Methodology

Transparent about how we rank and how we know what we know. This is also our ranking disclosure under EU consumer law (Omnibus Directive).

How we rank

In Match, the engine picks the best-fitting tool for each step of your task. The main parameters, in order of weight:

Fit: is this the tool's core capability or a secondary one?
Your requirements: budget, level, EU data residency, language, platform, watermark and open source. Hard requirements filter; if nothing qualifies, we say so.
Data quality: confidence, freshness and the status of the price.
Price: the monthly price of the cheapest fitting plan; a free plan counts.
Simplicity: a tool covering several steps counts in the "Fewest tools" variant and slightly in "Recommended".

On capability pages, tools with this as their core capability come first, then ordered by confidence, freshness and entry price. In the explorer you choose: relevance, name, price or freshness.

Money never changes the order

Affiliate links and sponsoring are not inputs to ranking. The ranking code technically cannot read monetisation data, and every change runs a test comparing results with and without affiliate links: they must be identical. Sponsored spots are always labelled "Sponsored" and kept apart from results. Disclosure

What the statuses mean

Verified: Confirmed on the vendor's official source, with a verbatim quote, and high confidence.
Supported: Several independent sources agree, or the official source mentions it but it hasn't been checked verbatim yet.
Community: Based on community or social sources only.
Unverified: A single secondary source or an editorial assumption that still needs checking.

Confidence (0–100)

Confidence is computed from evidence, not a feeling. Factors: the type of source (official weighs most), whether the quote was found verbatim on the page, the age of the observation, the method (an automated check weighs more than web research) and the number of independent sources that agree. If a sanity check fails (e.g. an impossible price jump), confidence stays low. A language model's self-reported certainty is never accepted as evidence.

Freshness

Every kind of data has a time limit. When it expires, we show "Information may be outdated" with the date of the last check. Current limits (fresh / aging):

Prices: fresh up to 14 days, then aging up to 45 days
Website reachability: fresh up to 7 days, then aging up to 21 days
Capabilities: fresh up to 60 days, then aging up to 120 days
Social signals: fresh up to 7 days, then aging up to 30 days
Videos: fresh up to 30 days, then aging up to 90 days

What our agents do automatically

Agents check official pricing pages, changelogs and websites, and discover new tools. A price change is only published after it has been measured twice, some time apart, with the same result. Large jumps, shutdowns without an official announcement and new tools go past a human by default. If a run sees unusually many changes at once, publishing freezes automatically. Every automated change is logged and can be reverted.

Where we use AI

To understand your question (when keyword analysis is unsure) and to extract data from official pages, always with a check that every value appears verbatim in the source. Text composed with AI is labelled. The recommendations themselves are computed, not generated.

Limitations

Prices are as shown by the vendor, usually excluding VAT and sometimes different per region. Dollar amounts are converted with the ECB daily rate and are indicative. Part of the data comes from web research across several sources and is therefore marked SUPPORTED until our agents confirm it on the official page.

Corrections

Something wrong? Report it with a source; we respond and keep a public corrections log. Corrections