Agencies pass an AI relevance, product capability, and evidence gate before scoring. Five dimensions are scored from 0 to 10 in half-point steps. The arithmetic mean becomes the displayed score. Ties are resolved by shipped AI evidence, audience fit, then recency. Payments and submissions cannot change a position.
An agency must show a shipped AI product or a detailed AI product case. Adding AI to a service list, publishing generated concepts, or naming an AI client without scope is not enough.
The work must include product or service design, not only model development, automation consulting, or marketing. Every profile needs at least two accessible sources.
Detailed cases and working products carry the most weight. Official service pages establish capability but do not prove delivery. Technical documentation, client announcements, and verified profiles can corroborate scope and responsibility.
Vendor results remain vendor claims unless another accessible source confirms them. The rating does not convert an agency's marketing figure into an audited outcome.
The five dimensions have equal weight. Editors score in 0.5-point steps, calculate the arithmetic mean, and round once to one decimal place. There is no hidden adjustment after the average.
If totals match, stronger shipped AI evidence wins. Audience fit is second and the recency of the strongest relevant work is third.
The rating date changes only after a substantive evidence review. An agency may submit a correction or new source, but submission does not guarantee inclusion or a score change.
Send the exact disputed statement, the source URL, and the date checked to inquiry@aiproductdesignagencies.com.
The score is an editorial assessment for AI product buyers. It is not a customer-review average, security certification, model benchmark, award, or guarantee of project performance.
The publication is an AI Ratings editorial project. Positions are not sold. Agency selection and evidence scoring require editorial judgment.
Live AI features and product cases with enough detail to distinguish delivery from a concept.
Evidence for prompts, context, review, failure, iteration, and operational use rather than a chat box alone.
How the work handles uncertainty, verification, permissions, disclosure, safety, and quality evaluation.
Evidence that real users, jobs, and organisational change shaped the product and its rollout.
Detailed cases, technical collaboration, accessible sources, recent dates, and outside corroboration.