Mavel Research

The AI Visibility Report.

The first in a planned quarterly series on how AI engines actually answer — computed from Mavel's daily tracking panel across 8 engines, published in the open with its full methodology. Not a sales asset: a reference other people can cite, check and argue with.

ungated web report numbered releases, quarterly methodology published 8 engines · 6 model providers — not two UIs on the same model

Next release

Report 01 · Q3 2026.

MAVEL AI VISIBILITY REPORT · 01IN PREPARATION

THE HEADLINE
One quotable stat, stated in the first sentence — chosen from the findings below once computed. The number the coverage cites.
ENGINE DIVERGENCE
Do the 8 engines agree on who wins a category? Share-of-voice overlap between ChatGPT, Gemini, Claude, Perplexity, AI Overviews, AI Mode, Copilot and Grok for the same brand set. No published study covers more than four.
CATEGORY CONCENTRATION
Top-3-brand share of AI citations per tracked category — which categories are already winner-take-most, and which are still open.
SOURCE MIX, PER ENGINE
Owned vs earned vs UGC/forum share of citations, engine by engine — testing whether the "Reddit is eating AI search" narrative is universal or engine-specific.
MOVERS
Largest quarter-over-quarter share shifts per category, and the source changes behind them.
METHODOLOGY
Panel size, prompt counts, engine list, date range, computation notes — published in full, prominently, so the numbers are citable with confidence.

Why this exists

The panel already runs. We publish what it sees.

Every research property in this category had to build its data pipeline to write its report. Ours is a byproduct of the product: the same daily 8-engine panel our customers use, aggregated and anonymised.

Six model providers

Published studies cover 3–4 surfaces, mostly two companies. Our 8 engines span OpenAI, Google, Anthropic, Perplexity, Microsoft and xAI — divergence a skeptic can't wave away as UI skins on one model.

Daily, longitudinal

Not a one-off crawl: the same prompts, every day, so quarter-over-quarter movement is real measurement, not sampling noise.

Open by default

Web version ungated, methodology first-class. The properties people actually cite (Seer, SparkToro) are the ones without a sales wall.

Methodology

How the numbers are made.

The panel

Engines: ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Claude, Perplexity, Microsoft Copilot, Grok — queried daily.
Prompts: the full tracked prompt panel across every monitored category — real buying-intent questions, not synthetic keyword expansions. Exact counts publish with Report 01.
Aggregation: all published figures are category-level aggregates, anonymised across workspaces. No customer-identifiable data is ever published.
Period: each report covers a full quarter of daily observations; movers are computed against the prior quarter's identical prompt set.
Limitations, stated: panel composition skews toward the categories our customers track; concentration stats are reported only for categories above a minimum prompt threshold.

Get the releases

One email per quarter.

Everything publishes here, in the open — bookmark the page, or get the one-email-per-quarter announcement.

NOTIFY ME — REPORT 01