Publication standard

How Trakkr publishes AI rankings

The method is designed to stop thin, stale and duplicative category pages from entering search. A URL is not a ranking until its current evidence passes every check.

[01]Source and observation

What the source provides

Trakkr groups observed AI answers by category, counts brand appearances, and ranks brands using the source feed’s 0–100 visibility score. The score reflects frequency and prominence; raw mentions and average position are shown beside it.

Observation window
Latest daily observation. This is a current snapshot, not a rolling market-share estimate.
Sample size
The sum of current brand mentions in the category. Ranked-brand count is reported separately.
Source-named models
ChatGPT, Claude, Gemini, Perplexity.

Latest source status

Data updated 18 August 2026. The source reports 9,195 brands across 500 categories.
Open live status JSON

[02]Measurement definitions

Visibility score

The source feed’s 0-100 score for how often and how prominently a brand appears. It is a relative visibility measure, not a percentage of people or revenue.

Observed mentions

Brand appearances reported in the latest daily category observation. This is the sample used by the publication gate.

Mention share

A page-level calculation: a brand’s current mentions divided by all current mentions shown for that category.

Average position

The source feed’s mean position for the brand in observed answers. Lower positions are more prominent.

Movement

The global feed supplies 24-hour and 7-day score movement. Category pages do not show movement because the public category feed does not provide it.

Editorial explanation

Context written by Trakkr to describe the question a category answers. It is not measured data and is kept outside the ranking table.

[03]Deterministic publication gate

All checks are required. Click-through rate alone never decides publication.

  1. [1]

    Fresh date

    The source update date must parse, must not be in the future and must be no more than 72 hours old.

  2. [2]

    Enough observations

    A category needs at least 30 current observed mentions.

  3. [3]

    Enough brands

    A category needs at least 10 ranked brands and at least 10 distinct normalized names.

  4. [4]

    Unique rows

    Duplicate brands or duplicate ranks fail the page.

  5. [5]

    Valid measurements

    Scores must be finite and between 0 and 100. Mentions cannot be negative, and average position must be positive.

  6. [6]

    Complete explanation

    The category must be in the curated set and have a stated question, method, window and useful category explanation.

  7. [7]

    No placeholders

    Empty, unknown or placeholder brand names fail the page.

  8. [8]

    Schema follows evidence

    Ranking and dataset schema are emitted only after every preceding check passes.

[04]Coverage and limits

The public feed does not expose the prompt list, prompt denominator, response text, citation evidence or a per-model split. Trakkr therefore does not claim that a category ranking explains why a model chose a brand.

Scores compare visibility within the source method. They do not measure product quality, customer satisfaction, sales, market share or a recommendation by Trakkr.

The current publication is intentionally selective. Categories outside the curated set, even when the API returns rows, are not automatically indexable.

[05]Corrections and history

Report a source issue

Send the category URL, brand name, source update date and the exact row that looks wrong through Trakkr support. The source JSON remains linked on every current page.

Contact support

Historical rankings

The archive remains policy-only until a complete, comparable time series can support dated historical tables without filling gaps or mixing methods.

Read the archive policy