Rating methodology

Every profile combines automated collection with editorial verification. Each rated field is labelled with its source so you always know whether a value is a machine estimate, editor-verified, or founder-submitted.

1. Data collection

• Official websites are crawled for titles, descriptions and headline claims, and re-checked weekly for availability.
• Pricing is normalised into type (subscription, seat, usage, revenue share, enterprise, open source), currency, range and billing unit.
• Products whose sites fail repeated checks are marked as unreachable or discontinued; their pages remain with alternatives.

2. AI Marketing Autonomy Standard (L1–L5)

1. L1 Prompt-assisted: the user writes a prompt; the tool returns a single output from a general model.
2. L2 Workflow copilot: templates and guided steps; a human reviews and advances every step.
3. L3 Conditional automation: event- or rule-triggered flows (e.g. abandoned-cart emails, lead routing); humans handle exceptions.
4. L4 Autonomous agent: end-to-end goal decomposition, external tool use and self-correction within one scenario.
5. L5 Self-evolving swarm: multiple agents negotiate, allocate budget across channels and learn in a closed loop.

A level is shown only after classification; unrated products display “Not yet rated”.

3. Moat tiers

• Thin wrapper: relies on third-party model APIs with no proprietary data; exposed to model upgrades.
• Workflow-embedded: built on general models but deeply integrated with customer systems (CRM, commerce, ad accounts), creating switching costs.
• Proprietary moat: proprietary vertical models, closed-loop outcome data, or specialised real-time infrastructure.

Tiers marked “Machine estimate” are preliminary and are replaced by editor ratings as reviews are completed.

4. What we do not do

• We do not publish star ratings until there are enough moderated, verified user reviews.
• Affiliate relationships and sponsorships never change ratings or ordering of comparison tables.
• We do not publish unverified founding dates or funding figures as facts.

5. Corrections

Vendors and users can request corrections. Verified changes are applied in the next weekly data release and the profile's “last verified” date is updated.