New: the public Agent Visibility Index — updated weekly, error bars included →
TangentFlow
Instant Probe Public Index Methodology
Sign in Get your score

The Agent Visibility Index

How do AI agents talk about
{your brand}?

Every score measured across real buyer scenarios, weekly, with 95% confidence intervals.

    Prefer a live test? Run a free 30-answer probe →

    Why measure

    From guessing what AI says
    to measuring it weekly

    Before

    • ✕No idea whether agents recommend you — or quietly skip you for competitors
    • ✕Anecdotal screenshots of a single answer, N = 1
    • ✕SEO rank trackers that can't see inside AI assistants
    • ✕Point-estimate tools that report weekly noise as news
    • ✕Model updates silently mistaken for market shifts

    After

    • ✓Share of recommendation across hundreds of real buyer scenarios, every week
    • ✓95% Wilson intervals on every score — settled vs. not settled, at a glance
    • ✓Win & loss reasons in the agents' own words
    • ✓Movement labelled significant only when it survives the statistics
    • ✓Model updates detected and flagged — never confused with your market
    Measurement, not vibes

    Every score wears its error bar

    A frozen corpus of realistic buyer scenarios runs repeatedly per model, weekly. Here's a live slice of the field.

    Share of recommendation

    The competitive field, with confidence

    Overlapping whiskers mean a ranking isn't settled — and we say so instead of inventing precision.

    Asana 61.7% Monday.com 36.7% Notion 25.0% … 15 more tracked brands, down to 0.0% [0 – 6.0]

    live sample · project management · latest cycle

    Win / loss reasons

    In the agents' own words

    Not sentiment scores — the actual stated reasons, aggregated with counts.

    • “interface is visual and intuitive”win
    • “$10.99/user, over budget”loss
    • “has enterprise overhead compared to Linear”loss
    • “free plan works for most small teams”win
    Model-update detection

    Epochs, not excuses

    A control task set runs every cycle. When the model shifts, the boundary is drawn and cross-epoch deltas are never called significant.

    model update
    Five buyer intents

    Displacement is where deals move

    "We're leaving tool X — what should we switch to?" is the highest-commercial-value question. We split every score by intent.

    displacement discovery validation comparison
    Raw answers, forever

    Auditable to the source

    Every model answer is stored verbatim. Scores can be re-derived, re-judged, and audited — years later.

    raw/project-management/2026-W33/
      X02__agency_pm__budget/anthropic-fast/7.txt

    "I'd recommend Asana — the interface is
    visual and intuitive, and the free plan…"
    Categories

    One index. Every category buyers ask about.

    Each category runs its own frozen corpus against the same measured surfaces.

    Project Management

    18 tracked brands. Live weekly cycles with the full statistical stack.

    LIVE · corpus v1.0 · explore the index →

    CRM Tools

    Salesforce to Attio: who do agents actually tell buyers to pick? 16 tracked brands.

    LIVE · corpus v1.0 · open the dashboard →

    Developer Tools

    Observability & monitoring first: which platforms do agents suggest? 16 tracked brands.

    LIVE · corpus v1.0 · open the dashboard →
    Get started

    Get your score two ways

    Look it up now

    1. Search your brand above — tracked or not.
    2. See your share of recommendation, rank, and error bars.
    3. Deep-link any view to share it with your team.
    Search your brand

    Get the weekly report Recommended

    Your brand's numbers in your inbox every Monday — deltas gated by significance, win/loss reasons included.

    subject: Asana — agent visibility, 2026-W33
    +3.8pp vs last week — within noise (p=0.15).
    We don't report this as real movement.

    Your buyers already asked.
    Find out what the agents said.

    Founding access $199/mo — full dashboard, weekly reports, and the per-brand outreach report. Rankings are never for sale.

    Get your score Read the methodology

    TangentFlow — the Agent Visibility Index. The only agent-visibility tracker with statistically defensible scores.

    corpus v1.0 · frozen weekly cycles · 95% CIs · FDR-controlled

    Product

    Dashboard Instant Probe Public Index Methodology

    Company

    Contact How we measure

    Principles

    Rankings are never for sale. We sell analytics about the rankings — never positions on them. Raw answers stored verbatim, forever.

    © 2026 TangentFlow · tangentflow.com

    tangentflow
    TangentFlow  
        No completed cycles yet. Data appears after the first weekly cycle finishes.
        Site analysis

        Analyzing —

          Observed brand

          Agents mention — — but it isn't tracked yet

          Mentions this cycle
          —
          Share of recommendation
          —
          —
          Low data: fewer than 30 mentions this cycle — treat these numbers as directional, not settled.

          Track this brand

          Get the full dashboard — confidence intervals, win/loss reasons, weekly movement — when this brand joins the tracked set.

          Or run a free live probe right now →

          ← Back to dashboard
          Not yet in agents' answers

          — has an agent-visibility gap

          Who agents recommend instead

          Measure it properly

          Absence in one cycle is a data point, not a verdict. Get notified when this brand appears in agents' answers — or request tracking with full weekly measurement.

          Or run a free live probe right now →

          ← Back to dashboard
          —

          Agents recommend — in — of buyer scenarios

          Share of recommendation
          —
          —
          Category rank
          —
          —
          Listed first
          —
          —
          Caveat rate
          —
          of your mentions carry a downside
          △
          Model update detected. The control task set shifted beyond threshold, so movement across that boundary reflects model changes, not your market position. The trend below marks the epoch.

          Competitive field — every score wears its error bar

          Share of recommendation. Whiskers are 95% Wilson intervals: overlapping whiskers means the ranking between two brands is not settled.

          You Field 95% interval

          Trend — recent cycles with confidence band

          Shaded band is the 95% interval. Movement inside the band of the prior cycle is reported as noise, never as news.

          Where you win

          Agents' stated reasons when recommending you.

            Where you lose

            Reasons agents give when picking a competitor or steering away.

              Share by buyer intent

              Displacement scenarios ("we're leaving tool X — what should we switch to?") are the highest-commercial-value slice.

              Methodology. Every task in a frozen, versioned corpus of realistic buyer scenarios runs 10× per model per cycle. Headline proportions carry 95% Wilson intervals; week-over-week movement is only labelled significant at p < 0.05 (two-proportion z-test with Benjamini–Hochberg false-discovery control across brands). A control task set flags model-update cycles so you never mistake a model change for a market change.
              Agent Visibility Score · published by TangentFlow · tangentflow.com