Features

AI visibility

See whether AI answers name your brand for the prompts your buyers ask, checked every day on the engines your plan includes.

Core Feature

AI Visibility Monitoring

Every day, millions of people ask AI assistants for product recommendations, service comparisons, and brand information. DiscoveredBy runs scheduled scans through the Perplexity Sonar, Anthropic and xAI APIs, and collects Google AI Overviews, Google AI Mode, ChatGPT's app and Gemini's app answers through a licensed data provider, according to your plan. The API outputs can differ from what people see in the consumer apps.

You get a clear picture of when you're mentioned, what's being said about you, and how your visibility changes over time. No more guessing whether AI engines know your brand exists: you'll have the data.

The Overview screen for the Quillstone demo project over the last 90 days: the headline says Quillstone leads AI visibility among 5 brands, above five headline numbers over answers from seven engines, a Visibility trend with one line per brand from August to October, and the brand ranking with Quillstone first of five.
Fictional demo brand.
  • Brand mention tracking: see when collected outputs for your tracked prompts name your brand, with evidence from those runs.
  • Visibility and citation rate: the share of analysed answers that name your brand, and the share of answers that cite your site, from prompts run every day.
  • Prompt-level breakdown: see which specific questions trigger mentions of your brand and which ones don't.
  • Bring your list in bulk: import CSV or Excel (.xlsx) prompts, countries and tags, with a preview that shows exactly what will be created and what each row costs before anything is saved. Excel reads the first worksheet only.
  • Sort your portfolio by intent: assign or clear intent, buyer stage, theme and branding across selected prompts, manage tags in bulk, and pause or resume the selection. Each classification is from a fixed list, and each one filters the list. On Starter, Growth and Pro, use on-demand AI suggestions to review intent, buyer stage, theme and branding before saving. Branded and unbranded labels are editable and filterable. Prompts we suggest arrive already classified; the ones you write or import can be classified manually or through reviewed AI suggestions.
  • See which prompts people already search for: connect Search Console, map its property to your project, and each prompt shows the Google impressions behind the queries it covers, or a dash when none matches, scored against your own portfolio. It measures Google demand, not AI demand, and says so on the screen.
  • Multi-engine comparison: understand how your visibility differs across ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Grok and Claude.
  • Google AI Overviews and AI Mode: measured through a licensed data provider that fetches Google's own results for each prompt. See how often Google showed an AI answer at all, and where you rank in the organic results next to whether the AI Overview cited you.
  • ChatGPT: what chatgpt.com shows a visitor who is not signed in, measured through the same licensed data provider, and the only way we measure ChatGPT. ChatGPT decides for itself whether to search, so see how often its answers cite any source at all.
  • Gemini: what gemini.google.com shows a visitor who is not signed in, measured through the same licensed data provider. City targets run on it at the city's coordinates, and the Google product and Maps place links among its sources count as shopping and local businesses, not as citations.

Competitive Edge

Competitive Intelligence

Knowing your own visibility is only half the picture. On Starter, Growth and Pro and during the product trial, DiscoveredBy tracks your competitors across the same AI engines and the same prompts, so you can see your Share of Voice relative to the brands you compete with.

You don't have to know that list up front. Because we read the brand names in every answer, not just the links, DiscoveredBy suggests the competitors already appearing alongside you, including the ones you had never thought to track.

When a competitor starts appearing more often in AI answers, or a brand you have never tracked starts turning up in them, you'll know about it. This gives you the intelligence to respond quickly and stay ahead in the AI discovery channel.

Settings, Brands for the Quillstone demo project, scrolled to the Competitors section: 4 of 10 competitor slots in use and an Add competitor button, then the Tallyhook card with its domain, See insights, Rename, Pause and Remove, its names, its sub-brand Tallyhook Pro and its social accounts, and the top of the Bramblewick card.
Fictional demo brand.
  • Your competitors in one place: Settings → Brands has a card for each competitor you track, with its domain, a link to its Brand insights, Rename, Pause and Remove, and its social accounts, as shown above. Below the cards it suggests competitors: the companies named in at least two of your answers in the last 28 days that you do not track yet, most named first. Track one, or reject it so it is not suggested again.
  • Share of Voice: for each brand family, yours or a tracked competitor's, count how many answers named it, treating a brand and its own sub-brands as one count per answer, then compare that against every active tracked family's own count added together.
  • Side-by-side comparison: Overview's Brand ranking puts your visibility, share of voice and average position beside every active tracked competitor's, and Brand insights sets any one competitor beside you under every filter, with its trend, a breakdown by engine, tag, country, persona or theme, and the prompts where it beats you.
  • Prompt by prompt: each prompt's Brands tab ranks you, your tracked competitors and the other companies its answers named, and keeps your own row in view with its real rank.

Brand Families

Sub-brands and aliases

On Starter, Growth and Pro, nest a sub-brand one level under your own brand or a tracked competitor, such as a product line with its own name. A brand family counts once per answer even when it and a sub-brand are both named, and each active sub-brand also reports its own answers and sentiment, shown alongside its family rather than added to it.

Settings, Brands, for the Quillstone demo project: Quillstone's own sub-brand Quillstone Cloud with its alias QCloud and a note that your own social accounts are kept in Project settings, then the Competitors section with 4 of 10 competitor slots in use and an Add competitor button, and the competitor Tallyhook with its domain, See insights, Rename, Pause and Remove, and its sub-brand Tallyhook Pro with its alias THP.
Fictional demo brand.
  • Aliases on every plan: give your own brand its other names, such as a product name, an abbreviation, a common misspelling or a domain, and each one is handed to mention extraction as another name for that brand. A tracked competitor takes aliases the same way, on Starter, Growth and Pro and during the product trial.
  • From a discovered name: any suggested competitor on Settings → Brands can become an alias of your brand or a tracked competitor or, where sub-brands are available, a sub-brand, instead of a separate competitor.
  • History comes with it: adding a name moves the project's past untracked mentions with exactly that name onto it, and the app reports how many moved. No model is run again on old answers.

Audiences

Buyer personas

On Starter, Growth and Pro, name a buyer profile and run a tracked prompt as that persona, alongside or instead of the general, unbranded audience. Each location (a country or a city), audience and language combination uses one prompt slot. A persona run adds one audience block to the message sent to each chat engine, Perplexity included; Google AI Overviews, AI Mode, ChatGPT (app) and Gemini receive only the prompt text, so persona variants do not run there.

The Personas settings screen showing two active persona cards, Indie writer and Procurement lead, with their variant counts, their edit, archive and suggest-prompts controls, and the Add persona button.
Fictional demo brand.
  • Prompts a persona would ask: each persona can suggest up to 10 buyer-intent questions that person would plausibly ask, without naming your brand. They wait in Suggestions for your review and are never tracked automatically.
  • Report by audience: the persona filter narrows every screen that takes it, Sentiment trends and Brand reasons each have a By persona view, and persona is an Explorer dimension.
  • History stays as it was sent: each run stores the exact message the engine received, audience block included, so rewriting a persona changes only future runs.

Languages

Run a prompt in another language

On every plan, name one or more of 48 languages for a tracked prompt and each chat engine is asked to write its whole answer in that language, while Google AI Overviews, AI Mode, ChatGPT and Gemini receive the language as a setting of the request; the prompt text itself is never translated. Choose which engines run a prompt too, at no extra slot cost, and save a countries/audience/language/engine combination as a reusable template. A server-side preview shows the new slots and the monthly estimate before you save.

The Add prompt form with the United States and the city of London selected, As written and French both selected, all engines chosen, and the live preview sentence counting 2 locations.
Fictional demo brand.
  • Templates for many prompts: start a new prompt from a saved template, or apply one to a selection of tracked prompts and preview what it changes across all of them before anything is written.
  • Report by language: the language filter narrows every screen that takes it, and any Explorer metric can be broken out by language.
  • Where a language cannot run: a Chinese variant does not run on Google AI Overviews, AI Mode, ChatGPT or Gemini. It still runs on every chat engine, and the prompt's page lists where it was not run.

Cities

Track a prompt from a city

On every plan, add cities such as London or New York to a tracked prompt, picked from a list of 344 major cities across our 18 countries. Each city, audience and language combination uses one prompt slot, the same as a country. The Local view compares each city's brand visibility, citation rate, share of voice and position with the same country country-wide. A city is information we pass to the engine, not a real person's location, and each answer records how its location was sent.

The Local page for the last 90 days, with the United States and the United Kingdom expanded into their country-wide and city rows, gap pills on each city, and New York's prompts listed underneath.
Fictional demo brand.
  • A gap on every city: each city row shows the difference from the same country's country-wide answers, counted over the engines that answered for that city, so like is compared with like.
  • Where a city differs most: open a city to list up to 10 prompts where its brand visibility differs most from the same prompt country-wide.
  • How each engine gets the city: Perplexity, Claude and Grok receive it in a search context block with the prompt and natively; Google AI Overviews, AI Mode and Gemini receive its coordinates; ChatGPT takes no city, so a city target does not run on it.

Under the Hood

See what the AI actually searched for.

Before some AI engines answer your prompt, they search behind the scenes, expanding one question into several of their own before composing a reply. DiscoveredBy captures those hidden searches and, on Starter, Growth and Pro and during the product trial, sorts each one into a category, so you can see whether a competitor is winning because they show up in a comparison search, a shopping search, or a search that names a brand you never thought to track.

The fan-out view of a tracked prompt: the searches Claude ran before answering, each labelled with a category, and the pages each search returned, without saying which of its queries found which page.
Fictional demo brand.
  • Eight categories, one order: shopping, comparisons, named-brand searches, personalization, and more, judged in a fixed order so a search lands in exactly one of the eight, or is marked unclear rather than guessed.
  • A permanent record: what the engine actually searched for is written once and never rewritten, so a better classifier later never touches the underlying history.
  • From search to tracked prompt: on Starter, Growth and Pro and during the product trial, send any search you find interesting into your suggestions queue, so you can start tracking it directly.

Inside the Answer

What each answer held

On every plan, see how often each engine's answers used web search and held shopping, local businesses, ads, video, images or tables, from the structured parts of the Google AI Overviews, ChatGPT and Gemini responses and the searches each engine recorded; an engine that does not report a feature shows no value, never 0%.

The Answer features card on a prompt's page: per engine, the share of its answers with web search, shopping, local businesses, ads, video, images and tables, with Google AI Mode reading Reports no features we can count.
Demo data: fixed fictional responses run through the product's own collection and detection code.
  • Answer by answer: a prompt's run history marks each answer with the features it held, with a count for shopping, local businesses, ads, images and tables.
  • In the Explorer: seven Answer features metrics, from answers using web search to answers with tables, can be charted and broken down like any other.
  • What is not counted: anything written into the answer text, such as product links or images in the text, apart from Google's product-viewer links, Google Maps place links among Gemini's sources and the YouTube videos an AI Overview links.

Portfolio

Find prompts you could pause

On every plan, Prompt coverage compares the last 28 days of answers to your running prompts and recommends ones you could pause together while every source and brand they see regularly is still seen by a prompt you keep. Prompts that fact checks, open fixes or citation gap work depend on, and the last prompt for any country, city, persona, language or engine, are never recommended. Preview any selection, with the slots it frees and what it would stop observing, then pause it from the same screen. Nothing is paused automatically.

The Prompt coverage screen for the demo project: the window line and totals, the Prompts you could pause card, the first rows of the running prompts table with the recommended ones ticked, and the preview of that set beside it.
Fictional demo brand.
  • Two stated bars: a source or brand is regular for a prompt when it appears in at least 10% of that prompt's answers on at least 2 days, and still covered when a prompt you keep sees it in at least 2.5% of its answers.
  • New prompts are safe: a prompt is judged only once it has analysed answers on at least 14 days, so one you just added is never the first offered for pausing.
  • No model, nothing stored: every number is worked out from answers already collected when you open the page; only a pause you confirm changes anything.

Docs

Learn more in the docs

Prompts

What a prompt is, how it differs from a keyword, how the ones you track get run, your figures and the leading brand on each one, countries, cities and buyer-persona audiences, the Google results card (shown rate and organic rank next to Overview citations), bulk organization, reviewable AI classification and prompt branding, and the Google demand behind it.

Brand insights

One brand at a time, yours or a competitor, under every filter: its figures beside yours, its trend, a breakdown by engine, tag, country, persona or theme, and where it beats you.

Brands and sub-brands

Settings → Brands: your brand and competitors, suggested competitors, social accounts, what an alias is, and how a sub-brand rolls up into its family.

Personas

Named buyer audiences a tracked prompt can run as, how they use prompt slots, what changes at execution, and where their results show up.

Languages and templates

Run a tracked prompt in a language, what the response-language instruction says and where it goes for each chat engine, how the two Google engines, ChatGPT and Gemini receive a language instead, choosing which engines run, the server-side slot and answer preview, and saving a country/audience/language/engine combination as a reusable template.

Local

Compare the answers for prompts tracked in a city with the same country country-wide: brand visibility, citation rate, share of voice and position, the gap for each city, and the prompts where a city differs most.

Fan-out

What each engine actually searched for behind your prompt, and the eight categories those searches fall into.

Answer features

Which engines put shopping, local businesses, ads, video, images or tables inside their answers, and which answers used web search.

Prompt coverage

Which running prompts you could pause while everything they see regularly, the sources their answers cite and the brands they name, is still seen by a prompt you keep: the last 28 days of answers, protections for prompts other work depends on, a preview of any selection, and pausing it from the same screen.

Engines and measurement

The seven AI engines we measure, how we query them (the two Google engines, ChatGPT and Gemini through a licensed data provider), which engines run for your project, how a country or city reaches each one, and how each answer is labelled.

Which plan includes each engine and feature is on the pricing page.

Ready to see your AI visibility?

Start monitoring how AI search engines talk about your brand.

Start free

Start monitoring your AI visibility.

See how AI search engines talk about your brand.

Free to start. No credit card required.