Subdomains Are Leverage, Not Duplication

Somebody trying to decide whether to keep running Power 9 past its end-of-life date is not asking the same question as somebody trying to figure out if Power 12 will even support their current FlashSystem media. Somebody pricing out IBM i software licensing isn't asking the same question as somebody pricing out the hardware underneath it. Somebody comparing IBM against Dell and HPE for a mixed-vendor RFP isn't reading either of those pages at all.

One generalist site trying to answer all of that on a handful of pages doesn't produce focus. It produces a page arguing with itself, Power 7 migration urgency sitting next to Power 12 roadmap speculation, hardware pricing bleeding into software licensing, because there was nowhere else for any of it to live.

So instead, each piece got its own room. Power 9 got dedicated lifecycle, sourcing, and migration content once the end-of-life articles showed real demand for it. The AI subdomain launched to cover IBM's AI-on-IBM-i push specifically, and caught itself inventing a feature that doesn't exist (something called 'Db2 for i AI system Learning') before it shipped, replacing it with IBM's actual flagship product instead. FlashSystem got its own subdomain, then got substantially deeper eleven days later once IBM shipped a new hardware generation and a competitor's own ranking gaps showed exactly which model pages were missing. Pricing split into two subdomains on purpose, hardware economics and software economics don't share a buyer's mental model, so they don't share a page. Security shipped only once the actual infrastructure, subdomain and SSL both, was really provisioned, not the moment the content was ready.

Even the two oldest, quietest properties in the family, Power 5 and Power 6, only got formally wired into the connective structure, the constellation, the cross-links, the portfolio listing, in the last few weeks. They existed before that. They just weren't pulling any weight in the network yet.

The first warehouse pull, through August 21, showed 81,615 impressions and 268 clicks across the 18 campaign properties. The traffic was uneven, which was expected. The three flagship sites carried roughly two-thirds of the impressions while the new specialist sites were still earning their first visibility.

A separate warehouse audit then fixed the missing subdomain log ingestion. The bot zeros were not quiet sites. They were missing log rows. Once the pipeline could see the cPanel subdomain logs, the AS400System family showed exactly what the server had been showing all along: search crawlers and AI crawlers were moving through the network.

Google's global August 2026 spam update finished on August 21. On August 22, every established property under as400system.com fell at the same time.

The family went from 9,278 impressions during August 15 through 21 to 1,322 during August 22 through 28. That is an 85.8 percent decline. Clicks went from 60 to 1. The root domain fell 80.1 percent while its average position moved from 50.5 to 66.4. The number of root pages receiving impressions fell from 117 to 48.

The individual subdomains dropped between 83.2 and 97.9 percent. AS400Software.com and AS400IBMSystems.com did not follow them down. Googlebot requests across the affected family increased from 938 to 984. Search Console still showed 683 indexed pages after the drop, with no manual action and no security issue.

That combination matters. This was not a server outage, deindexing event, demand collapse, or broken sitemap. It was a synchronized ranking reclassification across one domain family, beginning the day after the spam update completed. The timing is evidence. It is not proof of which individual signal Google disliked. Google's incident record confirms the update window.

The obvious reaction is to start undoing the architecture until Google looks happier. I am not doing that.

These subdomains were not built to disguise duplicate pages. They were built because Power 7 history, Power 9 lifecycle planning, Power 11 buying decisions, FlashSystem hardware, IBM i security, AI modernization, and pricing economics are different jobs. The network still has those jobs after a ranking system changes its opinion.

Google is one discovery channel. It is not the board of directors for the portfolio. The same properties are still indexed, still receiving Googlebot, still being crawled by OpenAI, Anthropic, Amazon, ByteDance, Perplexity, and other AI readers, and still building a structured record of the IBM Power market.

So the architecture stays. Planned subdomains keep shipping when they have a real role. The measurement changes from 'did Google reward this immediately' to 'does a legitimate, differentiated network regain visibility as its history, links, citations, and human use accumulate.' That is a better long-term experiment anyway.