Editorial review · 260529-004
How XCHO’s piece on Sundar says they love it. The install charts say otherwise. scored.
Read the article →Solid reporting. Some issues but credible overall. The reader is well-served.
Accuracy
Headline statistics on installs and noai.duckduckgo.com visits are attributed to TechCrunch and PC Gamer reporting from late May 2026, which sits past my verification cutoff (-3 minor, post-cutoff source attributed). Google Q1 2026 search revenue growth of 19% is asserted without citation in the sources block (-5). Specific model names routed by Duck.ai (Claude Haiku 4.5, GPT-5 mini, Llama 4 Scout) are unsourced specifics (-5).
Balance
The piece is opinion but seriously engages Google's counter-argument, granting that Q1 revenue is the strongest rebuttal and offering three competing readings of the data. It names the vocal-minority and Hacker News amplification risk against its own thesis. Loaded framing ("force-feeding", "enshittification" in glossary) leans against Google without equivalent treatment, but the body argues fairly (-5).
Concerns (4)
- minoraccuracy
“Google's Q1 2026 search revenue grew 19% year-over-year”
Specific figure asserted with no citation in sources or footnotes.
Evidence: No Alphabet earnings link provided; figure is load-bearing for the counter-argument.
- minoraccuracy
“Duck.ai routes users to Claude Haiku 4.5, GPT-5 mini and Llama 4 Scout”
Specific model versions stated without source.
Evidence: No DuckDuckGo documentation or Duck.ai page cited for the routing claim.
- minoraccuracy
“US app installs +30.5% on 25 May; iOS peak +69.9%”
Post-cutoff, source attributed to TechCrunch and DuckDuckGo.
Evidence: Cannot verify against live sources from this review surface.
- minorbalance
“(glossary entry for enshittification)”
Loaded framing in glossary applied only to Google's side of the argument.
Evidence: No equivalent critical term defined for DuckDuckGo's marketing posture.
Reproducibility
How this review works: read the methodology. Each published Dispatch is scored by a single primary reviewer (Claude Opus 4.7) against the public rubric. A second model (Gemini 2.5 Pro with Google Search) runs the same prompt as a variance signal and is shown above only when the two scores diverge by more than ten points.