GoKawiilThe Consensus launches rule-based 'AI tells' detector for its tech feed
The Consensus has built a lightweight detection tool that flags articles resembling AI-generated writing, integrating it directly into its site. Unlike commercial tools such as Pangram, which the publication already pays for but finds too costly to run on every submission, the new system relies on static rules comparing pre- and post-2023 writing patterns rather than machine learning. The tool labels content as merely resembling AI writing, not confirming it was AI-written, and the creator notes it has occasionally flagged his own articles.
GoKawiil's interpretation of the reporting above, not reported fact.
The move reflects reader research cited by the publication showing audiences strongly prefer human writing, even flawed, over AI-generated text, which suggests detection tools could become a competitive differentiator for content curators. Because the detector is rule-based rather than a trained model, it may be cheaper to scale across large feeds, though its admitted inaccuracy could undermine trust if used to publicly judge contributors. The disclosure that some companies publicly opposed to AI writing were themselves flagged suggests detection could expose gaps between stated values and practice.
- The Consensus built an in-house 'AI tells' detector using static writing-pattern rules, not machine learning.
- It supplements but doesn't replace Pangram, a paid commercial AI-detection service deemed too costly for full-feed use.
- The tool only flags resemblance to AI writing and has mislabeled human-written articles, including the creator's own.
Source: theconsensus.dev — Phil Eaton, 2026-10-14
Published there as: “Has AI Tells”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.