A Good Lighthouse Score Did Not Imply Agent Readiness
Only 16 of 215 complete-case VCL project URLs with Lighthouse performance of at least 90 reached Agent Readiness L2. A focused analysis of the v9 overlap table.
*Updated October 25, 2026.*
Among 215 custom-domain project URLs with a Lighthouse performance score of at least 90 and all three recorded measurements, 16 also reached Agent Readiness L2 or higher: 7.4%. A Good performance score left the agent-readiness question largely unanswered in this VCL sample.
This is a new calculation from the published version 9 overlap table, not a new scan. The table contains 936 project URLs with recorded Domain Rating, Lighthouse performance and Agent Readiness at the October 1, 2026 cutoff.
Start with one threshold and check the others
The report’s headline counts the eight URLs meeting all three thresholds. A different question is useful when a builder already knows one result: what else did URLs with that result meet?
| Starting group within the 936 complete cases | URLs in group | Also met the first other threshold | Also met the second other threshold | Met both other thresholds |
|---|---|---|---|---|
| Lighthouse performance ≥90 | 215 | Ahrefs DR ≥10: 85 (39.5%) | Agent Readiness ≥L2: 16 (7.4%) | 8 (3.7%) |
| Ahrefs DR ≥10 | 447 | Lighthouse performance ≥90: 85 (19.0%) | Agent Readiness ≥L2: 64 (14.3%) | 8 (1.8%) |
| Agent Readiness ≥L2 | 106 | Ahrefs DR ≥10: 64 (60.4%) | Lighthouse performance ≥90: 16 (15.1%) | 8 (7.5%) |
Percentages use the starting group in that row. The rows overlap and must not be added together. These are descriptive proportions within the measured sample, not predictions for a new project.
The first row means 199 of the 215 URLs with Good Lighthouse performance were below L2 Agent Readiness. The second means 362 of the 447 URLs reaching DR 10 were below the Lighthouse Good threshold. Neither measure supplied a substitute for the other check.
How to reproduce the calculation
The source contains eight mutually exclusive combinations. The Lighthouse group is the sum of Lighthouse only (122), DR plus Lighthouse only (77), Lighthouse plus Agent Readiness only (8), and all three (8): 215.
Of those, the two rows meeting the agent threshold contain 8 plus 8 URLs: 16 / 215 × 100 = 7.4%, rounded to one decimal place.
The DR group is 306 + 77 + 56 + 8 = 447. Its Lighthouse overlap is 77 + 8 = 85, giving 19.0%. The Agent Readiness group is 34 + 56 + 8 + 8 = 106.
All calculations can be repeated from the version 9 source CSV. No unpublished project-level data is needed.
What each result actually describes
Lighthouse measures a submitted page in a desktop lab run. Agent Readiness is the provider’s level for that URL; L2 is the report’s chosen threshold beyond L0 and L1. Ahrefs DR describes backlink strength at the registrable-domain level, with 10 chosen as a low reference point.
Those are different scopes. Several project URLs can inherit one domain’s DR, and the measurements can come from different dates. The pattern does not establish statistical independence, causation or a product-quality ranking. Agent Readiness also does not measure an agent successfully completing an authenticated task.
The denominator is especially important: these 215 Lighthouse-Good URLs are part of the complete-case custom-domain comparison. The broader Lighthouse chart has 309 Good pages across 1,292 measured project URLs. Substituting 309 here would mix populations.
The practical consequence
A builder can use a strong score as evidence about the thing that was measured. It still leaves other questions open. For a launch review, record page performance, domain visibility and agent access separately; then test the actual product task a person or agent needs to complete.
Data and source: VCL Research, State of Vibe Coding 2026, version 9, cutoff October 1, 2026 at 00:00 UTC. The complete-case comparison covers 936 of 1,020 eligible custom-domain project URLs. The 84 without all three measurements are excluded, not classified as failures. The cohort describes VCL listings, not all AI-built products. See the methodology.