Early accessEvery number comes with its margin of errorMargin of error built in →

See what AI says about you. Prove what changed it.

For brands, SaaS teams and agencies. Massyl asks up to six AI engines your buyers' questions twenty times a month, shows how often you are named with an honest range, and proves which of your changes actually moved it.

Get a free starter kit for your site instantly: your buyers' questions, an llms.txt file, schema and a checklist. The full 20-run report arrives at launch.

ChatGPTGoogle AI OverviewsGoogle AI ModePerplexityGeminiClaudeGrok, DeepSeek, Meta AI, Mistral next
Halden Coffee · Example checkExample
Run 15 of 20
best place to buy espresso beans in austin?
Latest runs
    Mention rate, ChatGPT
    40%
    range 20–64%, n = 15
    0%50%100%
    Named in the same answers
    Cuvée
    65%
    Halden
    40%
    Fleet
    30%
    The product

    Every number you can trust, and check.

    One weekly report across up to six engines. Each rate comes with its range, and each range opens the answers that produced it.

    Example data for a fictional business.

    6AI engines, 4 on every plan
    20×runs per question, per engine
    95%range on every number
    Anylanguage, written natively
    Before and after

    How most teams check AI today, and with Massyl.

    Screenshots in a spreadsheet and a Slack thread asking why a competitor shows up. One answer per check, no way to tell if anything changed.

    Today
    With Massyl
    ChatGPT · espresso beans austinweek 46
    40%8 of 20 answers · range 22–61%
    Why Cuvée wins: 13 of 20 answers cite cuveecoffee.com/roast-schedule. Your site has no roast-day page.
    Review draft fixOpen 20 answers
    Measured, not guessed

    One answer is a coin toss. Twenty is a measurement.

    AI engines answer the same question differently every time. Most trackers ask once and report what came back. Massyl asks twenty times on every engine, so each rate comes with a range, and you can tell a real change from noise.

    • 20 runs per question, per engine, every month
    • 95% range on every rate, so a "+9 points" means something
    • "Not enough data" instead of a confident number when the range is too wide

    Try it: a business that is truly named 40% of the time

    Asked once
    100%
    Recent checks: 100%

    Named this time. Next week the same check may say 0%.

    Asked twenty times
    40%
    range 22–61%, n = 20

    Named in 8 of 20 answers. The true rate is very likely between 22% and 61%.

    Proof on every number

    Click any rate and read the answers behind it.

    Every answer is stored with its date, engine, country and language, and has its own link you can send to a client. The words that name you, or a competitor instead, are highlighted. When AI gets a fact wrong, you see its words next to yours.

    • Every answer kept, dated, with a shareable link
    • Wrong facts flagged: hours, prices, locations, features
    • How it was asked: app or API, shown on each answer
    Wrong fact · opening hours7 of 20 ChatGPT answers
    ChatGPT says

    “Halden is open daily until 6 pm, so plan an early visit.”

    Your site says

    Open daily, 7 am to 8 pm.
    haldencoffee.com/visit

    1. Run 3 · 12 Nov…closes at 6 pm on weekdays…/a/3Lw8
    2. Run 9 · 12 Nov…open until 6, worth the trip for…/a/9Hc1
    3. Run 12 · 12 NovDoes not mention opening hours./a/12Rd
    Fixes on your own pages

    See why a competitor wins, then fix your page.

    For each question, Massyl shows which competitor page the engines cite and what yours is missing. It drafts the change to your existing page first, and a new comparison page only where you have none. Nothing goes live until you approve it.

    • Before and after view of every edit, with undo
    • Publishes to WordPress, Webflow, Shopify, Framer, Wix, HubSpot, GitHub
    • Re-checked after indexing, with the change shown as a range
    haldencoffee.com/visitDraft ready

    Cited instead: cuveecoffee.com/roast-schedule. Your page is missing roast days and has opening hours only inside an image.

    − [image: opening hours]
    + Open daily, 7 am to 8 pm
    + We roast on Tuesdays and Fridays, and ship the same day.
    Approve and publishAsk for changes
    Proof that it worked

    Changed pages, compared with pages you left alone.

    AI answers drift on their own. So Massyl keeps a set of questions untouched as a control, and reports the gain of your fixed pages over that control, with its range. If the ranges overlap, it tells you the change is not proven yet.

    • Held-back control on every fix campaign
    • Gain shown as a range, never as a single lucky number
    • Annotated timeline of what was published and when
    Mention rate, 24 questions+14 pts over control (range +6 to +22)
    Questions you fixedControl, left alone95% range
    Every language, written natively

    Your buyers don't all search in English.

    Track any country and language in one project. Fixes and new pages are written in that language from the start, not translated afterwards. Arabic and Hebrew pages are laid out right to left.

    USbest espresso beans in austin40%
    FRmeilleur café en grains à Lyon15%
    AEأفضل قهوة مختصة في دبي10%
    MXmejor café en grano en CDMX25%
    DEbester Espresso in Berlin kaufen20%
    Sentiment
    Positive 62%Neutral 29%Negative 9%
    ChatGPT
    70%
    Perplexity
    58%
    Gemini
    55%
    ChatGPT · run 4negative
    “Halden is a bit pricey but the beans are always fresh.”
    Perplexity · run 11positive
    “Locals rate Halden’s single-origin roasts among the best in Austin.”

    62% positive, 29% neutral, 9% negative. Top negative: “pricey”.

    • Every answer tagged positive, neutral or negative
    • The exact phrases that cost you
    • Split by engine and by language
    Question demand
    W40W43W46

    “espresso beans austin” searched about 1,900 times a month, rising.

    • Monthly volume for each question
    • Rising questions flagged every week
    • Suggested questions to start tracking
    AI crawler visits
    W40W43W46

    OAI-SearchBot read 41 pages this week; PerplexityBot 18.

    • Which AI bots read which pages
    • Pages the bots never reach
    • Daily counts per bot
    Alerts
    Perplexity stopped naming you for “espresso beans austin” · 70% → 35%
    ChatGPT now names you for “coffee subscription austin” · 15% → 50%

    Sent to Slack or email when a drop is larger than the margin of error.

    • Only when a change beats the margin of error
    • Slack or email
    • A weekly digest, always
    Agency workspaces
    Halden Coffee 40%Leeds Dental 45%Nord Bikes 12%

    Every client in one view, each with its own branded report.

    • Every client in one login
    • Branded reports per client
    • Access set per client
    Benchmark
    Peer median 24%
    Halden
    38%
    Cuvée
    41%
    Fleet
    19%

    You are 14 points above the peer median. Cuvée still leads by 3.

    • You against your top competitors, per engine
    • The median for your category
    • Who gained, who lost, and why

    Everything in the product: 59 features at launch

    Pick an area to see what it includes.

    Track

    BuyingCompareLocalUSFR+43 more
    • Suggested questions, bulk import and tags
    • Search volume for each question
    • Brand name variations and misspellings
    • Branded, intent and persona tags
    • Several locations per brand
    • Any country and language
    • Related questions and query fan-out for each topic
    • Competitors detected for you automatically
    • Sponsored answers and ads inside AI results
    • Product-level tracking for online stores, with Shopify import
    • Keyword research for AI prompts and Google keywords

    Understand

    Share of voice 22%Sentiment 62%Sources 31
    • Share of voice and competitor auto-detection
    • Position and sentiment of every mention
    • Benchmarks against your industry
    • Cited sources, grouped by site
    • Wrong facts, traced to the page AI learned them from
    • Ask the data in chat, or through MCP
    • Answer history: see how each answer changed week to week
    • Citation share by source type: your site, reviews, forums, news
    • Brand attributes and objections AI repeats about you
    • Backlink and brand-mention monitoring

    Fix

    7 tasks3 drafts1 approved
    • Task list ranked by expected impact
    • Edits to your pages, with approval and undo
    • Comparison and FAQ pages, written natively
    • AI crawler, robots.txt and llms.txt checks
    • IndexNow pings after each publish
    • Earned-media tasks: Reddit, reviews, listings
    • Content briefs for new comparison and FAQ pages
    • Schema and internal-link suggestions per page
    • Article drafts in your brand voice, 8 a month on Grow
    • Alternatives and listicle pages, written at scale
    • Outreach to the sites AI cites, with drafted emails you approve
    • Autopilot: auto-publish low-risk fixes, with undo and a held-back control
    • AI-agent-ready pages: lightweight versions served to AI crawlers

    Report

    • AI referral traffic from GA4
    • AI crawler visits from server or CDN logs
    • Weekly written report, PDF and share links
    • Looker Studio, CSV, API and webhooks
    • Slack and email alerts, and a Slack bot
    • Scheduled PDF and slide exports
    • A one-page summary for your boss or client
    • Share-of-voice benchmarks for your industry
    • Google rankings next to AI answers for the same question

    Agencies

    HaldenLeeds DentalNord Bikes+9
    • Client workspaces, sub-brands and a cross-client view
    • Reports on your own domain and brand
    • Client approvals by link, no logins
    • Pitch reports for prospects
    • Done-for-you work to resell
    • Client portal by magic link
    • Prospect audits to win new clients
    • Lead widget that runs a free check on your agency site

    Security and data

    Read-onlyExportDeleteRoles
    • Read-only analytics access
    • Export or delete everything, any time
    • Seats and roles that grow with your plan
    • Encrypted in transit and at rest
    • Your data is never used to train AI models
    • Data kept in the EU or US
    • Audit trail of every approved change
    • SSO on the Agency plan

    Next, after launch

    RoadmapBuilt in publicIn this order
    • Help-center checks: find articles that contradict your product or AI answers
    • Grok, DeepSeek, Meta AI and Mistral
    • SOC 2
    Built for how you sell

    Different businesses lose AI answers in different ways.

    A question your buyers ask

    What Massyl often finds

    What Massyl does

    Example result
    How we measure

    The maths is public. Check it yourself.

    A mention rate is a share: how many of the answers named you. With twenty answers, the 95% range comes from the Wilson score interval, a standard method that stays honest at small samples and near 0% or 100%. When the range is wider than 40 points, Massyl labels the result "not enough data" and asks for more runs.

    This calculator is free and works right now.

    Margin of error calculator Free, works now
    Mention rate40%
    95% range22–61%

    A real range. Enough to compare with next week.

    Free tools

    Six free tools you can use right now.

    No sign-up. Everything runs in your browser and nothing you type leaves this page. Pick a tool below. The sixth, a margin-of-error calculator, sits in the “Measured, not guessed” section.

    AI crawler checker

    Paste your robots.txt to see which AI engines can read your site.

      Tip: open yoursite.com/robots.txt, copy everything, paste it above.

      Question generator

      The questions buyers ask AI about a business like yours, ready to track.

        AI readiness score

        Tick what is true for your site. You get a score out of 100 and the three fixes worth doing first.

        Your score0/ 100

          llms.txt generator

          Write the file that tells AI assistants what your site is and which pages matter. Upload it as yoursite.com/llms.txt.

          llms.txt
          
                

          Schema generator

          Structured data that states your facts plainly for AI engines: business details and FAQ. Paste it into your page's <head>.

          JSON-LD
          
                
          How it works

          Measure, fix, then prove it worked.

          The same set of questions is asked every week, so before and after are compared fairly.

          Suggested from haldencoffee.com48 questions found
          best espresso beans in austinBuyingUS
          halden vs cuvée coffeeCompareUS
          coffee subscription that ships same dayBuyingUS
          meilleur café en grains à LyonBuyingFR
          Week 46 · 48 questions × 6 engines × 20 runs5,760 answers
          ChatGPT8 of 2040% 22–61
          Perplexity14 of 2070% 48–85
          Google AI Mode5 of 2025% 11–47
          Gemini6 of 2030% 15–52
          Task 1 of 7 · expected impact: highDraft ready

          Add roast days and text opening hours to /visit. Cited instead: cuveecoffee.com/roast-schedule.

          − [image: opening hours]
          + Open daily, 7 am to 8 pm. We roast on Tuesdays and Fridays.
          Approve and publishAsk for changes
          Fix campaign · 24 questions · 4 weeks after publishing+14 pts over control
          Questions you fixedbefore 24%after 44%
          Control, left alonebefore 25%after 27%
          Gain over control95% range+6 to +22 pts
          Done for you, and for agencies

          Rather not do the work? We will.

          Our team writes and publishes the fixes, and you approve them in the app. Everything happens in writing, so there is nothing to schedule.

          Done for you

          See the plan →
          • Everything in Grow, run by our team
          • Page edits and comparison pages, published after your approval
          • Listings, review sites and community answers where engines look
          • A written report every week, in your language
          M
          Massyl team
          ChatGPT names Cuvée for “espresso beans austin” and cites their roast-schedule page. You don't have one. Draft ready: /roast-schedule.
          You
          You
          Add that we ship the same day.
          Approve and publishAsk for changes

          For agencies

          See the plan →
          • A workspace for each client
          • Reports under your name, logo and link
          • Every number links to the answers behind it
          • Before and after packs that show what your work changed
          Client report · Perplexity · “dentist open saturday leeds”
          Before
          10%
          After
          45%
          The change is larger than the margin of error.
          MCP server

          Massyl inside your AI assistant.

          Connect Massyl to Claude, ChatGPT or Cursor through MCP and ask about your visibility in plain words. Your assistant reads the same numbers, ranges and answers you see in the app, and can draft fixes for you to approve.

          Works withClaudeChatGPTCursorAny MCP client

          30 tools, included on every paid plan and read-only on Free. A REST API and webhooks are there too.

          # In Claude, with the Massyl MCP server connected
          you › How visible were we on Perplexity this week?
          massyl.get_mention_rate(engine="perplexity", week="2026-W46")
          Halden was named in 70% of answers (range 48–85%, n = 20). Up from 45% last week. The change is larger than the margin of error.
          you › Which questions do we lose to Cuvée? Draft a fix for the biggest one.
          massyl.list_gaps(competitor="cuvee") → massyl.draft_fix(question="espresso beans austin")
          Biggest gap: “espresso beans austin” on ChatGPT, Cuvée 65% vs Halden 40%. Draft ready for /visit: adds roast days and text opening hours. Approve it in Massyl.
          get_mention_rate

          Rate and 95% range for any question, engine and week

          get_answers

          The stored answers behind a number, with dates and links

          list_gaps

          Questions where a competitor is named and you are not

          list_wrong_facts

          Facts AI states differently from your site

          draft_fix

          A draft edit for one of your pages, ready to approve

          get_report

          This week's written report, in your language

          30 tools in total. The six above, plus:

          get_share_of_voicelist_competitorsget_sourcesget_sentimentlist_questionsadd_questionssuggest_questionsget_answer_historycompare_enginesget_crawler_visitsget_ai_trafficlist_tasksapprove_fixpublish_fixundo_fixget_control_gainlist_alertsset_alertexport_reportlist_clientsget_client_reportget_benchmarkcheck_robotslist_languages
          • Google Analytics 4
          • Search Console
          • Looker Studio
          • Slack
          • REST API
          • Webhooks
          • Server logs
          • CSV export
          Pricing

          Priced by questions, not credits.

          One question means twenty answers on each engine, every month, spread across the weeks so each number comes with its range. No credits, no per-answer surprises.

          Free
          $0
          At launch. No card, no time limit

          See where you stand.

          • 10 questions, one report
          • 800 answers
          • 4 engines
          • 1 competitor
          • Answers behind every number
          Reserve my free report
          Track
          $99/mo

          The number, with its range.

          • 100 questions
          • 8,000 answers a month
          • ChatGPT, Gemini, AI Overviews, AI Mode
          • 3 competitors
          • Wrong-fact alerts
          • Any language
          Reserve this plan
          Grow Most teams
          $299/mo

          Tracking plus the fixes.

          • 250 questions
          • 24,000 answers a month
          • Adds Claude and Perplexity
          • 10 competitors
          • Page fixes and drafts
          • Publish with approval
          • GA4 and Search Console
          Reserve this plan
          Agency
          $499/mo

          For client work.

          • 500 questions, shared
          • 46,000 answers a month
          • Client workspaces
          • Reports under your brand
          • Full API and MCP write access
          Reserve this plan
          Done for you

          Everything in Grow, with our team writing and publishing the fixes. Weekly written report. No calls.

          $1,200/mo
          Compare every feature by plan
          FeatureFreeTrackGrowAgency
          Track
          Questions tracked10, once100250500
          Answers per month800 once8,00024,00046,000
          Engines446 (Claude and Perplexity on 100 key questions)6 (Claude and Perplexity on 150 key questions)
          Competitors1310Unlimited
          Languages and countries1AnyAnyAny
          Locations per brand115Unlimited
          Understand and fix
          Answers behind every number
          Wrong-fact alerts—
          Sources and sentiment
          Page fixes and drafts——
          Publish with approval and undo——
          Report
          Google rankings next to AI answers——
          Weekly written report—
          GA4 and Search Console——
          API, MCP and webhooksMCP read
          Agencies and data
          Client workspaces and white label———
          Seats1310Unlimited
          Export and delete everything
          SupportWrittenWrittenWrittenWritten, priority

          Planned founding prices for early-access users. They may change before launch.

          Get your free 20-run report.

          Twenty answers per question, every answer, every range. Free when Massyl opens, no card.

          Reserve my free report
          Compared

          What Massyl adds.

          From the public websites and docs of 17 AI visibility tools, read on 14 November 2026. Tools change, so check before you buy.

          Typical AI visibility trackerMassyl
          Answers per questionOne per check, at most tools we readTwenty per engine, every month
          Margin of errorNone shown by any tool we readA range on every rate
          EvidenceSome keep the answer textEvery number opens its answers, each with a dated link
          FixesGeneral advice or new articlesEdits to your existing pages, tied to the competitor page that wins
          LanguagesTracking in many, content mostly EnglishTracking and writing in your customers' language
          Proof of a fixBefore and after chartsGain over a held-back control, with its range
          ServiceOften a strategist call or a Slack channelWritten briefs and approvals. No calls.
          Who is behind it

          No logo wall yet. Here is what you get instead.

          Massyl is new and in build. Rather than borrowed logos, here is who is building it and how your data will be handled.

          MX@muxbuilds
          LLM and search engineer

          Building Massyl in public. Progress, numbers and mistakes are posted on X every week.

          Follow the build →

          Your data

          We keep your questions, the engines' answers and, if you connect it, read-only traffic. Nothing goes on your site without your approval. Export or delete everything, any time.

          Open method

          Every rate shows its sample size and range, and every answer says whether it came from the engine's app or its API. The formula is on this page.

          See how we measure →

          Build log

          1. Studied 17 AI visibility tools, feature by feature
          2. Shipped this page, a free starter kit and six free tools
          3. The measurement engine, in public on X. Every feature on this page is what ships at launch.

          No calls, ever

          Setup, support, approvals and the done-for-you plan all run in writing. You will never need to book a demo to see a price.

          FAQ

          Questions

          MXBuilt in public
          Questions, progress and early access, all on X.
          Ask @muxbuilds
          Why twenty answers and not one?

          AI engines answer the same question differently each time. With one answer, a business named 40% of the time shows up as 0% or 100%. Twenty answers give a rate and a range, so you can tell a real change from noise.

          What does the range mean?

          It is a 95% range. If the true rate were outside it, results like these would be very unlikely. When the range is too wide to act on, Massyl says "not enough data" instead of showing a confident number.

          How are the answers collected?

          Through each engine's official API where one exists, and through the public app where the app answers differently. Every stored answer says which method produced it.

          Will you change my website without asking?

          No. Every edit is shown as a before and after and nothing is published until you approve it. Published changes can be undone.

          What data will you keep?

          The questions you track, the answers the engines give, and read-only traffic numbers if you connect Google Analytics or Search Console. You will be able to export everything and delete your account at any time.

          Which engines do you cover?

          ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Claude at launch. Copilot, Grok, DeepSeek, Meta AI and Mistral are next. Each engine can be switched on or off per question.

          Do you track my local business in several cities?

          Yes. One brand can have several locations, each with its own questions, country and language.

          What if I need more questions than my plan has?

          You can add questions in blocks of 25 at any time, or move up a plan. Nothing is cut off mid-week; the extra questions start on the next run.

          Is there a trial, and can I pay yearly?

          The Free plan gives you one full report with no card and no time limit. Yearly billing gives you two months free. You can cancel any time.

          Will I ever need a call?

          No. Setup, reports, approvals and support all happen in writing, including on the done-for-you plan.

          When can I use it?

          Massyl is being built now, in public on X. Early access and the first free reports go to people following the build.

          In buildBuilt in public

          Find out what AI tells your customers.

          Follow the build on X for early access and a free report when Massyl opens.

          Free 20-run report on 4 engines when Massyl opens. No card, no sales calls. One email when it is ready, nothing else.