/glossary/mention-rate.md): The percentage of scored AI answers that name the tracked brand or competitor. - [Average position](https://refd.ai/glossary/average-position.md): The average order in which the brand appears when it is mentioned. Position 1 means it was named first. - [Prompt result signals](https://refd.ai/glossary/prompt-result-signals.md): M is mention rate and C is citation rate for each AI surface. None means the answer was scored but contained neither signal. - [Share of voice](https://refd.ai/glossary/share-of-voice.md): Of all tracked brand mentions, the percentage that belongs to this brand or competitor. - [Share of voice gap](https://refd.ai/glossary/share-of-voice-gap.md): Your brand's share of voice minus the strongest competitor's. A positive value means your brand is ahead. - [First named](https://refd.ai/glossary/first-named.md): When a tracked brand is named first, the percentage of those answers led by this brand. - [Surface leaders](https://refd.ai/glossary/surface-leaders.md): The brand or competitor with the highest mention rate on each AI surface in the selected range. - [Citation rate](https://refd.ai/glossary/citation-rate.md): The percentage of scored AI answers that cite a page from this brand or competitor. - [Attributed citations](https://refd.ai/glossary/attributed-citations.md): Citations that can be assigned to a recognizable domain. Opaque redirects are excluded. - [Brand URL citations](https://refd.ai/glossary/brand-url-citations.md): The total number of times an exact page from your domain appeared as a cited source. - [Source gap](https://refd.ai/glossary/source-gap.md): Domains cited in answers where your brand was neither mentioned nor cited. - [Answers citing](https://refd.ai/glossary/answers-citing.md): The number of distinct AI answers that cite at least one page from this domain. - [Citations](https://refd.ai/glossary/citations.md): The total cited URLs from this domain. One answer can contain more than one citation. - [Sentiment](https://refd.ai/glossary/sentiment.md): How AI answers portray this brand or competitor when it is mentioned, as positive, neutral, and negative shares of classified mentions. - [Positive sentiment](https://refd.ai/glossary/positive-sentiment.md): Of the answers that mention the brand, the percentage whose portrayal is classified as positive. - [Material change](https://refd.ai/glossary/material-change.md): A movement large enough to outrank routine noise, measured between seven-day windows of runs rather than between two single runs: 5 points for mention and citation rates, 4 points for share of voice, 5 points for sentiment shares, and a quarter of a rank for average position. - [Workspace](https://refd.ai/glossary/workspace.md): An isolated monitoring project for one brand and its competitive landscape. - [Brand](https://refd.ai/glossary/brand.md): The primary entity a workspace is created to monitor. - [Competitor](https://refd.ai/glossary/competitor.md): A company or product tracked alongside the workspace brand for comparison. - [Tracked entity](https://refd.ai/glossary/tracked-entity.md): A brand or competitor that refd looks for in AI answers and citations. - [Prompt](https://refd.ai/glossary/prompt.md): A question sent to each enabled AI surface during a collection run. - [Domain](https://refd.ai/glossary/domain.md): A website address assigned to a tracked entity for citation attribution. - [Alias](https://refd.ai/glossary/alias.md): An alternate name that should count as a mention of a tracked entity. - [AI surface](https://refd.ai/glossary/ai-surface.md): An AI answer experience monitored as a separate data source. - [AI answer](https://refd.ai/glossary/ai-answer.md): The response text and source list returned for one prompt on one AI surface in a run. - [Mention](https://refd.ai/glossary/mention.md): An occurrence of a tracked entity name or alias in the AI answer text. - [Citation](https://refd.ai/glossary/citation.md): A valid web source URL associated with an AI answer. - [Source](https://refd.ai/glossary/source.md): A web page cited as supporting material for an AI answer. - [Cited domain](https://refd.ai/glossary/cited-domain.md): The website hostname extracted from one or more cited source URLs. - [No AI Overview](https://refd.ai/glossary/no-ai-overview.md): A valid Google search result in which Google did not show an AI Overview. - [Run](https://refd.ai/glossary/run.md): A collection batch covering the active prompts and enabled AI surfaces at a point in time. - [Run trigger](https://refd.ai/glossary/run-trigger.md): The event that started a collection run. - [Collection unit](https://refd.ai/glossary/collection-unit.md): One result for a specific prompt and AI surface within a run. - [Scoring](https://refd.ai/glossary/scoring.md): The process that turns a collected answer into entity and citation signals. - [Raw answer payload](https://refd.ai/glossary/raw-answer-payload.md): The original provider response retained behind a scored collection unit. - [Answer engine](https://refd.ai/glossary/answer-engine.md): A search product that returns a written answer instead of a list of links. - [Google AI Overview](https://refd.ai/glossary/google-ai-overview.md): The AI-generated summary Google places above the classic result list for some queries. - [Google AI Mode](https://refd.ai/glossary/google-ai-mode.md): Google's dedicated conversational search surface, separate from AI Overviews. - [Zero-click search](https://refd.ai/glossary/zero-click-search.md): A search that ends without the user visiting any result, because the answer was given on the results page. - [Retrieval-augmented generation](https://refd.ai/glossary/retrieval-augmented-generation.md): Fetching documents at question time and giving them to a language model to write its answer from. - [Grounding](https://refd.ai/glossary/grounding.md): Tying an answer to retrieved sources so its claims can be traced rather than invented. - [Answer variance](https://refd.ai/glossary/answer-variance.md): The tendency of an answer engine to give different answers to the same question. - [Hallucination](https://refd.ai/glossary/hallucination.md): A confident claim in an AI answer that no source supports. - [Answer engine optimization](https://refd.ai/glossary/answer-engine-optimization.md): The practice of trying to influence whether and how a brand appears in AI-generated answers. - [Generative engine optimization](https://refd.ai/glossary/generative-engine-optimization.md): A near-synonym for answer engine optimization, usually abbreviated GEO. - [llms.txt](https://refd.ai/glossary/llms-txt.md): A proposed file at a site's root that lists its documents in a form built for language models to read. ## Project - [Security at refd](https://refd.ai/security.md): How refd protects hosted accounts, isolates workspace data, stores monitoring evidence, authorizes agent access, and handles vulnerability reports. - [Support](https://refd.ai/support.md): How to get help with hosted refd, account access, monitoring results, MCP connections, privacy requests, and self-hosted deployments. - [Open-source refd](https://refd.ai/open-source.md): How refd's MIT-licensed code, auditable metrics, three-Worker Cloudflare architecture, hosted service, and contribution path fit together. - [Source code](https://github.com/emaitchess/refd): The MIT-licensed implementation of the whole stack. - [Design system](https://github.com/emaitchess/refd/blob/main/docs/DESIGN.md): The public UI and charting rules the dashboard follows. ## Account - [Create an account](https://dash.refd.ai/auth/create-account): Start a hosted workspace. - [Sign in](https://dash.refd.ai/auth/sign-in): Return to an existing workspace. ## Optional - [Full text of every public page](https://refd.ai/llms-full.txt): The same documents listed above, expanded inline in one file. - [Privacy policy](https://refd.ai/privacy.md): How the hosted refd service collects, uses, shares, stores, and deletes account, workspace, analytics, and connector data. - [Terms of service](https://refd.ai/terms.md): The terms that govern access to refd's hosted website, dashboard, API, monitoring service, and remote MCP connector. - [RSS](https://refd.ai/rss.xml): Subscribe to published research and documentation. - [Issues](https://github.com/emaitchess/refd/issues): Report a bug or request a feature.