Tone Is a Lever: The BigTechReseller Competitor-Seeding Campaign

The starting brief was a documented handoff: seed two new vendors (Everpure, Nutanix), three new categories (AI Infrastructure, Hybrid Cloud, Cyber Resilience), and five comparison articles, closing every explicit gap in a competitor-seeding plan. That part shipped clean in two rounds. What happened next is the actual case study.

Rounds 1-2 delivered the seeding brief as specified. Round 3 began after Krisada reviewed the live pages himself and called them 'vanilla slop,' generic vendor copy with no sales psychology behind it. That feedback triggered a rebuild, not a rewrite: related-article and highlight cross-linking got built directly into bootstrap.php and the vendor and product templates, and all nine vendor descriptions, not just the two new ones, got rewritten with actual sales-psychology specificity. Rounds 4-5 punched up opening lines with real energy and added a boxed excerpt treatment using the site's existing orange brand color, reusing an established callout pattern instead of inventing a new one. Round 6 built a real site_autolink_glossary() function and ten new glossary terms so vendor jargon linked to real definitions automatically, site-wide. Round 7 wired that same autolinking into article bodies, and that's where the campaign stopped being a content project and started being a bug hunt: the autolinking pass surfaced a glossary-accuracy error where IBM Power Systems and AS400 had been treated as synonyms when they're related but distinct, every genuine page-title H1 got recolored orange site-wide, and fixing that heading color surfaced a real semantic-HTML defect in vendor.php, the actual h1 tag was holding the small muted subtitle instead of the visually prominent vendor name, backwards from what a screen reader or a search engine would expect. SAP and Oracle's descriptions, still generic from the original seed, got rewritten to match the other seven vendors, and the whole standard was written down as docs/BigTechReseller-Editorial-Voice-Guide.md. Round 8 made the standard durable by adding a mandatory read-first rule to the site's hardlinked CLAUDE.md and AGENTS.md files, so any future AI author, Claude or Codex, picks it up automatically instead of needing the same correction repeated.

None of the three real defects, the heading-hierarchy CSS bug affecting every vendor page site-wide, the glossary alias conflating two related-but-distinct IBM product terms, and the semantic-HTML mismatch in vendor.php, were flagged by an automated check. All three surfaced because Krisada looked closely at the live pages and said something felt off, first about the writing, then, once the writing pass touched headings and links, about the structure underneath it.

That sequencing matters. A purely additive build, seed the vendors, ship the categories, call it done, would have shipped all three defects silently and moved on. The voice correction is what put a person back in front of the live pages long enough to notice the second and third problems, which had nothing to do with tone at all.

Every round validated cleanly on its own: JSON decode and PHP lint before each deploy, local rendering for any shared template or CSS change, a live curl sweep after deploy culminating in a 29-URL full-site sweep by round 7. That verification caught zero defects. A person reading the actual output caught three.