The full picture: quality, reliability, and cost.
Ten search providers, 216 test searches, and two AI judges. Provider identities were hidden from the judges (blind inputs). Each chart below has a one-line explainer — and every score comes with a confidence range showing how sure we are.
Loading list-price vs. paid-cost context…
What you pay vs. what you get
Each dot is a provider: further left means cheaper per search, higher up means better average quality score. The best value sits toward the top-left.
Cost per search vs. average quality score
One dot per provider with rankable Search data. The dashed line traces the "best value" providers — nothing is both cheaper and better than them.
How to read the axes: X = what one search costs (left = cheaper). Y = quality score 0-100 (how often this provider gives the better result vs an average one; 50 = average, higher = better). Best value = top-left.
Linkup isn't plotted: no rankable Search categories (insufficient evidence).
Quality scores, with how sure we are
Pick a search category to compare providers. Taller bars mean better quality; the whiskers show the confidence range — a wide whisker means less certainty. Providers without enough data aren't shown.
Quality by category
Loading scores…
How to read the axes: Each bar = quality score 0-100 for that category (50 = average provider). Whiskers = confidence range.
How often they deliver, and what it costs
Reliability and price are tracked separately from judged quality. There's no speed chart because this release didn't measure response times.
Reliability
How often each provider successfully returned results, out of 216 test searches
Estimated cost per search
What one search roughly costs with each provider, in US dollars (list prices)
Top fact-recall scorers
Fact recall is the share of facts correctly recovered from the page; higher bars are better.
Fact recall by provider
Loading Extract data…
This is a separate lane from Search: Search measures finding the right pages, while Extract measures pulling the facts out.
These are rankings with confidence ranges — small differences may not be meaningful. Provider identities were hidden from the judges (blind inputs), but human spot-checks of the AI judges are still pending. We're not yet telling you to switch your provider: this is a first look, not a final verdict.
How to read these numbers →