How TOP SCORE™ is built

Aggregated third-party ratings from independent review platforms.

• Every source is rescaled to a common 0–100 scale at ingestion, so a 4.6/5 and an 8.9/10 are comparable.
• Small samples are pulled towards the category prior. A 4.9 from 12 reviews does not outrank a 4.7 from 40,000.
• Each source is weighted by its editorial quality, its review volume and how recently it was checked.
• Where sources disagree sharply, the score is penalised and confidence drops — disagreement is information, not noise.

How visible, recommended and cited a platform is inside answer engines.

• Each category has a fixed set of about twenty buying questions, written the way a real person would ask them and containing no brand names. The set stays stable so this month's result is comparable with last month's.
• We put those questions to an answer engine and record every product it names, in what order, and whether it was actually recommended rather than mentioned in passing.
• Both a searching engine and one answering from memory alone are measured. Being findable today and being known already are different kinds of standing.
• Citation rate is the share of answers where the engine actually retrieved the platform's own site while answering.
• We only measure engines we can query, and every reading records which engine produced it. A narrow panel lowers the confidence rating rather than being presented as a full picture.

Independent coverage across the open web, weighted by who is doing the covering.

• We search for coverage of each platform inside its own category and record the publishers writing about it.
• Every publisher is classified once — major publication, trade press, credible independent, community, vendor, or an SEO farm — and that classification is reused everywhere, so two platforms are never judged on different scales.
• An affiliate listicle counts for very little however professional it looks. The vendor's own domain is excluded entirely.
• We do not operate a backlink index and do not pretend to one. This measures the quality of who is publishing about a platform, not raw link counts.
• Repeat sweeps only record coverage we have not seen before, which is what makes citation velocity meaningful over time.

Our own rubric, written per category.

• Each category defines its own criteria and weights: regulation and payout record matter for a broker, credit burn and output quality for an AI tool.
• Every criterion score carries a one-line justification tied to retrieved evidence.
• An unverifiable claim scores badly on transparency rather than being ignored.
• Assessments age: an old assessment loses confidence even if it is not replaced.

First-party reviews written by TOP10 users.

• Recent reviews count for more than old ones, on a decay curve rather than a cut-off.
• Verified users count for more; reviews carrying abuse signals count for less.
• The average is stabilised against a prior, so a handful of five-star reviews cannot move a score far.
• Only reviews that have passed moderation are counted.