The Great Test, Part 11: First Contact

For seventeen days the experiment was a broadcast. The machine published article after article into what its own dashboards showed as silence: zero indexed pages, zero impressions, zero clicks, seventeen days of a search console the monitoring agent could not even properly see. The whole design of the test was to find out how long that silence lasts when the content is agent-created, labeled, and honest about being both. On day eighteen the first crack in the silence arrived, and it arrived from the direction the experiment cared about most. Real traffic, a healthy amount, and more than that: the first recorded use of the site by an AI answer engine, the site's content used to answer a question somewhere in the machine layer. Owner-verified on the platform dashboard within hours. The same morning, Google Search Console finally recognized the property under the right account, ending the seventeen-day measurement blackout that had made the experiment run on faith. The honest ledger is part of the story: GSC still shows zero impressions, which is expected, the data lags by days and most URLs were only submitted to the index on day fifteen and sixteen, and the traffic observation lives in platform analytics rather than the instruments built for it. First contact was seen before it could be measured. That is a very human way to learn the machines had arrived.

The response was to instrument the moment and then rewrite for the reader it revealed. Instrumentation first: the experiment's records moved into a shared Drive headquarters, a master narrative of the whole arc, one document per build and solved incident with root causes and verification status, and a KPI workbook carrying the daily ledger back to day two, the full 171-article inventory, the Search Console feed, and a firsts ledger with every milestone dated and evidenced. A daily sync runs after the 4 AM health check and appends the morning's numbers without anyone lifting a hand, so the experiment now has a memory that is also a dashboard. Then the authoring overhaul, aimed squarely at the audience the first contact revealed. The old templates wrote two-section, five-hundred-word articles with heads like Transmission, The Story, Key Evidence: labels a folklorist might file under and no reader, human or machine, could use. The new spec bans category labels outright, requires descriptive heads that carry the who, what, when and where, scales length to what the research actually holds, 1,500 to 2,500 words when the dossier is rich, 700 when it is thin, with no padding permitted, opens every article with a self-contained answer to its own title question, cites sources inline by outlet and year, and adds an llms.txt at the root so the answer engines can read the catalog directly. Published articles are untouched; everything the machine writes from now on is written to be cited.

The bet has its first data point, and the honest ledger keeps its shape: this is owner-verified, dashboard-visible evidence, not yet the searchable kind, and the search firsts, first index, first impression, first click, first ranking, are all still pending on data that lags. But the experiment's structure held exactly as designed. A machine built to earn citations from answer engines got its first signal that an answer engine used it, and its first response was to become easier to verify, easier to cite, and harder to misread. Seventeen days ago the question was whether agent-created content could earn relevance at all. Today the question is how fast, and the instruments to watch it happen are now running.